The DGX Spark platform is familiar territory for us at this point. We’ve reviewed the Dell, ASUS, Acer, and Gigabyte takes on NVIDIA’s GB10 Grace Blackwell reference design, and the core ingredients are consistent across all of them: 1,000 TOPS of FP4 compute, 128GB of unified LPDDR5x memory, and dual 200GbE networking in a 150mm chassis. HP’s ZGX Nano G1n AI Station builds on that foundation, but the way HP has built around it sets this unit apart from the rest of the Spark field.

The most visible differences are in materials and construction. HP wraps the ZGX Nano in a chassis built from up to 75% recycled aluminum and 20% recycled steel, with packaging that carries up to 93% recycled content. The internal layout splits the chassis into upper and lower halves, making it easier to access components like the SSD and coin-cell battery than on several of the Spark units we’ve tested. Thermally, HP rates the system at 22 dBA idle and 27.6 dBA under intensive workloads, quiet for a system dissipating approximately 780 BTU/hr at peak.

Security is where HP pushes furthest past the reference platform. The ZGX Nano ships with TPM 2.0 operating in FIPS 140-2 certified mode, meets Common Criteria EAL4+, and includes BIOS-level secure boot and PXE controls. Storage is factory-installed as a self-encrypting OPAL NVMe drive. Taken together, HP is positioning this unit not only as a developer desk-side AI node but also as a system that can operate within regulated environments where supply chain certifications, encryption at rest, and tamper resistance matter for procurement.

Specification	HP ZGX Nano G1n AI Station
Overview
Product Name	HP ZGX Nano G1n AI Station
Form Factor	Mini
Operating System	NVIDIA DGX OS 7 / Ubuntu 24.04 NOTE: This product does not support Microsoft Windows.
Hardware
Processor	NVIDIA GB10 Grace Blackwell Superchip Blackwell Architecture GPU 20-core Arm CPU (10x Cortex-X925 + 10x Cortex-A725) Blackwell CUDA Cores 5th Gen Tensor Cores 4th Gen RT Cores 1x NVENC 1x NVDEC
Memory	128GB LPDDR5x, unified, 16 channels, soldered
Memory Bandwidth	273 GB/s
Storage (Internal I/O)	1x M.2 PCIe Gen5 x4 Options: 2TB or 4TB PCIe Gen4 x4 NVMe (2242, SED OPAL TLC)
Networking & I/O
Rear I/O Ports	1x USB-C power (240W) 3x USB-C 20Gbps (DisplayPort 1.4a, 30W total) 1x HDMI 2.1a 1x 10GbE RJ-45 2x QSFP 200GbE (ConnectX-7)
Network Controllers	Realtek RTL8127-CG 10GbE NVIDIA ConnectX-7 200GbE
WLAN & Bluetooth	AzureWave AW-EM637 Wi-Fi 7 + Bluetooth 5.4
Performance
AI Compute	Up to 1,000 TOPS (FP4)
Model Capacity	Up to 200B parameters
Physical & Power
Dimensions (H x W x D)	2.01″ (no feet) / 2.1″ (with feet) 5.9″ x 5.9″
Weight	Starting at 1.25kg (2.76 lbs)
Power Supply	240W USB-C external adapter, 89% efficiency, active PFC

Build and Design

The HP ZGX Nano G1n takes a noticeably different approach to the DGX Spark design compared with the other systems we have looked at so far (see our Dell/ASUS/Acer/Gigabyte reviews). Instead of the more common build, where the internals feel tucked into a top cover, HP splits the chassis into upper and lower halves, making the internal layout easier to understand once inside. What first appears more complicated turns out to be fairly practical, with straightforward access to parts like the coin-cell battery and SSD after removing just a handful of screws. That more considered internal structure also carries over to the outer build, where HP places greater emphasis on how the system is constructed and the materials used throughout.

That said, HP wraps it in a sleek black case with a 150mm-square footprint and relies heavily on recycled materials. Specifically, the build uses up to 75% recycled aluminum, 20% recycled steel, and significant amounts of post-consumer recycled plastics. Even the packaging reflects this commitment. Corrugated materials contain up to 93% recycled content, and plastic packaging incorporates at least 30% recycled content.

Thermally, the system relies on forced-air cooling. This is a notable engineering choice given the density of the NVIDIA GB10 Grace Blackwell Superchip. Despite its compact footprint, HP specifies a full thermal envelope. Under maximum load, the system dissipates up to approximately 780 BTU/hr, depending on configuration. Peak system power draw reaches approximately 228W. Furthermore, HP advertises relatively low noise levels, rated at 22 dBA at idle and 27.6 dBA under intensive workloads.

Physically, the unit measures 5.9 x 5.9 x 2.01 inches without feet, firmly placing it in ultra-compact territory. HP explicitly states that the unit is not rack-mountable, reinforcing its role as a desk-side AI node rather than traditional data center infrastructure. Serviceability is minimal by design. Users need a #1 Phillips screwdriver to access internal components, and most components, including memory, are non-user-replaceable.

Internally, the ZGX Nano uses NVIDIA’s reference board design, as do many other OEMs building on the DGX Spark platform. The LPDDR5x memory is soldered directly to the board and runs at up to 8533 MHz. Overall, the platform prioritizes efficiency and density over modularity.

Security and Upgradability

HP locks down the ZGX Nano G1n by design. It features an integrated TPM 2.0 module that operates in FIPS 140-2-certified mode, meets Trusted Computing Group specifications, and is Common Criteria EAL4+ certified. BIOS-level protections include secure boot controls, PXE-based remote boot capabilities, and the ability to disable boot from removable media entirely.

From a hardware standpoint, HP is explicit: this system is not upgradeable. The 128GB of LPDDR5x unified memory sits soldered directly to the board. Additionally, buyers must select storage at the time of purchase. While the single M.2 slot supports PCIe Gen5 x4 electrically, factory configurations ship with PCIe Gen4 x4 NVMe SSDs. These come in 2TB or 4TB capacities and are all self-encrypting OPAL drives.

HP notes that spare parts will remain available for up to five years after production ends. Nevertheless, this is fundamentally an appliance-style system rather than a modular workstation.

I/O and Expansion

The front of the unit is minimalist, featuring only a power button and a status LED. On the back, the system offers a dense array of high-performance connectivity options. HP delivers power via a standard NVIDIA-recommended 240W USB-C adapter and warns that third-party adapters may cause degraded performance or instability.

Three USB 3.2 Type-C ports provide USB connectivity, each operating at 20 Gbps and supporting DisplayPort 1.4a Alt Mode. A dedicated HDMI 2.1a port provides additional display output. For networking, the system includes both a Realtek RTL8127-CG 10GbE controller and an NVIDIA ConnectX-7 controller, providing dual 200GbE QSFP112 ports, each with 200 Gbps throughput.

The networking stack supports a wide range of enterprise features. These include PXE boot, Wake-on-LAN, VLAN tagging (802.1Q), time synchronization (802.1as/1588), and full-duplex operation across all supported speeds. Additionally, a Wi-Fi 7 (802.11be) 2×2 module with Bluetooth 5.4 provides wireless connectivity and supports MU-MIMO, WPA3 security, and operation across the 2.4GHz, 5GHz, and 6GHz bands.

Graphics and Audio

The integrated NVIDIA Blackwell GPU in the GB10 Superchip handles all graphics tasks. The system supports up to 8K output at 60Hz via USB-C DisplayPort 1.4a and 8K at 30Hz via HDMI 2.1a. HP recommends using direct cable connections for 8K output, as adapters or docks may cause instability or degrade signal quality.

Audio runs over HDMI, with no dedicated analog audio outputs. This aligns with the system’s positioning as a compute node rather than a traditional multimedia workstation.

Thermals Testing

CPU Temperature

During CPU thermal testing, the HP ZGX Nano G1n reached a peak temperature of 77.3°C during the workload’s more intense bursts. This places HP below the hottest systems in the comparison stack during peak transitions, as other units climbed into the 90°C range. As the workload transitioned into Equal ISL/OSL and then Decode Heavy, CPU temperatures stabilized rather than continuing to rise sharply.

At the lower end, the CPU recorded a minimum temperature of 36.4°C during light-load conditions. This means the HP has effective heat dissipation when the system is not under heavier computational stress. Overall, the ZGX demonstrated controlled burst CPU thermal behavior with stable sustained-load performance.

 

GPU Temperature

GPU thermals followed a similar pattern. During periods of heavy acceleration, the GPU reached a maximum temperature of 69°C. This positions HP on the cooler side of the comparables during peak burst conditions, with several other systems (like the Dell, ASUS, and Founders Edition) running noticeably warmer at the top end. As activity shifted into Equal ISL/OSL and Decode Heavy phases, GPU temperatures leveled off and remained stable.

The GPU recorded a minimum temperature of 34°C during lighter phases, indicating solid idle thermal capabilities.

NVMe Temperature

During the Equal phase, the NVMe drive reached roughly 42°C, showing only a gradual rise from its resting baseline. As the workload shifted to Prefill Heavy, the storage temperature rose noticeably, ranging from 42°C to 47°C. In Decode Heavy, the drive operated in its warmest range, 47°C to 54°C, where it peaked, yet remained noticeably below most other Spark systems.

NIC Temperature

During the Equal phase, NIC temperature ranged from 39°C to 52°C, showing a steady climb, indicating moderate thermal buildup as network activity ramps up early in the run.

In Prefill Heavy, NIC thermals increased, ranging from 48°C to 64°C, because this phase places much more sustained pressure on the networking subsystem. During Decode Heavy, NIC temperature was in its warmest range, 52°C to 68°C, where the peak was reached. Nonetheless, thermal behavior remained stable throughout the test.

GPU Power Consumption

During the Equal phase, GPU power consumption ranged from 2.86W to just over 40W, placing the HP ZGX Nano G1n in the middle of the pack.

In Prefill Heavy, GPU power started at roughly 37W, dipped to as low as 35W, and spiked to as high as 69W, making this the most power-intensive phase of the run.

During Decode Heavy, GPU power consumption settled into a lower, more stable range of 35W to 46W, indicating that power demand eased as the workload shifted away from the more aggressive burst behavior.

Thermal Summary

Under load, the ZGX Nano G1n operates within a tightly controlled thermal envelope. Maximum system power consumption is approximately 228W, and heat dissipation is approximately 780 BTU/hr. By contrast, idle power draw remains low at approximately 36–38W, which indicates efficient power scaling when the system is not active. The forced-air cooling solution maintains stable operation within HP’s specified range of 5°C to 30°C.

HP ZGX Nano AI Performance Testing

To evaluate the HP ZGX Nano with GB10, we tested Spark units using the vLLM Online Serving benchmark, the most widely adopted high-throughput inference and serving engine for large language models. The vLLM online serving benchmark simulates real-world production workloads by sending concurrent requests to a running vLLM server and measuring key metrics, including total token throughput (tokens per second), time to first token, and time per output token, across varying load conditions.

Our testing spanned a range of models, including dense architectures and micro-scaling data types, and evaluated performance across three workload scenarios: Equal ISL/OSL, Prefill Heavy, and Decode Heavy. These scenarios represent distinct real-world serving patterns, from balanced input and output loads to compute-intensive prompt processing and memory-bandwidth-bound token generation.

In addition to the HP ZGX Nano with GB10, we benchmarked other OEM systems from Dell, ASUS, Acer, and Gigabyte. This allowed us to place HP’s results within the broader competitive landscape and understand where it leads, keeps pace with the pack, or trails across different models and workloads.

GPT-OSS-120B

With GPT-OSS-120B, the HP ZGX Nano G1n posts its strongest results in Prefill Heavy, where throughput climbs from 304.5 tok/s at batch 1 to 2773.3 tok/s at batch 64. Equal ISL/OSL also scales steadily, rising from 69.6 tok/s to 722.9 tok/s across the sweep. Decode Heavy is much lighter by comparison, starting at 183.7 tok/s in batch 1, dipping slightly in batch 2, then recovering to 262.9 tok/s by batch 64.

 

GPT-OSS-20B

With GPT-OSS-20B, HP’s highest numbers come from Prefill Heavy, but the scaling is less linear than with the other models. Prefill starts at 1626.6 tok/s at batch 1, climbs to 1980.3 tok/s at batch 2, drops sharply to 1120.3 tok/s at batch 4, then recovers to 4345.1 tok/s by batch 64. Equal ISL/OSL scales more smoothly from 92.6 tok/s to 1550.6 tok/s, and Decode Heavy rises from 94.4 tok/s to 670.4 tok/s.

Qwen3 Coder 30B A3B FP8

For Qwen3 Coder 30B A3B (FP8), HP again excels in Prefill Heavy, with throughput increasing from 432.2 tok/s at batch size 1 to 2069.4 tok/s at batch size 64. Equal ISL/OSL rises from 104.2 tok/s to 1274.4 tok/s, while Decode Heavy improves from 55.9 tok/s to 480.4 tok/s. This is among HP’s stronger overall results.

Qwen3 Coder 30B A3B Base

On Qwen3 Coder 30B A3B (Base), HP delivers steady growth across all three phases, although the topline remains in the Prefill Heavy phase. That phase increases from 258.6 tok/s at batch 1 to 1629.4 tok/s at batch 64. Equal ISL/OSL scales from 60.3 tok/s to 690.3 tok/s, while Decode Heavy rises from 33.0 tok/s to 331.8 tok/s.

Llama 3.1 8B Instruct FP4

With Llama-3.1-8B-Instruct (FP4), HP shows a clear step up in throughput. Equal ISL/OSL climbs from 76.4 tok/s at batch 1 to 2774.1 tok/s at batch 64, making it the strongest of HP’s three phases on this model. Prefill Heavy also scales aggressively, rising from 316.8 tok/s to 2397.1 tok/s at batch 32 before slipping to 2270.4 tok/s at batch 64. Decode Heavy increases from 40.7 tok/s to 547.6 tok/s across the sweep.

Llama 3.1 8B Instruct (Base)

On Llama-3.1-8B-Instruct (Base), the HP ZGX Nano G1n scales cleanly across all three phases. In Equal ISL/OSL, throughput rises from 28.2 tok/s at batch 1 to 1298.6 tok/s at batch 64. In Prefill Heavy, HP increases from 123.2 tok/s to 1759.5 tok/s, with gains remaining strong throughout the sweep before tapering slightly at the top end. Decode Heavy is much lighter by comparison, rising from 15.5 tok/s at batch 1 to 366.4 tok/s at batch 64.

GPU Direct Storage

How GPU Direct Storage Works

Traditionally, when a GPU processes data from an NVMe drive, the data must first pass through the CPU and system memory before reaching the GPU. This process creates bottlenecks because the CPU acts as a middleman, adding latency and consuming system resources. GPU Direct Storage eliminates this inefficiency by allowing the GPU to access data directly from the storage device over the PCIe bus. This direct path reduces data movement overhead, enabling faster, more efficient transfers.

AI workloads, especially those involving deep learning, are highly data-intensive. Training large neural networks requires processing terabytes of data, and any delay in data transfer leads to underutilized GPUs and longer training times. Accordingly, GPU Direct Storage addresses this challenge by delivering data to the GPU as quickly as possible, minimizing idle time and maximizing computational efficiency.

In addition, GDS benefits workloads that stream large datasets, such as video processing, natural language processing, and real-time inference. By reducing CPU reliance, GDS accelerates data movement and frees CPU resources for other tasks, further enhancing overall system performance.

GDSIO Read Throughput 16K

Looking at GDSIO Read Throughput 16K, the HP ZGX Nano G1n starts at 0.70GiB/s with 1 thread, placing it among the stronger low-thread performers in the group. It dips to 0.41GiB/s at 2 threads, then climbs back to 0.86GiB/s at 4 threads, showing the same small early-thread inconsistency seen in a few of these systems. From there, scaling becomes much more consistent. Throughput rises to 1.6GiB/s at 8 threads and 2.2GiB/s at 16 threads, then continues upward to 3.0GiB/s at 32 threads. At the higher queue depths, the HP keeps gaining ground, reaching 3.9GiB/s at 64 threads and peaking at 4.6GiB/s at 128 threads.

GDSIO Read Average Latency 16K

Looking at GDSIO Read Average Latency (16K), the HP ZGX Nano G1n starts at approximately 0.02ms with 1 thread and remains low through 2 threads (0.08ms) and 4 threads (0.07ms). Latency edges up slightly at 8 threads (0.08ms) and 16 threads (0.11ms), then increases more noticeably at 32 threads (0.16ms) and 64 threads (0.25ms). At 128 threads, latency reaches 0.42ms, still a bit below the highest results in the group while tracking the system’s steady throughput scaling across the test.

GDSIO Write Throughput 16K

Looking at GDSIO Write Throughput 16K, the HP ZGX Nano G1n starts at 0.84GiB/s on 1 thread, rises to 1.4GiB/s on 2 threads, and reaches 2.2GiB/s on 4 threads. Performance continues to scale strongly at 8 threads (3.0 GiB/s) and reaches 3.3GiB/s at 16 threads, where it effectively levels off. From there, throughput remains nearly flat at 3.3GiB/s with 32 and 64 threads, then eases slightly to 3.2GiB/s with 128 threads, indicating the platform reaches its write ceiling relatively early and sustains that level consistently through the rest of the sweep.

GDSIO Write Average Latency 16K

Looking at GDSIO Write Average Latency (16K), the HP ZGX Nano G1n starts at approximately 0.02ms with 1 thread and remains very low through 2 threads (0.02ms) and 4 threads (0.03ms). Latency rises modestly at 8 threads (0.04ms) and 16 threads (0.07ms), then jumps at 32 threads (0.15ms) and 64 threads (0.30ms). At 128 threads, latency reaches 0.61ms, still fairly well controlled overall, though the upward trend aligns with the point where write throughput has already flattened at higher thread counts.

GDSIO Read Throughput 1M

Looking at GDSIO Read Throughput 1M, the HP ZGX Nano G1n starts at 3.2GiB/s on 1 thread and rises to 4.1GiB/s on 2 threads. Performance continues to climb at 4 threads (5.2GiB/s) and 8 threads (5.5GiB/s), after which the platform effectively reaches its ceiling. Throughput then holds essentially flat at 5.5GiB/s for 16, 32, and 64 threads, before easing slightly to 5.3 GiB/s at 128 threads, indicating a strong early ramp followed by a very stable high-thread plateau.

GDSIO Read Average Latency 1M

Looking at GDSIO Read Average Latency (1M), the HP ZGX Nano G1n starts at approximately 0.31ms with 1 thread and remains relatively low at 2 threads (0.47ms) and 4 threads (0.76ms). Latency increases with concurrency, rising to 1.4ms at 8 threads, 2.9ms at 16 threads, and 5.9ms at 32 threads. The trend continues at 64 threads (12.8ms) and reaches 27.2ms at 128 threads, tracking the higher queue depths even though throughput had already flattened much earlier in the sweep.

GDSIO Write Throughput 1M

Looking at GDSIO Write Throughput 1M, the HP ZGX Nano G1n starts at 3.1GiB/s with 1 thread and rises to 3.5GiB/s with 2 threads, then holds that level at 4, 8, and 16 threads. Performance dips slightly to 3.3GiB/s at 32 threads before returning to 3.5GiB/s at 64 threads. At 128 threads, throughput increases to 3.7GiB/s, indicating a mostly flat write profile across the sweep with only minor variation and a small uptick at the highest thread count.

GDSIO Write Average Latency 1M

Looking at GDSIO Write Average Latency (1M), the HP ZGX Nano G1n starts at approximately 0.31ms with 1 thread, rising to 0.57ms with 2 threads and 1.1ms with 4 threads. Latency continues to climb as concurrency increases, reaching 2.2ms with 8 threads, 4.4ms with 16 threads, and 9.4ms with 32 threads. The upward trend continues at 64 threads (17.7ms) and reaches 37.3ms at 128 threads, reflecting steadily increasing queue pressure even though write throughput itself remains fairly flat through most of the sweep.

Conclusion

HP’s ZGX Nano G1n carries the DGX Spark platform’s expected performance profile and adds engineering choices that set it apart from the other Spark systems in the field. In our testing, CPU temperatures peaked at 77.3°C and GPU temperatures at 69°C, both on the cooler side of the Spark units we’ve benchmarked. vLLM performance was strongest in Prefill Heavy workloads across all six models we tested, with scaling that held cleanly through higher batch sizes. GPU Direct Storage read throughput reached 4.6 GiB/s at 16K and 5.5 GiB/s at 1M block sizes, and write throughput plateaued early but held that level consistently across the remaining thread counts.

Where the ZGX Nano G1n separates itself from the rest of the Spark field is in the work HP did around the reference design. The recycled-materials content, the upper/lower-chassis split that improves internal serviceability, and the acoustic envelope that holds at 27.6 dBA under load all reflect deliberate engineering choices beyond what the GB10 platform itself requires. The security stack follows the same pattern. TPM 2.0 in FIPS 140-2 mode, Common Criteria EAL4+, and SED OPAL storage push this unit past a developer appliance and toward a system that can clear procurement in regulated environments.

Like other Sparks, this is not a general-purpose workstation, and HP does not position it as one. For developers, small teams, and organizations that need local AI compute with credible sustainability and security stories behind the purchase, the ZGX Nano G1n is a clear differentiated option within the Spark lineup. For shops where those criteria do not apply, the underlying platform is the constant across all five OEM systems we’ve reviewed, and the decision comes down to ecosystem, support, and price.

Product Page – HP ZGX Nano G1n AI

The post HP ZGX Nano G1n AI Station Review: A Secure, Sustainable Desk-Side AI Node appeared first on StorageReview.com.

NVIDIA and Google Cloud used Google Cloud Next in Las Vegas to outline a new phase of their long-standing engineering partnership, introducing updates to the Google Cloud AI Hypercomputer platform to scale agentic and physical AI for production environments. The companies continue to co-design infrastructure spanning silicon, systems, networking, and software to support increasingly complex AI workloads, including autonomous agents, robotics, and digital twins.

Vera Rubin-Based A5X Infrastructure Targets Large-Scale AI Factories

Google Cloud introduced A5X bare-metal instances built on NVIDIA Vera Rubin NVL72 rack-scale systems. These systems are designed to significantly improve inference economics and efficiency, delivering up to 10x lower cost per token and 10x higher token throughput per megawatt than the prior generation.

The A5X platform integrates NVIDIA ConnectX-9 SuperNICs with Google’s next-generation Virgo networking stack. This architecture enables cluster scaling to 80,000 Rubin GPUs within a single site and to 960,000 GPUs across multi-site deployments. The design targets hyperscale AI training and inference environments where network performance and system-level optimization are critical.

Google Cloud emphasized that tightly integrated infrastructure and managed AI services are required to support the next wave of AI workloads. The combined stack enables customers to train, fine-tune, and deploy models with an emphasis on performance, efficiency, and operational scalability.

Broad Blackwell Portfolio Enables Right-Sized Acceleration

Google Cloud also outlined its portfolio of NVIDIA Blackwell-based instances, spanning a wide range of deployment sizes and performance profiles. Offerings include A4 VMs based on NVIDIA HGX B200 systems, A4X and A4X Max configurations built on the GB200 and GB300 NVL72 platforms, and fractional GPU access via G4 instances with RTX PRO 6000 Blackwell Server Edition GPUs.

This range allows organizations to align infrastructure with workload requirements. Configurations range from fractional GPUs for lighter inference tasks to full NVL72 racks with 72 GPUs interconnected via fifth-generation NVLink and NVLink Switch technology. At the high end, deployments can scale to tens of thousands of GPUs for large-model training and distributed inference.

These systems are designed to support a range of AI workloads, including mixture-of-experts (MoE) models, multimodal inference, large-scale data processing, and simulation workloads for robotics and physical AI.

Early adopters are already leveraging the platform. Thinking Machines Lab is using GB300 NVL72-based A4X Max instances to scale training for its Tinker API, while OpenAI is running large-scale inference workloads, including ChatGPT, on GB200 and GB300-based instances on Google Cloud.

Confidential AI Extends to Blackwell GPUs

Google Cloud is extending confidential computing capabilities to its AI infrastructure. Gemini models running on NVIDIA Blackwell and Blackwell Ultra GPUs are now available in preview on Google Distributed Cloud, enabling organizations to deploy models closer to sensitive data sources.

NVIDIA Confidential Computing enables encrypted execution environments in which prompts and fine-tuning data remain protected from unauthorized access, including by cloud operators. This capability is also coming to multi-tenant environments through Confidential G4 VMs with RTX PRO 6000 Blackwell GPUs.

This marks the first confidential computing implementation for Blackwell GPUs in the public cloud, targeting regulated industries that require strict data protection while maintaining access to high-performance AI infrastructure.

Open Models and Managed RL Pipelines for Agentic AI

The platform supports a broad model ecosystem, including Google’s Gemini and Gemma models and NVIDIA’s Nemotron open models. NVIDIA Nemotron 3 Super is now integrated with the Gemini Enterprise Agent Platform, enabling developers to build and deploy reasoning-driven agentic workflows.

Google Cloud is also introducing Managed Training Clusters with a reinforcement learning API built on NVIDIA NeMo. This service automates cluster provisioning, job orchestration, and fault handling, enabling large-scale RL training. The goal is to reduce operational complexity and allow teams to focus on model behavior and optimization.

CrowdStrike uses NVIDIA NeMo tools, including Data Designer, Automodel, and Megatron Bridge, to generate synthetic data and fine-tune domain-specific cybersecurity models. These workflows run on Blackwell-based infrastructure, accelerating threat detection and response pipelines.

Expanding Industrial and Physical AI Workloads

The joint platform also targets industrial and physical AI use cases. Applications from Cadence and Siemens Digital Industries Software are now available on Google Cloud with NVIDIA acceleration, supporting design, simulation, and manufacturing workflows across industries such as semiconductors, automotive, aerospace, and heavy equipment.

NVIDIA Omniverse libraries and Isaac Sim are available on Google Cloud Marketplace, enabling the development of physically accurate digital twins and robotics simulation pipelines. These tools allow organizations to simulate and validate systems before deployment.

In addition, NVIDIA NIM microservices can be deployed on Vertex AI and Google Kubernetes Engine to support vision AI and robotics workloads. These services enable capabilities such as real-time video analytics, robotic planning, and automated data processing.

Platform Focus: From Experimentation to Production

The updates position Google Cloud AI Hypercomputer as a full-stack platform for moving AI workloads from research to production. With tightly integrated compute, networking, software, and security capabilities, the platform is designed to support large-scale agentic systems, industrial automation, and real-time AI applications.

The post NVIDIA and Google Cloud Expand AI Hypercomputer Platform at Next 2026 appeared first on StorageReview.com.

LaCie has been a fixture in our lab for well over a decade. From the 8big Rack Thunderbolt 2 we covered in 2014 through the many generations of 5big, 6big, 8big, and Rugged devices that have followed, the formula has been consistent: premium Neil Poulton-designed enclosures, Seagate drives inside, Mac-centric polish, a solid warranty, and a clear focus on creative professionals. The new LaCie 8big Pro5 carries that pedigree forward in build quality, design, and purpose, and arrives at a notable inflection point for high-capacity direct-attached storage.

With eight 32TB HAMR-based Seagate IronWolf Pro drives on board, the 8big Pro5 tops out at 256TB of raw capacity. As far as turnkey desktop DAS products go, nothing else on the market ships at that capacity today. Competing 8-bay Thunderbolt enclosures from OWC, Sabrent, and others cap out at around 192 TB with the previous-generation PMR drives. While it is technically possible to roll your own by pairing a bare enclosure with eight 32TB IronWolf Pros, that DIY route leaves you stitching together the warranties across vendors. Seagate backs the complete LaCie kit end-to-end, including the drives, which is an advantage at this capacity point and for the value of the workloads involved.

Heat-assisted magnetic recording has been more than two decades in the making, and it has finally moved from hyperscale sampling to a product that a creative professional can put on a desk. For teams working with multi-stream 4K and 8K RAW footage, large photogrammetry or virtual production asset libraries, or AI-assisted content pipelines that consume storage faster than any prior generation, the jump from 24TB-era PMR drives to 32TB HAMR in the same eight bays is a meaningful change. We walked through the technical foundations of HAMR with Seagate’s Colin Presly on Podcast #124: The Path to 50TB HDDs with Frickin Lasers. The roadmap Colin laid out then is now shipping as product, with Mozaic 3+ drives at 30TB and up, Mozaic 4+ pushing to 44TB, and a longer arc toward 100TB drives as platter density continues to climb.

Around that storage core, LaCie delivers the rest of the package you would expect. The 8big Pro5 connects via Thunderbolt 5, which Seagate quotes at up to 80Gbps bidirectional for data, with additional headroom when combined with display traffic. In practice, the ceiling for a hard-drive array is set by the drives themselves. The IronWolf Pro 32TB is rated for up to 285 MB/s sustained, so eight drives in parallel have a theoretical maximum of about 2.2 GB/s before caching effects are taken into account.

The host port delivers up to 140W of power to a connected laptop, with two downstream Thunderbolt 5 ports rated at 30W each and a USB 20Gbps port rated at 15W for daisy-chained peripherals and displays. The LaCie 8big Pro5 ships preconfigured as a single RAID 5 array for 224TB of usable capacity, with RAID 0, 1, 6, 10, 50, and 60 available through LaCie RAID Manager. Build quality, thermals, and design are vintage LaCie, which we will cover in detail throughout the rest of this review. Pricing starts at $5,979 for the 32TB base configuration, with SKUs available up to 64TB, 128TB, 192TB, and 256TB.

LaCie 8big Pro5 – Build and Design

At the front of the LaCie 8big Pro5, the unit features a clean, minimal industrial design that aligns with its professional focus. It measures 11.69 inches in length, 9.13 inches in width, and 8.46 inches in height, giving it a compact yet substantial footprint for an eight-bay system.

Our review unit shipped fully populated with eight of Seagate’s new IronWolf Pro 32TB drives, for a total raw capacity of 256TB. With all drives installed, the system weighs just over 29 pounds, underscoring both its density and solid construction.

The enclosure itself is crafted from a single-piece aluminum chassis finished in metallic gray, giving it a premium, durable feel. Up front, each drive bay is tool-less, allowing quick, easy access to swap or service drives. Each tray is paired with an individual status LED, providing clear, at-a-glance visibility into drive activity and health without requiring interaction with the software.

At the rear, the LaCie 8big Pro5 maintains the same clean, functional design, with heavy perforations across the back panel to support airflow in a fully populated chassis. Power is handled via a standard C19 input and a physical power switch, confirming that the power supply is fully integrated into the unit rather than relying on an external brick.

Connectivity centers on four USB-C ports, each clearly labeled for its role. The leftmost port serves as the primary host connection, operating over Thunderbolt 5 with up to 80Gbps bandwidth and delivering up to 140W of power, making it well-suited for powering and connecting a laptop with a single cable.

Next to it are two additional Thunderbolt 5 downstream ports. These ports enable expansion beyond the enclosure, supporting external storage devices or displays while also delivering up to 30W of power to connected peripherals. This makes the unit function as both a high-capacity storage array and a compact docking hub.

The final USB-C port supports a 20 Gbps connection, intended primarily for additional storage expansion. It also provides up to 15W of power, which is sufficient for bus-powered drives and similar accessories.

To round things out, there is a Kensington lock slot for physically securing the device, a practical addition for shared workspaces or studio environments where the unit may not always be in a controlled rack or locked room.

From a wider rear view, the airflow design becomes much more apparent. The majority of the back panel is perforated, allowing the system to move a significant amount of air across all eight drives. Cooling is handled by a three-fan setup, with two larger fans serving the primary drive bay area and a smaller fan dedicated to the lower section housing the controller and power components. This separation helps ensure consistent airflow across both the storage and internal electronics. This is especially important in a fully populated 256TB configuration where thermal buildup can become a limiting factor over sustained workloads.

You can also see the subtle branding here, with “LaCie – design by Neil Poulton” centered along the upper portion of the rear panel, reinforcing the industrial design heritage that has been a hallmark of LaCie systems for years.

Up top, LaCie adds a simple yet practical touch with the integrated handle cutouts. Machined directly into the aluminum, these recessed grips provide a secure way to lift and move the unit without compromising the clean design language.

Given that the system weighs just over 29 pounds when fully populated, a built-in grip like this makes a noticeable difference during deployment or repositioning. It is a small detail, but one that reflects an understanding that this is not a lightweight desktop accessory and will occasionally need to be handled with a bit more care.

LaCie 8big Pro5 – LaCie RAID Manager software

To manage the 8big Pro5’s storage configuration, LaCie requires its RAID Manager software. This utility is available for Windows and macOS and is necessary to configure the array in RAID modes or switch the unit to JBOD, depending on your deployment needs.

Through RAID Manager, users can choose from a full range of RAID levels, including RAID 0, RAID 1, RAID 5, RAID 6, RAID 10, RAID 50, and RAID 60. This flexibility allows the unit to be tailored for everything from maximum performance to high levels of redundancy and fault tolerance. As shown here, a RAID 5 configuration using all eight 32TB drives yields 224TB of usable capacity and provides single-drive fault tolerance through parity.

In addition to RAID configuration, the software also allows you to format the array in either APFS for macOS environments or NTFS for Windows deployments, making it easy to integrate into mixed or platform-specific workflows. The interface itself is straightforward, providing visibility into drive status, serial numbers, and overall array health, while also confirming valid configurations before deployment.

LaCie 8big Pro5 – Performance

For Windows testing, we leveraged a Dell Pro Max 14 with the following configuration:

• Intel Core Ultra 9 285H
• NVIDIA RTX PRO 2000 8GB GDDR7
• 64GB LPDDR5X-8400
• 1TB SSD

For macOS testing, we used an M4 MacBook Air.

To evaluate the performance of the 8big Pro5, we began testing in a Windows environment with ExFat, configuring the array in RAID 5. This setup reflects a common balance of capacity, performance, and redundancy for general-purpose use. In this configuration, we ran a series of benchmarks, including IOMeter for synthetic workload analysis, Blackmagic Disk Speed Test for media-focused throughput, and PCMark 10 Disk Benchmark to capture more real-world application behavior.

After completing Windows testing, we switched to a macOS environment using RAID 5 and ExFAT. This allowed us to measure the performance of the same configuration across Windows and Mac environments. In this configuration, we reran Blackmagic Disk Speed Test to compare results in a macOS-native workflow and added ATTO Disk Benchmark to analyze performance across varying transfer sizes.

Blackmagic Disk Speed Test

The Blackmagic Disk Speed Test benchmarks a drive’s read and write speeds to estimate its performance, especially for video editing tasks. It helps users ensure their storage is fast enough for high-resolution content, such as 4K or 8K video.

The Blackmagic results show clear, real-world performance gains across RAID configurations. In RAID 5 in Windows, the 8big Pro5 delivers 1,418.4 MB/s read and 2,061.5 MB/s write speeds, offering a strong balance of performance and data protection. When moved to macOS, read performance remains nearly identical at 1,414.9 MB/s, while write speeds are 1,751.3 MB/s, reflecting some platform differences rather than a limitation of the array itself.

Looking at the Blackmagic workload breakdown, RAID 5 still proves more than capable for high-resolution media workflows. At these speeds, the array comfortably supports formats up through 8K, including 8K DCI and even 12K playback in several codecs, with consistent results across ProRes 422 HQ and H.265. This reinforces that RAID 5 is not just a safe option, but a practical one for professional video editing where both performance and redundancy matter.

In practice, RAID 5 delivers more than enough performance for demanding video workflows while maintaining data protection.

Blackmagic (higher is better)	LaCie 8big Pro5 – Windows Raid 5 ExFat	LaCie 8big Pro5 – macOS Raid 5 ExFat
Read	1,418.4 MB/s	1,414.9 MB/s
Write	2,061.5 MB/s	1,751.3 MB/s

PCmark 10 Storage

PCMark 10 Storage Benchmarks evaluate real-world storage performance using application-based traces. They test the system and data drives, measuring bandwidth, access times, and consistency under load. These benchmarks offer practical insights beyond synthetic tests, enabling users to compare modern storage solutions effectively.

The PCMark 10 result of 717 gives a useful look at how the 8big Pro5 behaves under real-world workloads rather than pure synthetic throughput. This benchmark incorporates traces from everyday applications, which tend to be more sensitive to latency and mixed I/O patterns than large sequential transfers.

PCmark 10 Storage (higher is better)	LaCie 8big Pro5 – Windows Raid 5 ExFat
Overall Score	717

IOMeter

We also ran the LaCie 8big Pro5 array through IOMeter. This lets us dig deeper into workloads, including random and sequential performance. We tested the 8big with a single queue to simulate lighter use and with four queue to see how the DAS handles heavier, more demanding scenarios.

At 1 queue, sequential performance is 1,752.2 MB/s read and 1,851.5 MB/s write, showing strong throughput even under a lighter load. Random 2MB performance lands at 233.8 MB/s read, and 654.1 MB/s write, while small-block 4K operations reach 297 IOPS read and 5,482 IOPS write.

IOMeter (1  queue)	LaCie 8big Pro5 – Windows Raid 5 Raw
Seq 2MB Read	1,752.2 MB/s
Seq 2MB Write	1,851.5 MB/s
Random 2MB Read	233.8 MB/s
Random 2MB Write	654.1 MB/s
Random 4K Read	297 IOPS
Random 4K Write	5,482 IOPS

Scaling to 4 queue, sequential reads increase to 1,949.1 MB/s, while writes remain steady at 1,873.6 MB/s, indicating the array is already near its write ceiling. Random 2MB performance improves more noticeably, with reads rising to 391.1 MB/s and writes to 980.5 MB/s. For 4K workloads, reads scale to 1,103 IOPS, while writes settle at 4,458 IOPS.

IOMeter (4 queue)	LaCie 8big Pro5 – Windows Raid 5 Raw
Seq 2MB Read	1,949.1 MB/s
Seq 2MB Write	1,873.6 MB/s
Random 2MB Read	391.1 MB/s
Random 2MB Write	980.5 MB/s
Random 4K Read	1,103 IOPS
Random 4K Write	4,458 IOPS

ATTO Disk Benchmark Summary (LaCie 8big Pro5 – macOS RAID 5, ExFat)

The ATTO results provide a clear picture of how the 8big Pro5 behaves in macOS when pushed to maximum throughput across a wide range of transfer sizes in a RAID 5 configuration.

At lower transfer sizes, performance ramps up gradually, as expected for an HDD-based array. Small-block operations (under 16KB) remain relatively modest, but once you move to larger transfer sizes, the system scales more effectively.

From around 64KB onward, throughput stabilizes and becomes a far more representative measure of real-world performance. Peak read speeds reach approximately 3.4 GB/s, while write performance settles slightly lower in the 2.7-3.1 GB/s range across larger block sizes.

Overall, the results show strong sequential performance, with the array delivering high read throughput and slightly lower, but still consistent, write speeds under sustained workloads.

Conclusion

The LaCie 8big Pro5 marks a meaningful leap forward for the line. At 256TB raw over Thunderbolt 5, with eight HAMR-based IronWolf Pro drives housed in a well-designed Neil Poulton enclosure, it is the first turnkey desktop DAS to deliver both a massive capacity jump and next-generation interface bandwidth to creative pros in a single box. The 8big formula is all here: premium build, thoughtful thermals, quiet operation, mature RAID management through LaCie RAID Manager, and a clear focus on the video, photo, and 3D asset workflows that have consistently outpaced the storage they rely on.

Performance lands where a well-tuned eight-bay array should. In RAID 5, the array comfortably handles multi-stream 4K and 8K editing with room to spare. Small-block random performance is modest, as expected for any HDD-based array, but that is not the workload profile this product is built for. For bulk sequential transfers, active project storage, and long-form media ingest, the array delivers the throughput that modern creative workflows need. The Thunderbolt 5 host port with 140W of power delivery, plus the two downstream TB5 ports and the 20Gbps USB-C, also make the unit a legitimate one-cable docking solution for a laptop-based edit bay, not just a storage target.

Pricing starts at $5,979 for the 32TB base configuration and scales up through 64TB, 128TB, 192TB, and 256TB tiers. That is a meaningful investment, but a 5-year warranty that covers both the enclosure and the drives end-to-end, Rescue Data Recovery Services, and the operational simplicity of a single-box deployment distinguish it from a DIY build using bare IronWolf Pros and a third-party enclosure. For creative professionals, production teams, and studios working at 4K, 8K, and beyond, and for anyone whose project data has outgrown what previous-generation PMR arrays could deliver in the same footprint, the 8big Pro5 is the most capable turnkey desktop DAS available today and earns the shortlist spot for high-end workflows that need both the capacity and the interface to match.

Product Page – LaCie 8big Pro5

The post LaCie 8big Pro5 Review: 256TB of HAMR-Powered Thunderbolt 5 DAS appeared first on StorageReview.com.

Seagate’s FireCuda X Vault is the gaming-flavored half of a two-drive launch that brings bus-powered USB-C to 3.5-inch external hard drives for the first time. Available in 8TB and 20TB capacities starting at $269.99, it runs on a single USB-C cable for both data and power, provided the host port can supply at least 15W. That’s the same category-first hook Seagate is pitching with the new One Touch Desktop HDD, but the FireCuda X Vault trades the One Touch’s clean-desk minimalism for customizable RGB with Windows Dynamic Lighting support, Xbox on PC certification, and a one-month Xbox Game Pass Ultimate trial.

The pitch here is overflow storage for buyers who’ve outgrown smaller drives and want a clean way to add serious capacity to a gaming PC or streaming rig. Large game libraries, captured gameplay, archived installs, and media collections are the target workloads. It’s worth being upfront about what it isn’t: the 5400 RPM drive inside won’t deliver SSD-like load times, so this isn’t the place to install the games you actually play. The better pairing is an internal NVMe for active titles and the FireCuda X Vault for everything else. Despite the Xbox branding on the box, the drive is PC-only and is not compatible with Xbox Series X/S. And because 15W USB-C delivery isn’t universal on older systems, it’s worth confirming your port can feed it before committing.

Seagate bundles Toolkit with the FireCuda X Vault, adding a decent set of storage management features beyond basic file transfers. Incremental backup copies only files that are new or changed after the first run, which helps reduce backup time for repeat jobs, and it supports both scheduled backups and manual runs. The software also includes folder mirroring for keeping selected directories synced, password protection on supported setups, and direct import from USB devices or memory cards.

The FireCuda X Vault 8TB model is estimated to hold roughly 110 to 145 games, based on installations ranging from 80GB to 150GB, along with about 800 hours of 1080p video or around 120 hours of 4K footage. The 20TB version increases that to around 275-360 games, about 2,000 hours of 1080p video, or roughly 300 hours of 4K video.

Backed by a 2-year warranty, Seagate includes the drive, a 0.5-meter USB-C cable, Toolkit software, a quick start guide, and two years of Rescue Data Recovery Services. Seagate also adds a one-month Xbox Game Pass Ultimate offer for new users and a two-month Adobe Creative Cloud Pro subscription, which makes sense given its gaming and content-creation use cases.

Seagate FireCuda X Vault Specifications

Specification/Feature	Seagate FireCuda X Vault
Overview
Product Name	Seagate FireCuda X Vault
Product Type	Bus-powered USB-C external hard drive
Form Factor	3.5-inch USB-C desktop drive
Target Audience	PC Gamers, Streamers, and Content Hoarders
Capacities offered	8TB, 20TB
Connectivity and Compatibility
Connection	USB-C
Power	Bus-powered, single-cable USB-C desktop storage, no external power required
USB-C power requirement	USB-C port must supply equal to or greater than 15W for drive operation
Operating System Compatibility	Compatible with most Windows and macOS systems
Time Machine	Reformatting required for use with Time Machine
Toolkit software compatibility	Toolkit software not compatible with ChromeOS
Xbox on PC	Designed for Xbox on PC
Software and Features
Toolkit included	Yes
Toolkit features	Incremental Backup: Keeps data protected while minimizing backup time by saving only new or changed files Scheduled or “Backup Now” Options: Supports both hands-off automation and manual control Mirroring (RealTime Sync): Maintains an always-updated copy of active folders on the drive Seagate Secure (Password Protection): Helps prevent unauthorized access if the drive is lost or shared Import from USB / Memory Cards: Simplifies photo and video offloads directly to the drive RGB: Allows for various RGB illumination customization options
RGB lighting	Customizable RGB lighting with Windows Dynamic Lighting support
Rescue Data Recovery Services	Included
Capacity Estimates
8TB	~800 hours (≈10 GB/hr) 1080p HD Video ~120 hours (≈60–70 GB/hr) 4K Video ~110-145 (≈80-150GB Each) Games
20TB	~2,000 hours 1080p HD Video ~300 hours 4K Video ~275-360 (≈80-150GB Each) Games
In the Box and Bundles
What’s in the box	Firecuda X Vault Main Unit 1.64-foot (0.5m) USB-C cable Toolkit software Quick start guide
Warranty	2-year limited warranty (may vary in region)
Data recovery coverage	2-year Rescue data recovery services (may vary in region)
Bundled offers	Free month of Xbox Game Pass Ultimate included in box (for new users) Complimentary 2-month subscription to Adobe Creative Cloud Pro (All Apps)

Seagate FireCuda X Vault Design and Build

The FireCuda X Vault has a very distinct desktop look. The front features vertical ribbing wrapped by the outer shell, with a distinct opening at the top where the LED emits light. It provides immediate power feedback via this LED, glowing white when the drive is getting enough power and red when the USB-C source is not supplying enough.

There are no ports or controls on the front panel. One side carries only the FireCuda X branding, while the rear has only a single USB-C port. The design is pretty basic, and the LED light may make it a bit much for some work environments; however, for gaming or home use, the drive will fit in well.

The outer shell is mostly plastic, and the base uses a high-friction material that helps keep the drive in place on a desk. It runs on bus power and passive cooling.

For everyday use, the single-cable design keeps setup simple, and the shape leaves enough open space around the ribbed sections, so placing two units one above the other does not appear to create an obvious airflow problem. However, the weak point is the RGB lighting. The top light bar fits the overall style, but the diffusion is uneven, so the glow looks patchy rather than smooth.

Seagate FireCuda X Vault Performance

To evaluate the performance of the Seagate FireCuda X Vault, we compared it against the Seagate One Touch Desktop HDD across a variety of benchmarks.

Here’s the high-performance test rig we used for benchmarking:

• CPU: AMD Ryzen 7 9850X3D
• Motherboard: Asus ROG Crosshair X870E Hero
• RAM: G.SKILL Trident Z5 Royal Series DDR5-6000 (2x16GB)
• GPU: NVIDIA GeForce RTX 4090
• OS: Windows 11 Pro

The drive inside our 8TB Seagate FireCuda X Vault self-reported as the Seagate SkyHawk (ST8000VX009) at 5400 RPM.

Blackmagic Diskspeed Test

First up is the Blackmagic test, where we evaluated the Seagate FireCuda X Vault against the One Touch Desktop HDD.

The Blackmagic Disk Speed Test benchmarks a drive’s read and write speeds to estimate its performance, especially for video editing tasks. It helps users ensure their storage is fast enough for high-resolution content, such as 4K or 8K video.

In this run, the FireCuda X Vault reached 222.4MB/s read and 158.9MB/s write. The read performance stands out here, coming in noticeably ahead of the One Touch’s 211.9MB/s, and landing fairly close to Seagate’s quoted maximums for its internal FireCuda drives. Write performance tells a different story, where the One Touch leads at 211.2MB/s, putting the FireCuda’s 158.9MB/s more in line with typical HDD behavior.

Blackmagic (higher is better)	Seagate FireCuda X Vault 8TB	Seagate One Touch Desktop HDD 8TB
Read	222.4 MB/s	210.9 MB/s
Write	158.9 MB/s	152.0 MB/s

IOMeter

In the 1-queue IOMeter test, the FireCuda X Vault demonstrated strong sequential performance, reaching 224.03 MB/s read and 223.37 MB/s write, outperforming the One Touch Desktop HDD, which came in at 211.26 MB/s read and 211.48 MB/s write. This reinforces the FireCuda’s advantage in sustained, large-block transfers.

Random 2MB performance was much closer between the two drives. The FireCuda posted 117.17MB/s read and 149.59MB/s write, while the One Touch slightly edged ahead in write performance at 150.06MB/s and trailed slightly in reads at 113.83MB/s. These small differences are within the margin expected for mechanical drives.

Small-block performance remained predictably low across both drives. The FireCuda delivered 429 IOPS in random 4K writes and 126 IOPS in reads, nearly identical to the One Touch at 424 IOPS in writes and 129 IOPS in reads. At this level, neither drive is designed for latency-sensitive workloads, and their performance is effectively comparable.

IOMeter Test	Seagate FireCuda X Vault 8TB	Seagate One Touch Desktop HDD 8TB
Seq 2MB Write	223.37 MB/s	211.48 MB/s
Seq 2MB Read	224.03 MB/s	211.26 MB/s
Random 2MB Write	149.59 MB/s	150.06 MB/s
Random 2MB Read	117.17 MB/s	113.83 MB/s
Random 4K Write	429 IOPS	424 IOPS
Random 4K Read	126 IOPS	129 IOPS

PCMark 10

PCMark 10 Storage Benchmarks evaluate real-world storage performance using application-based traces. They test the system and data drives, measuring bandwidth, access times, and consistency under load. These benchmarks offer practical insights beyond synthetic tests, enabling users to compare modern storage solutions effectively.

In PCMark 10’s Data Drive Benchmark, both drives performed nearly identically, with the Seagate One Touch Desktop HDD scoring 750 and the Seagate FireCuda X Vault close behind at 746. This minimal difference indicates that, in trace-based workloads, there is no meaningful performance gap between the two.

As expected for high-capacity HDDs, both drives are better suited for bulk storage tasks such as backups, media libraries, and large file transfers rather than latency-sensitive workloads. Overall, this result shows that real-world responsiveness between the two is effectively on par in this test.

PCMark 10 Storage (higher is better)	Seagate FireCuda X Vault 8TB	Seagate One Touch Desktop HDD 8TB
Overall Score	746	750

Conclusion

The FireCuda X Vault’s appeal comes down to the same category-first hook as its One Touch sibling: a 3.5-inch desktop HDD that runs off a single USB-C cable with no power brick in the mix. For gamers and streamers who want to add significant capacity to a PC or laptop setup without another power supply on the floor, that’s a quality-of-life improvement over every desktop external HDD that came before it.

Performance lands where it should, for a 5400-RPM hard drive. Sequential read and write throughput sits in the 220 MB/s range; random workloads are modest; and small-block IOPS behave like the mechanical storage they are. Those numbers are fine for bulk transfers and archival use, but they confirm this isn’t a drive for running modern games directly. Pair it with an internal NVMe for active titles and use the FireCuda X Vault for everything that doesn’t need fast access.

Starting at $269.99 for 8TB, the pricing is competitive with other high-capacity external HDDs and considerably less than that of equivalent external SSDs. The RGB execution could be cleaner, the USB-C cable is short, and buyers should verify their host port can deliver 15W before committing. Those caveats aside, the FireCuda X Vault earns its spot on the shortlist for PC gamers, streamers, and media collectors who need ample local storage with minimal cable clutter.

Product Page – Seagate FireCuda X Vault

The post Seagate FireCuda X Vault Review: 20TB of Single-Cable Storage for Massive Game Libraries appeared first on StorageReview.com.

Seagate’s new One Touch Desktop HDD sidesteps one of the staples of the desktop external drive category: the power brick. The refreshed lineup runs 8TB, 20TB, and 24TB in a 3.5-inch chassis, but instead of a DC input and wall adapter, it draws everything it needs over a single USB-C cable. Seagate bills it as the industry’s only bus-powered USB-C desktop HDD, which is a meaningful shift in a segment where cable count and desk clutter have long been accepted costs of doing business. Pricing starts at $259.99 for 8TB and tops out at $619.99 for 24TB.

Beyond the cable story, the One Touch Desktop HDD is straightforward mechanical storage aimed at backup and archive workloads. It slots between the complexity of a NAS and the cost of high-capacity SSDs, working well as a companion to a smaller internal NVMe or as a bulk offload destination for photos, video, and project files. The bus-powered design also opens up use cases that traditional desktop drives can’t cover, such as pulling footage off a laptop in the field with no outlet nearby. Pair that with Windows and Mac support, Seagate’s Toolkit for backup and mirroring, and two years of Rescue Data Recovery Services, and the pitch comes down to storage headroom, data safety, and a cleaner desk at a competitive cost per terabyte.

Design & Features

The One Touch Desktop HDD features a refined, premium aesthetic, combining aluminum and plastic for a solid, high-quality feel. Rubber feet on the bottom also help stabilize the device and prevent unwanted movement during operation. To keep things clean and minimal, Seagate has also avoided adding unnecessary lighting elements.

For connectivity, the drive uses a single USB-C cable and does not require a separate power adapter, provided the host port can supply at least 15W. While this requirement may be a limitation for older systems, it ultimately simplifies setup for modern devices. A small front-facing status light is the only visual indicator, blinking red if insufficient power is detected.

Getting started is pretty straightforward; simply plug in the cable and wait for the volume to mount. You can optionally install the Seagate Toolkit software, but it works out of the box with both Windows and macOS. Time Machine users will need to reformat before initial use, though.

Inside the box, Seagate includes a (0.5m) USB-C cable, Toolkit software, a quick-start guide, and a 2-year limited warranty. In addition, users receive 2-year Rescue Data Recovery Services, which include one in-lab recovery attempt, with recovered data returned on an encrypted device if the attempt is successful. The turnaround time for the recovery service is about 30 days, which provides peace of mind for anyone relying on the drive for long-term storage.

For creatives, Seagate provides a complimentary 2-month trial subscription to Adobe Creative Cloud Pro (All Apps). This inclusion gives users access to tools they might otherwise pay for separately, making the overall package more compelling.

Feature	8TB	20TB	24TB
Specifications
Connector	USB-C
Interface	USB 3.2 Gen 1 (up to 5Gb/s)
Power	Bus-powered via USB-C (≥15W required)
Compatibility	Windows & macOS (Time Machine requires reformat; ChromeOS not supported for Toolkit)
In the Box & Software
What’s in the Box	One Touch HDD, 1.64ft USB-C cable, Toolkit software, Quick Start Guide
Included Software	Seagate Toolkit, 2-month Adobe Creative Cloud Pro (All Apps) trial
Support & Pricing
Warranty	2-year limited (may vary by region)
Rescue Data Recovery	2-year included (may vary by region)
MSRP	$259.99	$519.99	$619.99

Toolkit Software

Seagate Toolkit is a bundled utility that enhances the One Touch Desktop HDD’s functionality without complicating the user experience. After the initial backup, its incremental backup feature saves only modified files, helping keep backup times and system load manageable. At the same time, the Mirroring (RealTime Sync) feature continuously maintains updated copies of selected folders in the background. Additionally, Seagate Secure provides password protection for supported drives, while the Import function automatically transfers files from connected USB devices or memory cards, making it especially useful for frequent media offloads.

Moreover, Toolkit supports both scheduled and manual backups. Users who prefer automation can rely on scheduled backups, while those who want more control can trigger backups manually. Either way, it delivers essential data protection features without requiring third-party software.

Capacity in Context

To better understand available capacities, Seagate provides real-world storage estimates for common file types. Although actual results will vary depending on codec, compression, and workflow, these figures still offer a helpful baseline for planning:

Capacity	1080p HD Video (approx.)	4K Video (approx.)	RAW Photos (approx.)
8TB	~800 hours	~120 hours	~200,000
20TB	~2,000 hours	~300 hours	~500,000
24TB	~2,400 hours	~360 hours	~600,000

Performance

To evaluate the performance of the Seagate One Touch Desktop HDD, we compared it against the Seagate FireCuda X Vault across a variety of benchmarks.

Here’s the high-performance test rig we used for benchmarking:

• CPU: AMD Ryzen 7 9850X3D
• Motherboard: Asus ROG Crosshair X870E Hero
• RAM: G.SKILL Trident Z5 Royal Series DDR5-6000 (2x16GB)
• GPU: NVIDIA GeForce RTX 4090
• OS: Windows 11 Pro

The drive inside our 8TB Seagate One Touch HDD self-reported as the Seagate SkyHawk (ST8000VX009) at 5400 RPM.

Blackmagic Disk Speed Test

The BlackMagic Disk Speed Test benchmarks a drive’s read and write speeds to estimate its performance, especially for video editing tasks. It helps users ensure their storage is fast enough to handle high-resolution content, such as 4K or 8K video.

In Blackmagic, the Seagate FireCuda X Vault posted the stronger read speed at 222.4 MB/s, edging out the Seagate One Touch Desktop HDD at 210.9 MB/s. Write performance also showed a similar edge, with the One Touch measuring 152.0 MB/s compared to 158.9 MB/s from the FireCuda X Vault. Overall, both drives landed in expected territory for high-capacity external hard drives, though the FireCuda showed slightly better read and write speed.

Blackmagic (higher is better)	Seagate One Touch Desktop HDD 8TB	Seagate FireCuda X Vault 8TB
Read	210.9 MB/s	222.4 MB/s
Write	152.0 MB/s	158.9 MB/s

IOMeter

In the 1-queue IOMeter run, the FireCuda X Vault led in sequential throughput, reaching 224.03 MB/s read and 223.37 MB/s write, compared to 211.26 MB/s read and 211.48 MB/s write from the One Touch Desktop HDD. Random 2MB performance was much closer. The One Touch slightly led in random 2MB writes at 150.06MB/s versus 149.59MB/s, while the FireCuda posted the better random 2MB read at 117.17MB/s versus 113.83MB/s.

Small-block performance remained low on both drives, as expected for HDD-based storage, with the FireCuda reaching 429 IOPS in random 4K writes versus 424 IOPS on the One Touch, while the One Touch narrowly led in random 4K reads at 129 IOPS versus 126 IOPS on the FireCuda. Overall, the FireCuda showed a modest advantage in sequential performance, while the two drives were very close in lighter random workloads.

IOMeter Test	Seagate One Touch Desktop HDD 8TB	Seagate FireCuda X Vault 8TB
Seq 2MB Write	211.48 MB/s	223.37 MB/s
Seq 2MB Read	211.26 MB/s	224.03 MB/s
Random 2MB Write	150.06 MB/s	149.59 MB/s
Random 2MB Read	113.83 MB/s	117.17 MB/s
Random 4K Write	424 IOPS	429 IOPS
Random 4K Read	129 IOPS	126 IOPS

PCMark 10 Storage

PCMark 10 Storage Benchmarks evaluate real-world storage performance using application-based traces. They test the system and data drives, measuring bandwidth, access times, and consistency under load. These benchmarks offer practical insights beyond synthetic tests, enabling users to compare modern storage solutions effectively.

In PCMark 10’s Quick System Drive Benchmark, both drives delivered nearly identical performance, with the Seagate One Touch Desktop HDD scoring 750 and the Seagate FireCuda X Vault coming in at 746. This narrow gap suggests that, in trace-based workloads, the two drives perform very similarly, with no meaningful advantage for either.

As expected for high-capacity HDDs, both are best suited for bulk storage tasks such as backups, media libraries, and large file transfers rather than latency-sensitive workloads. Overall, this result shows that real-world responsiveness between the two is effectively on par in this test.

PCMark 10 Storage (higher is better)	Seagate One Touch Desktop HDD 8TB	Seagate FireCuda X Vault 8TB
Overall Score	750	746

Conclusion

The Seagate One Touch Desktop HDD is a category-first product in a commoditized space. Bus-powered USB-C on a 3.5-inch desktop drive genuinely changes how the drive fits on a desk or travels in a bag, and it’s the feature most likely to sway buyers who’ve grown tired of juggling bulky power bricks. Cross-platform support, Toolkit for backup and mirroring, and two years of Rescue Data Recovery Services round out a package that covers the basics without asking for much from the user.

Performance lands where it should for 5400 RPM mechanical storage. Sequential throughput sits in the low 200s MB/s, random workloads are modest, and small-block IOPS are firmly in HDD territory. That rules it out for anything latency sensitive or for active video editing off the drive, but those aren’t the workloads this product targets. For backup, archive, media libraries, and bulk offload, it does the job.

At $259.99 for 8TB and $619.99 for 24TB, pricing is competitive against other high-capacity external HDDs, and the single-cable design is a real differentiator rather than a marketing one. For users who want maximum capacity with minimum desk footprint and cable clutter, the One Touch Desktop HDD earns its spot on the shortlist.

Product page – Seagate One Touch Desktop HDD

The post Seagate One Touch Desktop HDD Review: 24TB Without the Power Brick appeared first on StorageReview.com.

When we published our story on Oregon State University’s plankton imaging research last November, the headline was the science: AI-accelerated infrastructure aboard research vessels, processing terabytes of ocean data in near real-time before the ship ever reached port. But something else happened quietly in the weeks that followed. Word spread across campus about what a single Dell PowerEdge XE7745 with eight Solidigm D5-P5336 E3.S SSDs and NVIDIA RTX PRO 6000 GPUs could actually accomplish. Other departments started asking questions. Then they started making calls. Christopher Sullivan, Director of Research and Academic Computing at OSU’s College of Earth, Ocean, and Atmospheric Sciences, now wants a rack of these servers to meet the growing AI demand across the university, and the story driving that ambition goes well beyond plankton.

Oregon State has established itself as one of the most forward-thinking universities in the nation in its adoption of AI for both research and academic use. The infrastructure decisions being made on campus today, along with the formation of partnerships with companies such as Metrum AI, Dell, NVIDIA, and Solidigm, are not just academic experiments. They lay the groundwork for a new way for universities to deliver education, assess learning, and protect their students. This is the story of how that model was developed.

The Problem That Generative AI Made Worse

For decades, written assignments were central to academic evaluation. Submit a paper, show understanding, and get a grade. Generative AI has fundamentally altered that system. Now, a student can craft a polished, well-structured essay with little real engagement with the material, and even seasoned faculty can’t reliably tell if it is authentic work. The evidence of genuine understanding that universities relied on for generations has weakened.

The obvious alternative is oral evaluation. Ask students to explain their reasoning out loud, walk through their analysis, and defend their conclusions. That is hard to fake. The problem is scale. A professor teaching 200 students cannot sit across from each one and conduct a substantive oral exam. In the modern university, that constraint has effectively shelved oral assessment as a primary evaluation tool. Metrum AI was built to change that equation.

What Metrum AI Built

Metrum AI, co-founded by CEO Steen Graham and CTO Chetan Gadgil, was built around a simple conviction: AI should do real operational work, not just demonstrate potential. The company deploys multimodal AI agents that reason across video, audio, documents, and structured data for customers in industries from insurance to manufacturing. Metrum has developed a close partnership with Dell Technologies, validating its platforms against Dell’s enterprise server infrastructure across a range of GPU configurations. The academic evaluation system at Oregon State is not a pivot for Metrum; it is the same underlying capability applied to a new problem domain, with the same on-premises, human-in-the-loop design philosophy that runs through everything the company builds.

Applied to academic assessment, the platform processes recorded student video presentations using multimodal AI and returns rubric-aligned draft evaluations for faculty review. The pitch is specific: give instructors an AI partner that handles the repetitive, time-consuming extraction work so they can focus on the judgment calls that truly require human adjudication.

At a functional level, the platform carries out three operations. It extracts multimodal artifacts from submitted videos, generating timestamped audio transcripts with OpenAI Whisper and capturing slide content via visual analysis powered by Qwen3-VL-30B. It then applies instructor-designed rubrics to the extracted content, using Qwen3-30B-A3B reasoning models that run on vLLM. Finally, it presents draft evaluations with evidence pointers, linking each score to a specific transcript timestamp or slide identifier for faculty review and approval before anything reaches a student.

That final step is crucial. No score, comment, or piece of feedback is visible to students until an instructor has reviewed it, made any necessary modifications, and explicitly approved it. The system is built around faculty authority. Additionally, the platform operates entirely on-premises. This decision influences everything about how the system functions, who trusts it, and what hardware it requires.

From a Professor’s Side Project to a Provost Mandate

Jonathan Kalodimos is an Associate Professor of Finance and the Harley and Brigitte Smith Fellow in the College of Business at Oregon State University. His background is not what you might expect from someone at the center of an AI infrastructure story. Before joining OSU, he was a financial economist at the U.S. Securities and Exchange Commission, where he served as lead economist on Dodd-Frank Act Section 954, which established rules around executive compensation clawbacks. His research on corporate governance and financial regulation has been cited in The Wall Street Journal, The New York Times, Bloomberg, and the Harvard Business Review. He also, it turns out, has a physicist’s instinct to measure things precisely.

About a year ago, Kalodimos coded a simple tool for his MBA class: an AI agent to evaluate the oral component of case study presentations. The students were impressed with the quality of feedback. He presented the project during AI Week at Oregon State. Dell took notice, connected him to Metrum AI, and a classroom experiment became something much larger.

“Once you have the tool, you can refine your teaching style and your teaching methods to leverage the strength of the tool to provide a better educational experience.”

— Jonathan Kalodimos, Associate Professor of Finance and Harley & Brigitte Smith Fellow, Oregon State University

What Kalodimos is building toward is what he calls evidence-based extraction, underpinned by what he describes as rubric engineering. This encompasses determining which features are extractable from a student presentation, aggregating those features into learning outcomes, and providing faculty with a structured view of where each student demonstrated understanding and where they fell short. “The way I explain this to skeptical students,” he said, “is if I had a very detailed checklist, and I went through your presentation checking off things you did, that’s what the system is doing. Obviously way more sophisticated than that, but it’s allowing me to see all the opportunities for the student to demonstrate that they know this material.”

He offered two examples that illustrate what the system changes in practice. In the first, a student condensed a ten-minute presentation into five minutes, spoke in a monotone, and spoke at a pace as if English were a second language. His delivery obscured his comprehension entirely. “Even though I was listening carefully,” Kalodimos said, “I just couldn’t, or wouldn’t, break it down into that level of granularity to overcome the delivery element so I could focus on the actual evidence.” When he later walked through the AI-generated evidence breakdown in fifteen-second increments, it became clear the student understood the material. The delivery had been graded, not the knowledge.

The second case involved a student who built a presentation slowly, with what seemed like disjointed slides, and only pulled the argument together on the final slide. Watching live, Kalodimos had already formed a low opinion of the presentation. The system evaluated the work as a complete arc and scored it well. “I didn’t even think that would be a benefit of this type of evaluation,” he said. “It’s getting away from the time element of evaluation.”

Christopher Sullivan, Director of Research and Academic Computing, is the infrastructure lead on the deployment. His involvement sharpened when a critical compliance gap emerged in the original Metrum and Dell design.

FERPA, Data Sovereignty, and Why the Cloud Is Not the Answer

When Sullivan stepped in to build OSU’s on-premises implementation of the Metrum platform, the first thing he identified was a problem nobody had fully solved: The Family Educational Rights and Privacy Act (FERPA).

FERPA is the federal law governing student education records. It establishes strict requirements around who can access student data, under what conditions, and how it must be protected. For a system like Metrum’s, one that ingests student video submissions, generates transcripts, produces evaluations, and stores the complete history of every grading decision, FERPA compliance is not a checkbox; it’s an architectural constraint.

“We needed to be able to bring something on-premises that would meet all of my FERPA conditions,” Sullivan said, “but also have a large amount of storage space.” Cloud processing was not compatible with that requirement. Routing student video files, audio transcripts, and evaluation records through external AI APIs would mean transmitting personally identifiable student information to third-party systems outside the university’s direct control. The contractual and technical complexity of maintaining FERPA compliance in that environment, across every vendor in the chain, made it a non-starter.

There is also a practical student experience dimension. Students submitting recorded presentations are offering something personal: their voice, their face, their reasoning under pressure, sometimes in their non-native language. When they understand that their video is stored on an OSU server, processed by a model running on OSU hardware, and governed by OSU’s own data policies, the dynamic changes. Kalodimos saw this play out directly during the pilot. “The idea that this was a local model with local storage and OSU has the student’s back,” he said, “was palpable. We really need to use the institutional trust that OSU has built, protecting our students, leveraging these on-prem solutions.”

Cloud AI platforms are easy and quick to deploy, but they require institutions to place trust in a contract rather than in their architecture. For students who are already wary of how their data is managed, that distinction can significantly influence their willingness to adopt. On-premises deployment isn’t just about compliance; it establishes a foundation of trust.

The Pilot, the Provost, and What Comes Next at OSU

The pilot is underway. Approximately 500 students across multiple sections are submitting final projects at the close of finals week. Graded evaluations must be returned within 4 days of the last submission. The AI-generated reports have to be ready before faculty begin grading. “There’s a human component running in parallel,” Kalodimos noted, “but they need the report first. If it takes two days to process all of these, then the human element is even more compressed.” The pressure is real, the deadline is fixed, and the infrastructure is doing its job.

The pilot has surfaced something else worth noting. When a professor was recently promoted to an administrative role mid-term, an instructor had to step in on short notice and finish out the course. Having a consistent AI evaluation framework already in place, with defined rubrics and an established review workflow, gave that instructor a thread of continuity that otherwise would not have existed. “Having a consistent AI evaluation companion,” Kalodimos said, “is going to definitely improve the student experience” in exactly those situations when continuity of human instruction cannot be guaranteed.

The story eventually reached OSU’s Provost. Kalodimos presented the full stack: the Dell system, Solidigm storage performance, developer capabilities, and infrastructure benchmarks, in what was supposed to be a ten-minute meeting. It ran for forty minutes. The Provost followed up with an email to the CIO, the CTO, and Sullivan. OSU is now planning a university-wide deployment to make the resource available to faculty starting the spring term, managed by a newly defined Research Computing office that sits under the Provost and the research office.

Sullivan is thinking about that deployment in terms of rack-scale infrastructure. The same XE7745 platform that anchored the plankton imaging work and is now powering the Metrum AI evaluation pipeline is the foundation he wants to scale. The goal is a rack of these servers, available to float between academic compute and research compute workloads as demand shifts. Ideally, the servers would be dedicated to the Metrum evaluation pipeline during midterm and final submission surges, and redeployed to research workloads during quieter periods of the academic calendar. “We can take machines from that set and shove them into the academic compute side for a period of time, and then bring them back and leverage them for the research compute,” Sullivan said. “We want to be able to redeploy them on the fly.”

The organic adoption is already underway. Faculty from the College of Health and the College of Engineering have independently approached Kalodimos after hearing about the project through informal channels. The platform has not been formally announced beyond the pilot. It found its audience anyway.

The Capacity Problem Behind the Grading Problem

There is a version of this story that is only about speeding up grading. That is an incomplete vision.

The larger version is about class capacity. Sullivan described a 100-level geology course, a class that OSU treats as part of its core educational mission and that every student is meant to take. It currently runs two sections of 300 students each, for a total of 600 per quarter. The instructors are at their limit. Adding sections is not feasible given current teaching loads. “I can’t have the teachers do more work,” Sullivan said. “I need to create pathways for us to either create more sections by reducing the load, or put more students in the sections we’ve got.”

Kalodimos similarly framed the College of Business dimension. Professors with sections capped at 45 students for fire code reasons would have the option to explore large-lecture formats with breakout-room support once individualized evaluation can scale. “It’s not just about packing bodies,” he said. “It’s about maintaining quality while exploring different delivery modes.” The AI evaluation layer is what makes individualized assessment at a lecture-hall scale operationally possible.

“The AI is helping us increase the numbers without changing the impact or the message or what’s being learned.”

— Christopher Sullivan, Director of Research and Academic Computing, Oregon State University

Storage Was the Missing Piece

When Sullivan assessed what it would take to bring the Metrum system on-premises at OSU with full FERPA compliance, the GPU side of the equation was already established. The Metrum and Dell reference architecture had demonstrated that the XE7745 with NVIDIA RTX PRO 6000 GPUs could handle the inference workload at scale. What remained unsolved was storage.

The XE7745 is a 4U air-cooled platform optimized for GPU density. That design is its strength, but it comes with a real constraint: drive bay count is limited. “I needed to put a lot of space into a single piece of equipment without compromising speed,” Sullivan said, “because I didn’t want to lose all the value of the GPUs and everything that XE7745 was worth. And there really weren’t a lot of large-capacity SSD solutions out there to do that in the box.”

The storage layer in a system like this carries more than the headline AI workload. Video files arrive from the student portal and need somewhere to buffer immediately. Extracted audio tracks and timestamped transcripts are stored as discrete artifacts for faculty review. Slide images and OCR output occupy their own tier. The Supabase database tracking submission metadata, draft evaluations, faculty edits, and approval records runs continuously. Model weights for Whisper, Qwen3-VL, and the reasoning model need to load quickly enough to avoid inference bottlenecks. And the full audit trail for every AI-generated draft, every faculty override, and every approval action must be retained as a queryable record for accreditation reviews, academic integrity investigations, and administrative reporting.

Every one of those workloads lives on storage. The GPU gets the credit for the AI output. Storage keeps the GPU fed and working continuously.

Sullivan’s team selected the Solidigm D5-P5336 in the E3.S form factor. The XE7745 holds eight of these drives. At 30.72 TB per drive, that is over 245 TB of flash storage in a single 4U chassis. The D5-P5336 uses QLC NAND with enterprise firmware tuned for sustained write performance and data integrity, which matters here because the system is not handling occasional bursts. During peak submission windows around finals, it simultaneously ingests videos, writes transcripts, logs evaluation output, and updates the database.

As we documented in our ocean research story covering this same hardware configuration, the Solidigm drives in RAID 10 delivered sustained read and write performance without falling behind the processing pipeline. Storage was not the bottleneck. The architecture exposed the actual workload constraints, so the team could tune them where they mattered. That validated conclusion carries directly into the academic evaluation deployment.

OSU as a Blueprint for AI-Ready Higher Education

Oregon State’s approach to AI infrastructure is deliberate and worth examining as a model for other educational institutions. Rather than deploying AI tools opportunistically through cloud APIs, the university has made a series of architectural decisions that treat AI as a durable institutional capability rather than a vendor service. Hardware is standardized around platforms that span research and academic compute workloads. Storage is on-premises, high-density, and compliant by design. Faculty retain final authority over every evaluation the system produces.

Kalodimos is explicit about wanting the platform to travel beyond OSU. “Not every university is going to be as well-resourced as us,” he said. “I want to make sure this technology is available to all universities.” That foundation is what makes the broader argument for educational equity credible. An instructor at a smaller institution with a heavier teaching load and fewer resources arguably needs this tool the most.

“I need storage, and I need that storage to be fast. AI is a dead technology without storage. It’s a data-driven system. We had the algorithms back in the 1960s and 70s. We didn’t have any data to do it because we didn’t have any storage to actually hold that data.”

— Christopher Sullivan, Director of Research and Academic Computing, Oregon State University

Sullivan frames the hardware planning question the same way he frames every infrastructure decision at OSU. Models will change. The types of input students submit will evolve. The evaluation techniques faculty want to run will grow more sophisticated. “I’m going to get a bigger fork, a bigger knife, or a bigger spoon,” he said, “but it’s still a fork, knife, and spoon on the hardware side. I’m going to be changing the models and the inputs dramatically in the years to come, and it’s more important to me right now that the hardware keeps ahead of whatever those are going to be.”

Every processed transcript, every extracted slide, every draft evaluation, every approved grade, and every audit record must all live somewhere. In this system, there is 245 TB of Solidigm QLC flash on-premises within OSU’s infrastructure, doing the quiet work that makes the visible AI possible. The rack Sullivan is planning will not be the last one. The university is watching what this pilot produces, and other institutions will be watching what OSU does. That is what it means to lead in AI.

The post How Metrum AI and Oregon State University Are Building the New Standard for Academic Assessment appeared first on StorageReview.com.

ASUSTOR has announced the availability of the Lockerstor 24R Pro Gen2 (AS7224RDX), a higher-end rackmount NAS for enterprise and multi-user deployments, pairing a 24-bay chassis with AMD’s Ryzen 7 Pro 7745 processor. The AMD chip delivers 8 cores, 16 threads, and boost speeds up to 5.3 GHz. The system is designed for virtualization, container workloads, and heavy use of shared storage.

ASUSTOR Lockerstor 24R Pro Gen2 (AS7224RDX) Use Cases

The ASUSTOR Lockerstor 24R Pro Gen2 (AS7224RDX) is designed for large enterprise and data center storage deployments requiring high-capacity, scalable shared storage with built-in data protection and redundancy. It is also suited for virtualization and high-performance computing workloads, including containerized environments and VMware, Citrix, and Hyper-V deployments.

Its RAID options, high-speed networking, PCIe expansion, and NVMe support also make it well-suited for video production and media workflows, as well as for database management and big data applications that benefit from higher throughput and multi-user access. The platform also supports secure backup and long-term data storage through features such as MyArchive cold backup, Snapshot Center, Cloud Backup Center, and DataSync Center.

ASUSTOR Lockerstor 24R Pro Gen2 (AS7224RDX) Features

The unit ships with 16GB of DDR5-4800 ECC memory and can be expanded to 192GB across four slots, a spec that points to more demanding business use cases where error correction and larger memory pools matter. Storage flexibility includes 24 drive bays, one M.2 2280 NVMe slot over PCIe 5.0 x4 for cache or faster storage, and support for RAID 50 and RAID 60 for organizations balancing performance, capacity, and drive-failure protection.

Connectivity includes dual 10GbE and dual Gigabit Ethernet ports with support for link aggregation and SMB multichannel, while PCIe Gen5 x8 and x4 slots leave room for faster 25GbE, 40GbE, or 50GbE networking, as well as SAS expansion. ASUSTOR says the system can also scale to a total of 40 drive bays with its Xpanstor 12R expansion unit.

It also includes dual redundant 550W 80 PLUS Platinum power supplies and a five-year warranty. Software and data-protection features include snapshot support, cloud backup options, encryption, and virtualization support.

ASUSTOR Lockerstor 24R Pro Gen2 (AS7224RDX)
CPU	AMD Ryzen 7 Pro 7745, x64, 8 cores / 16 threads, 3.8GHz base, up to 5.3GHz
Memory	16GB ECC DDR5-4800 UDIMM, expandable to 192GB (4 slots, mixed capacity supported)
Flash Memory	32GB SATA DOM
Drive Bays	24
M.2	1 x M.2 2280 NVMe, PCIe 5.0 x4
Supported Drives	3.5-inch SATA HDD, 2.5-inch SATA HDD, 2.5-inch SATA SSD, M.2 2280 NVMe
RAID	RAID 50, RAID 60
Networking	2 x 10GbE RJ45, 2 x Gigabit Ethernet (1G/100M, WOL), link aggregation, SMB multichannel
Expansion Slots	1 x PCIe Gen5 x8, 1 x PCIe Gen5 x4; supports 10/25/40/50GbE NICs and SAS cards
USB	4 x USB 3.2 Gen 1 Type-A
HDMI	1 x HDMI (service only)
Expansion Capacity	Up to 40 drive bays with expansion unit; supports Xpanstor 12R via SAS
Power Supply	Dual redundant 550W PSUs, 80 PLUS Platinum
Cooling	4 x 60mm fans
Warranty	5 years
Power Consumption	253W (operation), 145W (disk hibernation)
Operating Temperature	0°C to 40°C
Storage Temperature	-20°C to 70°C
Humidity	5% to 95% RH
Dimensions	172(H) x 439(W) x 576(D) mm
Weight	18.6kg / 41lb
Certifications	FCC, CE, VCCI, BSMI, C-TICK, KCC, BIS, UKCA

ASUSTOR Lockerstor 24R Pro Gen2 (AS7224RDX) Availability

Backed by a 5-year warranty, the ASUSTOR Lockerstor 24R Pro Gen2 is now available from online retailers for about $7,100.

ASUSTOR Lockerstor 24R Pro Gen2 (AS7224RDX) Product Page

The post ASUSTOR Lockerstor 24R Pro Gen2 Available Now for Enterprise Storage and Virtualization appeared first on StorageReview.com.

TrueNAS has introduced the V160, a new addition to its enterprise appliance portfolio designed to support larger, more dynamic workloads. The platform targets organizations that balance performance requirements with cost control, especially as flash pricing volatility continues to affect infrastructure planning.

The V160 is built on dual fifth-generation TrueNAS controllers powered by AMD EPYC processors, with PCIe 5.0 connectivity and up to 768GB of DDR5 memory per controller. The system uses a hybrid architecture that combines high-capacity HDD tiers with NVMe flash and a large adaptive cache layer. TrueNAS reports up to 60 GB/s of throughput, driven by memory bandwidth, and up to 24 TiB of cache.

The hybrid design lets administrators tune the balance between NVMe and SAS HDD media across 24 internal bays and supports expansion to more than 1,400 drives in a single system. Additional scale-out options include up to six NVMe flash shelves or fourteen 102-bay SAS HDD shelves, enabling configurations with up to 20 PiB of flash or more than 35 PiB of HDD capacity. The platform does not impose capacity-based licensing, allowing organizations to scale storage without incremental software costs tied to capacity growth.

From a platform perspective, the V160 consolidates file, block, and S3-compatible object storage into a single system. This unified approach aims to reduce operational overhead from managing multiple storage silos, tools, and support models. The system supports a broad virtualization and container ecosystem, including VMware, Proxmox, Hyper-V, Xen, OpenShift, and Kubernetes. It includes high-availability and failover capabilities designed for large-scale virtual machine deployments.

Capability	TrueNAS V160
Raw Capacity	20 PiB NVMe, 35PiB HDD
Throughput	Up to 60 GB/s
Network	4x 100/200, 2×400 GbE, 4x16Gb FC, 2x32Gb FC
Hybrid Cache	24 TiB
RAM	Up to 768GB DDR5
High Availability	> 99.999% Uptime

The system is also optimized for high-throughput workloads such as media production and AI pipelines. For media environments, including 4K and 8K video workflows, the platform supports real-time editing and ingest without relying on proxy workflows. In AI and machine learning use cases, NVMe tiers can be used for active model serving, while HDD tiers provide lower-cost capacity for training datasets and archival data.

Access	Protocols
File	SMBv2, SMBv3, NFSv3, NFSv4 w/RDMA
Block	iSCSI, iSER, FC, NVMe-oF/RDMA
Object	S3-Compatible with Immutable Locking

Data-efficiency features are built into the platform through TrueNAS Adaptive Compression and Fast Deduplication. Compression is applied selectively to reduce capacity consumption without affecting throughput on incompressible data, while deduplication targets redundant data before it is written to disk. These capabilities are included in the base system rather than as licensed add-ons.

The V160 runs TrueNAS Enterprise 25.10, with all major features enabled by default. These include snapshots, replication, multiprotocol access, and integration with TrueCloud backup. TrueNAS maintains a seven-year lifecycle for enterprise deployments, with no required hardware refresh cycles or additional feature licensing during that period. The company positions this as a predictable cost model driven by architecture and media flexibility rather than incremental licensing.

At the filesystem level, the V160 uses OpenZFS, enabling data portability across TrueNAS systems and other OpenZFS-compatible platforms. This approach avoids proprietary data formats and helps organizations retain control over data placement and migration strategies.

The TrueNAS V160 is available immediately. TrueNAS also indicated that its next major software release, TrueNAS 26, is currently in beta and expected to reach enterprise availability later in 2026.

The post TrueNAS Expands Enterprise Portfolio with V160 Hybrid Storage System appeared first on StorageReview.com.

Following its recent collaboration announcement with Google Cloud, NetApp introduced new capabilities to simplify how enterprises use existing data for AI workloads in cloud environments. The updates focus on reducing the operational complexity and cost associated with moving and managing data across hybrid and multi-cloud infrastructures.

At a high level, the joint effort targets a common enterprise challenge. Organizations want to apply AI to existing datasets, but data gravity, fragmentation, and migration overhead often slow adoption. NetApp and Google Cloud are positioning their integration to allow customers to bring data into Google Cloud once and use it across services without repeated movement or duplication.

Google Cloud NetApp Volumes and NetApp Migrator

Central to this approach is Google Cloud NetApp Volumes, which allows enterprises to run file- and block-based workloads in Google Cloud without rearchitecting applications. NetApp stated that customers can migrate existing datasets directly into the service and immediately access Google Cloud-native services, including AI and analytics tools, against that data. This reduces the need for parallel data pipelines and minimizes the latency and cost associated with data duplication.

NetApp also announced general availability of NetApp Data Migrator (NDM), a multi-cloud data migration service designed to move data across environments without requiring specialized expertise. The service is intended to simplify data mobility between on-premises systems and cloud platforms, enabling more consistent access to data for AI and other advanced workloads.

Pravjit Tiwana, Senior Vice President and General Manager of Cloud Storage and Services at NetApp, stated that customers can easily transfer their enterprise data into Google Cloud NetApp Volumes and access Google Cloud services, including AI applications, without moving or duplicating the data. These updates cut costs, delays, and complexities in AI adoption.

Flex Unified Service Level for Google Cloud NetApp Volumes

On the Google Cloud side, the companies highlighted the general availability of the Flex Unified Service Level for Google Cloud NetApp Volumes. This introduces a single storage pool that supports both file and block workloads across all Google Cloud regions. The unified model is designed to support a range of enterprise use cases, including databases, high-performance computing, electronic design automation, and VMware environments, without requiring application changes or separate storage architectures.

Sameet Agarwal, Vice President and General Manager of Storage at Google Cloud, emphasized that AI innovation hinges on AI-ready data supported by flexible, unified architectures that prevent data silos. He highlighted that their partnership with NetApp minimizes data migration barriers, enabling organizations to rapidly leverage Google Cloud’s advanced data and AI capabilities for business innovation.

Overall, the integration extends NetApp’s hybrid cloud storage model into tighter alignment with Google Cloud’s AI and data services. The combined approach focuses on enabling enterprises to retain data within a unified storage layer while leveraging cloud-native AI capabilities, rather than repeatedly moving data between systems.

The post NetApp Expands Google Cloud Integration to Streamline Enterprise Data for AI appeared first on StorageReview.com.

HighPoint Technologies has introduced a new line of PCIe Gen5 x16 NVMe Boot RAID and Switch add-in cards and adapters for server and storage deployments.

The lineup is split into two product families: The Rocket 7600A Series is focused on boot-drive redundancy for Hyper-V and Proxmox systems, using NVMe RAID 1 to help keep hosts online if a boot drive fails. The Rocket 1600 Series is designed for software-defined storage environments, using PCIe Gen5 switching and native NVMe driver support for platforms such as S2D, ZFS, and Ceph.

Rocket 7600A Series

Rocket 7608A

The Rocket 7608A is a PCIe 5.0 x16 NVMe RAID add-in card with eight onboard M.2 slots, each connected via a dedicated PCIe 5.0 x4 lane. It supports up to 8 SSDs in the 2242, 2260, and 2280 form factors and is designed for systems that need higher-density local NVMe storage with RAID support on a single card. HighPoint lists Linux and Windows support for x86 (Intel and AMD) and ARM platforms.

The card supports RAID 0, 1, and 10, as well as single-disk and boot RAID modes. Added management and monitoring features include tools such as WebGUI, CLI, API, and UEFI HII, as well as SMART monitoring, storage health inspection, email alerts, alarm buzzer support, and rebuild controls. Other features include SafeStorage OPAL SED support, hardware secure boot, external power via a PCIe 2×3 connector, and a full-height cooling design with a heatsink, low-decibel fan, and thermal padding.

Main specs

• PCIe 5.0 x16 RAID AIC
• 8x M.2 NVMe ports
• Dedicated x4 lanes per port
• Up to 8 devices
• Supports M.2 2242, 2260, and 2280
• RAID 0, 1, 10, and single-disk
• Boot RAID support
• Linux and Windows support
• Supports x86 Intel/AMD and ARM platforms
• External power support via PCIe standard 2×3 connector
• Full-height, single-width form factor
• Card dimensions: 284mm x 110mm

Rocket 7604A

The Rocket 7604A is a PCIe 5.0 x16 RAID add-in card featuring four onboard M.2 NVMe slots, each with a dedicated x4 PCIe lane. It supports up to four SSDs in 2242, 2260, and 2280 form factors, and is aimed at systems that need dense local NVMe storage without relying on drive bays or cabling. The card supports up to 32TB of total storage and is designed for Linux and Windows environments on x86 Intel, AMD, and ARM platforms.

The card uses Broadcom’s 48-lane PEX89048 PCIe Gen5 switch, with x16 lanes upstream and x4 lanes per M.2 port. HighPoint rates the 7604A for real-world transfer speeds of up to 56GB/s, with support for RAID 0, 1, 10, as well as single-disk configurations, including bootable RAID. Other listed features include sensor logging and analysis, proactive cooling, SafeStorage SED support, firmware-based diagnostics, and customizable firmware options.

Main specs

• PCIe 5.0 x16 RAID AIC
• 4x M.2 NVMe ports
• Dedicated x4 lanes per port
• Up to 4 devices
• Supports M.2 2242, 2260, and 2280
• RAID 0, 1, 10, and single-disk
• Bootable RAID support
• Linux and Windows support
• Supports x86 Intel/AMD and ARM platforms
• Up to 32TB capacity
• Up to 56GB/s real-world transfer speed

Rocket 7628A / Rocket 7628U

The Rocket 7628A and 7628U are PCIe 5.0 x16 NVMe RAID adapters built for larger U.2, U.3, and E3.S SSD deployments. Each card provides four MCIO 8i ports with dedicated x8 lanes per port, supporting up to 8 drives directly and up to 32 drives via a backplane. The cards are designed for server and data center use, with support for Linux and Windows on x86, Intel, AMD, and ARM platforms.

HighPoint lists RAID 0, 1, and 10 modes, along with single-disk mode, boot RAID support, hot-plug and hot-swap capability, out-of-band management over PCIe, and storage health monitoring. The two models share the same core design, but the 7628U is the TAA-compliant version. Other features include SafeStorage OPAL SED support, hardware secure boot, UBM and VPP backplane support, and a low-profile MD2 form factor aimed at enterprise systems.

Main specs

• PCIe 5.0 x16 RAID adapter
• 4x MCIO 8i ports
• Dedicated x8 lanes per port
• Up to 8 direct devices
• Up to 32 devices via backplane
• Supports U.2, U.3, and E3.S NVMe SSDs
• RAID 0, 1, 10, and single-disk
• Boot RAID support
• Linux and Windows support
• Supports x86 Intel/AMD and ARM platforms
• TAA-compliant variant: Rocket 7628U

Rocket 7624A

The Rocket 7624A is a PCIe 5.0 x16 NVMe Pro/RAID adapter with two MCIO 8i ports, each carrying a dedicated x8 connection. It supports up to four NVMe SSDs directly, or up to 16 through a UBM backplane, and is designed for U.2, U.3, E3.S, and E1.S enterprise SSDs. It is a lower-port-count alternative to the 7628A/7628U for workstations, edge servers, and other systems that still need server-grade Gen5 RAID connectivity.

The card supports RAID 0, 1, and 10, as well as single-disk modes, for Linux and Windows, and works with x86 Intel, AMD, and ARM platforms. It uses Broadcom’s 48-lane PEX89048 PCIe switch and can deliver up to 32GB/s with four NVMe SSDs. Other features include hot-pluggable connectivity, SafeStorage SED support, active sensor logging and analysis, proactive cooling, and customizable firmware options.

Main specs

• PCIe 5.0 x16 Pro/RAID adapter
• 2x MCIO 8i (SFF-1016) ports
• Dedicated x8 lanes per port
• Up to 4 direct devices
• Up to 16 devices via UBM backplane
• Supports U.2, U.3, E3.S, and E1.S NVMe SSDs
• RAID 0, 1, 10, and single-disk
• Linux and Windows support
• Supports x86 Intel/AMD and ARM platforms
• Up to 32GB/s with 4 NVMe SSDs

Rocket 1600 Series

Rocket 1608A

The Rocket 1608A is a PCIe 5.0 x16 NVMe switch add-in card featuring eight onboard M.2 slots, each with a dedicated x4 connection for an SSD. It supports up to eight drives in 2242, 2260, and 2280 form factors, and is designed as a plug-and-play NVMe expansion card rather than a RAID product. It is natively supported by mainstream operating systems and compatible with platforms that already include native NVMe driver support, including x86 Intel, AMD, and ARM systems.

The card is designed for dense local NVMe expansion, combining eight device channels on a single full-height, single-width card. It includes external power support through a standard PCIe 2×3 power connector, toolless SSD mounting, self-diagnostic LEDs, FRU support, and hardware secure boot. Cooling is handled by a full-length aluminum heatsink with a low-decibel fan, thermal padding, and copper SSD contact points, which reflects the higher thermal and power demands of an eight-drive Gen5 M.2 design.

Main specs

• PCIe 5.0 x16 switch AIC
• 8x M.2 NVMe ports
• Dedicated x4 lanes per port
• Up to 8 devices
• Supports M.2 2242, 2260, and 2280
• Plug-and-play NVMe storage expansion
• Natively supported by operating systems with NVMe driver support
• Supports x86 Intel/AMD and ARM platforms
• External power support via PCIe standard 2×3 connector
• Full-height, single-width form factor
• Card dimensions: 284mm x 110mm

Rocket 1604A

The Rocket 1604A is a PCIe 5.0 x16 NVMe switch add-in card with four onboard M.2 slots, each linked over a dedicated PCIe 5.0 x4 connection. It supports up to four SSDs in 2242, 2260, and 2280 form factors, and is designed as a plug-and-play NVMe expansion option for systems that rely on native OS NVMe support rather than RAID-specific drivers.

HighPoint lists the card as compatible with x86 Intel, AMD, and ARM platforms, and supports any operating system that includes native NVMe drivers. The Rocket 1604A uses a full-height, half-length design and includes toolless SSD installation, diagnostic LEDs, FRU support, and software secure boot. HighPoint rates the card at up to 64GB/s of data transfer and uses a large heatsink with an integrated 8010 fan to manage thermal performance.

Main specs

• PCIe 5.0 x16 switch AIC
• 4x M.2 NVMe ports
• Dedicated x4 lanes per port
• Up to 4 devices
• Supports M.2 2242, 2260, and 2280
• Up to 64GB/s data transfer rate
• Plug-and-play NVMe storage expansion
• Native NVMe OS support
• Supports x86 Intel/AMD and ARM platforms
• Full-height, half-length form factor
• Card dimensions: 167mm x 110mm

Rocket 1628A

The Rocket 1628A is a PCIe 5.0 x16 NVMe switch adapter designed for high-density enterprise storage expansion. It uses four MCIO 8i ports with dedicated x8 lanes per port, supporting up to eight NVMe SSDs directly or up to 32 through UBM and VPP backplanes. The card supports U.2, U.3, E3.S, and E1.S SSDs and is designed for server, workstation, and datacenter deployments that require additional NVMe connectivity without a separate storage driver stack.

The adapter is built around Broadcom’s 48-lane PEX89048 switch and is natively supported by mainstream operating systems for driverless deployment. HighPoint rates it for up to 60GB/s from a single PCIe 5.0 x16 slot, with hot-swap and hot-plug support, multi-adapter support, and compatibility with x86 Intel, AMD, and ARM platforms. The LP-MD2 form factor is intended to fit standard server and workstation designs.

Main specs

• PCIe 5.0 x16 switch adapter
• 4x MCIO 8i (SFF-1016) ports
• Dedicated x8 lanes per port
• Up to 8 direct devices
• Up to 32 devices via backplane
• Supports U.2, U.3, E3.S, and E1.S NVMe SSDs
• Plug-and-play NVMe storage expansion
• Native OS support for driverless deployment
• Supports x86 Intel/AMD and ARM platforms
• Up to 60GB/s throughput
• Hot-swap and hot-plug support
• LP-MD2 form factor

Rocket 1624A

The Rocket 1624A is a PCIe 5.0 x16 NVMe switch adapter with two MCIO 8i ports, each carrying a dedicated x8 connection. It supports up to four NVMe SSDs directly and up to 16 through a backplane, and is designed for U.2, U.3, E3.S, and E1.S SSDs in enterprise and edge systems. Unlike the RAID models, the 1624A is (as mentioned in the other cards) a plug-and-play expansion card that relies on native NVMe OS support.

HighPoint indicates the 1624A is a broader PCIe expansion option, saying it can connect not only NVMe storage but also up to two PCIe x8 devices, such as GPUs or NICs. The card uses Broadcom’s 48-lane PEX89048 switch and is rated for up to 32GB/s with up to 16 NVMe drives. Other listed features include hot-swap and hot-plug support, multi-adapter support, customizable firmware, and compatibility with Linux and Windows on x86 Intel, AMD, and ARM platforms.

Main specs

• PCIe 5.0 x16 switch adapter
• 2x MCIO 8i (SFF-1016) ports
• Dedicated x8 lanes per port
• Up to 4 direct devices
• Up to 16 devices via backplane
• Supports U.2, U.3, E3.S, and E1.S NVMe SSDs
• Supports up to 2 PCIe x8 devices
• Plug-and-play NVMe storage expansion
• Native OS support for driverless deployment
• Supports x86 Intel/AMD and ARM platforms
• Up to 32GB/s throughput
• Hot-swap and hot-plug support

Rocket 7638D

The Rocket 7638D is a PCIe 5.0 x16 switch adapter built around a hybrid external-and-internal expansion design. It combines one external CDFP-CopprLink port with two internal MCIO 8i ports. It uses a 48-lane Gen5 PCIe switching architecture to provide a total of x32 downstream lanes: x16 to the CDFP connection and x8 to each MCIO port. HighPoint says the 7638D provides a way to link GPUs and NVMe storage more directly within high-performance systems.

Its design is intended to reduce bandwidth bottlenecks in workloads that move large amounts of data between accelerators and storage, including AI, machine learning, HPC, scientific research, and media production. By creating a direct path between hosted GPUs and attached NVMe storage, the Rocket 7638D is aimed at systems where both compute and storage need to run at full speed without competing for limited connectivity. It is compatible with both x86 and ARM platforms.

Main specs

• PCIe 5.0 x16 switch adapter
• 1x external CDFP-CopprLink port
• 2x internal MCIO 8i ports
• 48-lane PCIe Gen5 switching architecture
• x32 downstream lanes total
• x16 to CDFP, x8 to each MCIO port
• Direct GPU-to-NVMe storage pathway
• Built for AI, ML, HPC, scientific research, and media production
• Compatible with x86 and ARM platforms

HighPoint Gen5 Pro/RAID AICs

HighPoint Gen5 Pro/RAID Adapters

HighPoint Gen5 Switch AICs

HighPoint Gen5 Switch Adapters

The post HighPoint launches Gen5 NVMe Boot RAID and Switch Adapters appeared first on StorageReview.com.

At SUSECON in Prague, SUSE introduced SUSE AI Factory with NVIDIA, a pre-validated enterprise AI software stack that simplifies deployment and operations from local development through production. Built on SUSE AI and NVIDIA AI Enterprise, the platform is designed for organizations seeking to build, govern, and scale AI workloads across edge locations, core data centers, and public cloud environments while maintaining tighter control over data and infrastructure.

The announcement positions SUSE AI Factory as a standardized software layer for enterprise AI rather than a standalone model or service. The stack integrates several NVIDIA components, including NIM microservices, open Nemotron models, NeMo for agent development and management, Run:ai for GPU orchestration, NVIDIA Kubernetes Operators, OpenShell for secure agent runtime support, and NemoClaw, which uses SUSE K3s as part of a reference architecture for more secure autonomous agent deployments.

SUSE said the offering is designed to address a common enterprise AI challenge: moving from experimentation to production without introducing excessive operational complexity or compromising governance. According to the company, development teams can build and validate workloads in sandboxed environments. In contrast, platform teams can manage deployments through a Rancher-based interface or GitOps workflows for large-scale rollouts and lifecycle management. The goal is to reduce the number of disconnected tools needed to stand up AI infrastructure and make deployments more repeatable across distributed environments, including air-gapped sites.

From a platform perspective, SUSE is emphasizing a few core areas. First, SUSE is promoting pre-validated blueprints for common AI deployment patterns, which should reduce integration work for enterprise teams assembling custom stacks. Second, SUSE is focusing on zero-trust security and observability, built around SUSE Rancher Prime and SUSE Linux Enterprise Server, with governance controls for NVIDIA-based AI services. Third, SUSE is prioritizing deployment and lifecycle consistency, providing a common operational model from developer workstations to edge clusters. SUSE is also framing the platform around digital sovereignty, supporting organizations that need to keep models, data, and operational control within their private infrastructure to meet internal policies or regulatory requirements, such as the EU AI Act.

Thomas Di Giacomo, Chief Technology and Product Officer at SUSE, discussed the challenges AI teams face. He emphasized the need for balancing rapid innovation with workload security and full auditability before deployment. He also highlighted SUSE AI Factory with NVIDIA as a comprehensive, secure, and sovereign solution supporting both current and future AI development.

John Fanelli, Vice President of Enterprise Software at NVIDIA, noted that Enterprise AI adoption is accelerating, underscoring the need for compliant, secure infrastructure. He explained that their collaboration with SUSE provides an open, comprehensive AI Factory on a secure, sovereign platform that meets technical and regulatory standards.

SUSE AI Factory with NVIDIA is expected later this year.

Digital Sovereignty and Resilience

The product launch was accompanied by new SUSE research on digital sovereignty and digital resilience. In a survey of 309 IT leaders across France, Germany, India, Japan, and the U.S., SUSE found that 98% of respondents consider digital sovereignty a priority. Yet, only 52% said they are actively implementing it. The data suggests that many enterprises now recognize sovereignty as strategically important, particularly as AI adoption accelerates, but execution remains uneven.

Regional differences were notable. Respondents in India reported the highest level of active investment in digital sovereignty at 62%, followed by Germany and Japan at 57%, the U.S. at 52%, and France at 39%. The survey also found that sovereignty is beginning to influence procurement, with 45% of respondents saying it was included in recent RFPs and 42% reporting that it ultimately affected vendor selection. Even so, 41% said action is typically driven only by customer requirements or regulatory pressure, suggesting that external mandates remain a stronger trigger than internal readiness.

SUSE’s survey also linked AI adoption directly to resilience planning. Sixty-four percent of respondents said AI transparency, including visibility into model training and provenance, will be the primary driver of digital resilience over the next five years. At the same time, SUSE found that when given an unexpected 20% budget increase, most organizations would prioritize AI initiatives over investments in sovereignty, underscoring a persistent gap between the urgency of deployment and governance maturity.

When asked how they define digital resilience, respondents consistently emphasized control over systems and infrastructure, as well as recovery posture. Cybersecurity and threat detection ranked highest at 63%, followed by multi-cloud or hybrid diversification at 52%. Backup and recovery at 45% and continuous monitoring at 44% were also key elements of the resilience strategy.

The survey also highlighted the ongoing role of hyperscalers in sovereign infrastructure planning. 65% of respondents said hyperscalers remain relevant for sovereign workloads, reflecting a practical tension between the scale and operational convenience of large cloud providers and the jurisdictional control many enterprises increasingly seek. That dynamic is likely to sustain high demand for open, portable infrastructure stacks that span private, hybrid, and hosted environments without locking customers into a single operational model.

SUSE said the findings reinforce the need for infrastructure that supports both AI adoption and sovereignty requirements, aligning with the positioning of its broader portfolio, including SUSE Linux, SUSE Rancher Prime, and SUSE AI.

The company’s Navigating Digital Resilience report was based on an independently conducted survey across 13 industries, and respondents were not informed that SUSE sponsored the research.

The full report is available from SUSE here.

The post SUSE Unveils AI Factory with NVIDIA, Highlights Enterprise Sovereignty Gap appeared first on StorageReview.com.

IBM introduced a content-aware storage (CAS) architecture that integrates AI data processing directly into the storage layer. The approach targets retrieval-augmented generation (RAG) workflows by embedding document vectorization within the storage system, reducing the need for external preprocessing pipelines.

CAS shifts a core RAG function, document embedding using large language model-based techniques, into storage infrastructure. This enables enterprises to process and index data in place, aligning storage systems with AI-driven workloads and reducing data movement across infrastructure tiers. IBM positions this as a way to simplify deployment while improving performance and data locality for AI applications.

Vector Database at Scale

At the center of IBM’s CAS implementation is a vector database optimized for semantic search. Vector databases support approximate nearest-neighbor search, enabling AI systems to retrieve relevant data chunks based on similarity metrics such as cosine similarity or L2 distance. This capability is foundational to RAG, where user queries are converted into vectors and matched against indexed enterprise data to provide context-aware responses.

IBM Research, working with Samsung and NVIDIA, demonstrated a prototype system capable of scaling to 100 billion vectors on a single server. The system achieved over 90 percent recall and precision, with an average query latency under 700 milliseconds. This scale targets enterprise environments where datasets can span billions of files and, once fully indexed, reach hundreds of billions of vectors.

RAG Pipeline Integration

RAG is emerging as a preferred approach for enterprise AI because it improves output accuracy without requiring model retraining. It operates by augmenting prompts with enterprise-specific data retrieved from a vector database.

The pipeline begins with data ingestion, where documents such as PDFs and presentations are parsed, chunked, and converted into embeddings. These embeddings are stored in a vector database that organizes data for efficient similarity search. During query time, user input is embedded and matched against stored vectors, with relevant content passed to the language model as context. This grounding mechanism reduces hallucinations and improves trust in AI-generated outputs.

IBM’s CAS integrates this pipeline directly into storage, consolidating ingestion, indexing, and retrieval closer to the data.

Addressing Scale and Cost Challenges

Enterprise storage systems already operate at the petabyte scale. When extended to CAS, each file can generate hundreds of vectors, rapidly increasing the dataset size. Traditional vector databases typically scale out across multiple servers, introducing cost and operational complexity. Indexing and reindexing large datasets also become time-intensive.

IBM’s approach focuses on improving vector density and reducing indexing overhead to limit infrastructure sprawl. The architecture decouples vector and index storage from query compute, allowing independent scaling of storage and compute resources. This is enabled by IBM Storage Scale and its high-performance parallel file system.

Storage and Hardware Architecture

The CAS implementation leverages the IBM Storage Scale System 6000 (ESS 6000), an all-flash platform designed for AI and high-performance workloads. The system supports up to 48 NVMe drives per 4U enclosure, with capacities ranging from 7 TB to 60 TB per drive. It integrates PCIe Gen5, 400 Gb InfiniBand, or 200 Gb Ethernet connectivity and delivers up to 340 GB/s read and 175 GB/s write throughput per node, with up to 7 million IOPS.

The platform also supports NVIDIA GPUDirect Storage, enabling direct data paths between storage and GPUs, as well as BlueField-3 DPUs for offloading network and data processing.

Samsung PM9D3a PCIe Gen5 NVMe SSDs provide high-throughput, high-density storage. Based on eighth-generation TLC V-NAND, these drives deliver up to 30.72 TB per device, with sequential read speeds up to 12 GB/s and write speeds up to 6.8 GB/s. The use of commercially available enterprise SSDs allows the architecture to scale using standard components.

Hierarchical Indexing and GPU Acceleration

To address indexing at scale, IBM developed a hierarchical indexing model comprising multiple sub-indexes that can be optimized independently. This structure allows incremental updates and localized reindexing without disrupting the entire dataset, improving both availability and operational efficiency.

GPU acceleration significantly reduces indexing time compared to CPU-only approaches. Tasks that would take hours on CPUs can be completed in minutes using NVIDIA GPUs. In testing, building indexes for 100 billion vectors took 4 days with 6 NVIDIA H200 GPUs, compared to an estimated 120 days on a dual-socket CPU system.

The full dataset, including vectors and indexes, consumed approximately 153 TiB of storage. Initial data loading and partitioning took nine days. The resulting system delivered an average query latency of 694ms with 90% recall, validated against brute-force ground-truth calculations.

Roadmap

IBM and NVIDIA continue to optimize the platform, focusing on reducing indexing and query latency. Current targets include indexing 100 billion or more vectors within a day, reducing data ingestion time from nine days to one day, and lowering query latency to the 50-100 millisecond range while maintaining 90 percent recall.

Integrating vector indexing into standard file systems aims to simplify deployment and lower barriers to enterprise AI adoption. By embedding RAG capabilities directly in storage, IBM is positioning CAS as a foundational layer for AI-enabled infrastructure.

The post IBM Introduces Content-Aware-Storage for RAG Workloads appeared first on StorageReview.com.

Dell Technologies has announced PowerMaxOS 10.4, the latest update to its flagship mission-critical storage platform. The release focuses on measurable performance improvements, enhanced ransomware detection, expanded replication capabilities, and tighter integration with VMware and Red Hat OpenShift.

Performance and Efficiency Improvements

PowerMaxOS 10.4 delivers up to 25% faster read response times for SRDF-protected workloads, addressing a key requirement for environments such as Oracle, SAP, Salesforce, and Epic. The improvement is tied to optimizations in replication-aware I/O handling, benefiting organizations operating in synchronous or asynchronous disaster recovery configurations.

Dell also positions the update as a cost efficiency play for the latest PowerMax 2500 and 8500 systems. The updated node-pair architecture is designed to increase IOPS density while reducing the total cost of ownership. This aligns with enterprise demand for scaling performance without proportional increases in footprint or power consumption.

Cyber Resilience and Data Protection

Security enhancements are a core component of the 10.4 release. PowerMaxOS now includes Advanced Ransomware Detection, which is designed to identify anomalous activity earlier in the attack cycle. The platform also expands identity integration by supporting SSO providers, including Okta, PingFederate, and Microsoft Entra ID, as well as private key support for OIDC workflows. These additions support Zero Trust architectures by tightening access controls without introducing operational friction.

Dell continues to build on its multi-site replication strategy. The platform supports a four-site SRDF configuration that combines SRDF/Metro for active-active replication within a region and SRDF/A for cross-region failover. The addition of SRDF/S as a synchronous option within a region provides more flexibility for consistency-sensitive workloads. The architecture is designed to support automated failover, load balancing, and full-scale recovery, with secure snapshots providing an additional layer of data protection.

VMware and OpenShift Integration

PowerMaxOS 10.4 introduces enhancements to simplify the transition from virtualized to containerized environments. VMware virtual machine migrations can be accelerated by up to 10 times using array-based XCOPY in conjunction with the Red Hat Migration Toolkit for Virtualization. This reduces migration windows and minimizes operational disruption during platform transitions.

For container environments, REST API improvements enable up to 7 times faster provisioning of storage clusters for Red Hat OpenShift. The enhancements are intended to streamline infrastructure deployment and reduce time-to-service for developers and platform teams.

Fabric and Infrastructure Readiness

The release adds support for Connectrix 128Gb Fibre Channel switches and directors, based on Broadcom Gen 8 SAN technology. This upgrade increases available bandwidth and port scalability, addressing growing data center throughput requirements. Integration with Connectrix B-Series Gen 8 fabrics also introduces always-on AES-256 encryption, enhanced cryptographic services, and AI-driven management capabilities.

PowerMaxOS 10.4 further aligns with regulatory requirements through FIPS 140-3 Level 2 certification for TLC flash drives, positioning the platform for use in regulated sectors including finance, healthcare, and government.

Availability

The update is available immediately and targets enterprises running latency-sensitive workloads alongside modernization initiatives.

The post Dell PowerMaxOS 10.4 Includes Performance Gains, Expanded Cyber Resilience, and Modern App Integration appeared first on StorageReview.com.

KIOXIA America has unveiled the EG7 Series, a new family of client SSDs that brings its BiCS FLASH generation 8 QLC with CMOS directly Bonded to Array (CBA) technology to this segment for the first time. KIOXIA says the new series is designed to help PC manufacturers bring high-performance, power-efficient storage to a wider range of systems at a more accessible price point. Capacity options will include 512GB, 1TB, and 2TB models.

KIOXIA EG7 Series Performance and Main Features

The EG7 Series uses 4-bit-per-cell (quadruple-level cell) NAND flash. QLC storage is typically used in lower-cost systems, though it has long faced skepticism about whether it can keep pace with TLC-based drives in everyday performance. KIOXIA is positioning the EG7 Series as a response to that skepticism, saying the new SSDs can deliver TLC-level performance while lowering the total cost of ownership for PC makers building mainstream and more price-conscious systems.

For performance, the drives offer up to 1,000 KIOPS for random reads and writes. Sequential read speeds reach up to 7,000MB/s, while sequential write speeds go as high as 6,200MB/s. This places the EG7 Series in the range of modern PCIe Gen4 client SSDs.

The EG7 Series also supports the NVMe 2.0d specification, which KIOXIA says will provide PC OEMs with added flexibility in device management and overall system design.

KIOXIA EG7 Series Form Factors, Design, and Security

KIOXIA is offering the drives in several M.2 form factors: Type 2230, Type 2242, and Type 2280. That range allows the EG7 Series to fit in compact devices with tight board space, as well as in more conventional notebook and desktop configurations. The smaller 2230 and 2242 formats are important because thin-and-light systems and compact PCs often use shorter SSD layouts.

Metric	512GB	1024GB	2048GB
Sequential Read	6,400 MB/s	7,000 MB/s
Sequential Write	5,000 MB/s	6,000 MB/s	6,200 MB/s
Random Read	550,000 IOPS	850,000 IOPS	1,000,000 IOPS
Random Write	850,000 IOPS	950,000 IOPS	1,000,000 IOPS

The EG7 Series also uses a DRAM-less design. Rather than relying on onboard DRAM, the drives use Host Memory Buffer (HMB) technology, which taps into a portion of system memory to help manage SSD functions. This design is common in more budget-conscious storage products, and can help lower the total cost of ownership and reduce power use without sacrificing responsive performance.

The drives support TCG Opal version 2.02 self-encrypting drive functionality, which is important for business systems that require hardware-based security.

Storage vendors are pushing higher-density flash into a wider range of devices as manufacturers seek more capacity without adding significant system costs. So, the EG7 Series gives OEMs another option for mainstream PCs that need to balance performance, power efficiency, and affordability.

KIOXIA EG7 Series Availability

The EG7 Series is currently sampling with select PC OEM customers. Systems with the new SSDs are expected to begin shipping in the second quarter of 2026.

KIOXIA

The post KIOXIA EG7 Series SSDs Bring QLC Storage to Mainstream PCs appeared first on StorageReview.com.

AMD is once again pushing the boundaries of the high-performance desktop market with the Ryzen 9 9950X3D2 Dual Edition, which launches at an MSRP of $899. When we reviewed the Ryzen 9 9950X3D in March 2025, it made a compelling case as the first 16-core X3D processor, with thermal and TDP constraints no longer forcing a meaningful trade-off between gaming and productivity. It brought 3D V-Cache to a 16-core design, including full overclocking support, and raised the TDP ceiling to deliver sustained performance that earlier X3D chips couldn’t match. The 9950X3D2 builds on that foundation, extending 3D V-Cache across both CCDs for the first time and increasing the total L3 cache from 128MB to 192MB. AMD provided us with a sample for evaluation against the full 9000-series X3D stack.

AMD Ryzen 9 9950X3D2: Solving the Asymmetry Problem

The core problem the 9950X3D2 solves is one the 9950X3D never fully escaped. Because the 9950X3D applied 3D V-Cache to only one of its two CCDs, threads migrating between dies during normal Windows load balancing would periodically lose access to the cache-rich CCD, causing unpredictable latency spikes. AMD’s chipset drivers helped manage this, but the asymmetry remained. The 9950X3D2 eliminates it. Each CCD combines 32MB of traditional 2D L3 cache with a 64MB 3D V-Cache stack, giving both CCDs an identical 96MB L3 pool and all 16 cores symmetrical, low-latency access to a combined 192MB total. For workloads sensitive to memory latency, particularly high-FPS gaming, this is a meaningful architectural improvement rather than a simple spec bump.

The underlying 2nd Gen 3D V-Cache design is the same as the under-die architecture introduced with the 9800X3D and carried through to the 9950X3D, with cache placed beneath the compute cores to keep the primary heat source close to the cooling solution. What changes with the 9950X3D2 is scope: that design now covers both CCDs, and the TDP rises from 170W to 200W to support the additional sustained throughput. Total on-chip cache reaches 208 MB across L2 and L3, up from 144 MB on the 9950X3D.

AMD Ryzen 9 9950X3D2 Specifications

Specifications	AMD Ryzen 9 9950X3D2	AMD Ryzen 9 9950X3D	AMD Ryzen 9 9900X3D	AMD Ryzen 7 9850X3D	AMD Ryzen 7 9800X3D
Cores/Threads	16/32	16/32	12/24	8/16	8/16
Platform	AM5	AM5	AM5	AM5	AM5
Max Boost / Base Clock	5.6 / 4.3GHz	5.7 / 4.3GHz	5.5 / 4.4GHz	5.6 / 4.7GHz	5.2 / 4.7GHz
L2 Cache	16MB	16MB	12MB	8MB	8MB
L3 Cache	192MB	128MB	128MB	96MB	96MB
Total Cache	208MB	144MB	140MB	104MB	104MB
Architecture	Zen 5	Zen 5	Zen 5	Zen 5	Zen 5
PCIe	Gen5	Gen5	Gen5	Gen5	Gen5
DRAM	DDR5	DDR5	DDR5	DDR5	DDR5
TDP / Default Socket Power (PPT)	200W / 270W	170W / 230W	120W / 230W	120W / 162W	120W /162W
Graphics	Radeon	Radeon	Radeon	Radeon	Radeon
AMD Recommended Cooler	Liquid cooler	Liquid cooler	Liquid cooler	Liquid cooler	Liquid cooler

Platform and Compatibility

The 9950X3D2 slots into the AM5 ecosystem without requiring a platform change. Like the 9950X3D, it supports existing A620, B650/B650E, X670/X670E, X870/X870E, B840, and B850-class motherboards with a BIOS update, making it a straightforward upgrade for users already invested in the platform. The higher 200W TDP does demand more from the cooling side, however. While the 9950X3D can be managed with a capable 240mm AIO, AMD recommends a 360mm liquid cooler for the 9950X3D2 to maintain sustained boost performance under heavy workloads.

AMD Ryzen 9 9950X3D2 Performance

To evaluate overall performance, we compared the AMD Ryzen 9 9950X3D2 against the AMD Ryzen 9 9950X3D, Ryzen 7 9850X3D, and Ryzen 7 9800X3D. While all four processors feature AMD’s 3D V-Cache design, the two Ryzen 9 models sit in a higher-performance tier, sharing a 16-core, 32-thread configuration. The Ryzen 7 chips, with 8 cores and 16 threads, sit a step below, with performance differences becoming more apparent in heavily threaded workloads while remaining relatively close in lighter tasks. All testing was conducted at stock settings (no overclocking) to ensure a consistent baseline across the stack.

AMD Consumer Test Platform

To keep the testing environment as consistent as possible, all CPUs have been tested across X870E-based motherboards at stock settings. The only changes above stock settings have been the same DDR5 memory and EXPO configuration. Here’s a full rundown of our testing rig in this review:

• Motherboard:  ASRock X870E Taichi (provided by AMD)
• Memory: G.SKILL Trident Z5 Royal Series DDR5-6000 (2x16GB), running on EXPO 1
• Cooling: NZXT Kraken Elite 360
• Operating System: Windows 11 Pro

3DMark CPU Profile

The 3DMark CPU Profile measures CPU performance across different workloads by testing 1, 2, 4, 8, 16, and max threads. It highlights how the CPU handles single-threaded tasks, gaming workloads, and multithreaded applications such as 3D rendering. The benchmark minimizes GPU impact, offering a clear view of the CPU’s performance in various scenarios.

In the 3DMark CPU Profile benchmark, the Ryzen 9 chips most clearly separate themselves as thread counts increase. The 9950X3D2 tops the chart with 17,672 points in the Max Threads test, about 6% ahead of the 9950X3D, while the 9950X3D still holds a sizable lead over the Ryzen 7 9850X3D and 9800X3D by roughly 63% and 67%, respectively. That gap narrows quickly under lighter workloads, when all four chips are much closer together, but the ranking still favors the two Ryzen 9 processors overall.

3DMark CPU Profile (higher is better)	AMD Ryzen 9 9950X3D2	AMD Ryzen 9 9950X3D	AMD Ryzen 7 9850X3D	AMD Ryzen 7 9800X3D
Max Threads	17,672	16,690	10,261	10,018
16 Threads	16,956	15,983	10,285	10,034
8 Threads	9,141	9,070	8,611	8,269
4 Threads	4,980	4,846	4,867	4,646
2 Threads	2,508	2,521	2,487	2,394
1 Threads	1,274	1,264	1,267	1,213

y-cruncher

y-cruncher is a popular benchmarking and stress-testing application that launched in 2009. This test is multithreaded and scalable, computing Pi and other constants up to the trillions of digits. Faster is better in this test.

In y-cruncher, both Ryzen 9 chips show a clear advantage in this long-running computational workload. The 9950X3D2 completes the 1-billion-digit test in 12.605 seconds, roughly 31% faster than the 9950X3D, which itself is about 12% faster than the 9850X3D and 31% faster than the 9800X3D. As the workload grows, the lead widens further, with the 9950X3D2 completing the 5 billion run about 41% faster than the 9950X3D, reinforcing its stronger sustained compute performance.

y-cruncher (lower time is better)	AMD Ryzen 9 9950X3D2	AMD Ryzen 9 9950X3D	AMD Ryzen 7 9850X3D	AMD Ryzen 7 9800X3D
1 Billion	12.605 s	16.450 s	18.503 s	21.487 s
2 Billion	34.925 s	48.047 s	52.589 s	64.273 s
5 Billion	77.370 s	109.343 s	115.581 s	143.891 s

y-cruncher BBP

This y-cruncher benchmark uses the Bailey-Borwein-Plouffe (BBP) formulas to compute a large number of hexadecimal digits of Pi, measuring the CPU’s total computation time, utilization, and multi-core efficiency.

Looking at the y-cruncher BBP test, the Ryzen 9 9950X3D2 again sets the pace, completing the 100 BBP run in 47.07 seconds, about 7% faster than the 9950X3D. The non-D2 9950X3D still maintains a major lead over both Ryzen 7 chips, finishing that same workload about 66% faster than the 9850X3D and 66% faster than the 9800X3D. Across the full sweep, the order stays consistent, with the two Ryzen 9 processors comfortably ahead.

y-cruncher BBP (lower time is better)	AMD Ryzen 9 9950X3D2	AMD Ryzen 9 9950X3D	AMD Ryzen 7 9850X3D	AMD Ryzen 7 9800X3D
1 BBP	0.384 s	0.426 s	0.669 s	0.671 s
10 BBP	4.173 s	4.538 s	7.501 s	7.497 s
100 BBP	47.070 s	50.291 s	83.719 s	83.345 s

Maxon Cinebench

Cinebench is a widely used benchmarking tool that measures the performance of CPUs and GPUs by rendering with Maxon Cinema 4D. It provides a score that allows you to compare the performance of different systems and components. We ran R23 and R24, both popular Cinebench versions, so you can compare the results with those on popular online leaderboards.

In Cinebench, the separation between the Ryzen 9 and Ryzen 7 parts is immediately clear in multi-core performance, while single-core results remain much tighter across the stack. In Cinebench R23, the Ryzen 9 9950X3D2 leads with a score of 42,555, about 6% ahead of the 9950X3D, while both Ryzen 9 chips nearly double the performance of the Ryzen 7 models, holding roughly an 87–99% advantage in multi-core workloads. Cinebench R24 shows the same trend, with the 9950X3D2 reaching 2,508, about 12% ahead of the 9950X3D and again maintaining a significant 80%+ lead over the Ryzen 7 parts.

Single-core results tell a different story. In R23, all three newer chips cluster closely, with the 9950X3D2 holding only about a 2% lead over the 9950X3D and effectively tying the 9850X3D. R24 tightens even further, where the 9950X3D2 and 9850X3D are nearly identical, and the 9950X3D trails slightly. This consistency highlights that lightly threaded performance is broadly similar across the lineup, with only small gains at the top end.

Cinebench R23

Cinebench R23 (higher is better)	AMD Ryzen 9 9950X3D2	AMD Ryzen 9 9950X3D	AMD Ryzen 7 9850X3D	AMD Ryzen 7 9800X3D
Multi-Core	42,555	39,993	21,382	22,718
Single-Core	2,248	2,200	2,216	2,089

Cinebench R24

Cinebench R24 (higher is better)	AMD Ryzen 9 9950X3D2	AMD Ryzen 9 9950X3D	AMD Ryzen 7 9850X3D	AMD Ryzen 7 9800X3D
Multi-Core	2,508	2,246	1,366	1,338
Single-Core	143	134	142	130

7-Zip Compression

The 7-Zip Compression Benchmark evaluates CPU performance during compression and decompression, measuring GIPS (Giga Instructions Per Second) and CPU usage. Higher GIPS and efficient CPU usage indicate superior performance.

In 7-Zip, the 9950X3D2 achieves the highest overall score, with a total rating of 233.09 GIPS, about 9% ahead of the 9950X3D. The non-D2 9950X3D still holds a commanding advantage over the Ryzen 7 chips, outperforming the 9850X3D by roughly 64% and the 9800X3D by about 69% in total rating. Compression and decompression follow the same general pattern, with the two Ryzen 9 processors well out in front.

7-Zip Compression	AMD Ryzen 9 9950X3D2	AMD Ryzen 9 9950X3D	AMD Ryzen 7 9850X3D	AMD Ryzen 7 9800X3D
Compressing
Current CPU Usage	2,736%	2,737%	1,394%	1,387%
Current Rating/Usage	7.132 GIPS	6.565 GIPS	8.864 GIPS	8.488 GIPS
Current Rating	195.145 GIPS	179.648 GIPS	123.563 GIPS	117.745 GIPS
Resulting CPU Usage	2,717%	2,727%	1,390%	1,393%
Resulting Rating/Usage	7.186 GIPS	6.531 GIPS	8.852 GIPS	8.466 GIPS
Resulting Rating	195.272 GIPS	178.094 GIPS	123.073 GIPS	117.895 GIPS
Decompressing
Current CPU Usage	3,148%	3,034%	1,564%	1,570%
Current Rating/Usage	8.674 GIPS	8.207 GIPS	8.821 GIPS	8.365 GIPS
Current Rating	273.103 GIPS	248.987 GIPS	137.919 GIPS	135.527 GIPS
Resulting CPU Usage	3,134%	3,036%	1,567%	1,564%
Resulting Rating/Usage	8.643 GIPS	8.242 GIPS	8.820 GIPS	8.663 GIPS
Resulting Rating	270.917 GIPS	250.233 GIPS	138.223 GIPS	135.448 GIPS
Total Rating
Total CPU Usage	2,926%	2,882%	1,479%	1,478%
Total Rating/Usage	7.915 GIPS	7.387 GIPS	8.836 GIPS	8.564 GIPS
Total Rating	233.094 GIPS	214.163 GIPS	130.648 GIPS	126.671 GIPS

UL Procyon

UL Procyon AI Inference is designed to gauge a workstation’s performance in professional applications. It should be noted that this test does not leverage multiple CPU capabilities. Specifically, this tool benchmarks the workstation’s ability to handle AI-driven tasks and workflows, providing a detailed assessment of its efficiency and speed in processing complex AI algorithms and applications.

UL Procyon shows a tighter spread, but the overall hierarchy still favors the Ryzen 9 chips. The 9950X3D2 posts the top overall AI Computer Vision score at 271, about 23% ahead of the 9950X3D, while the 9950X3D itself remains 5% ahead of the 9850X3D and 17% ahead of the 9800X3D. Model-level results are more mixed, particularly in lighter tasks like MobileNet V3. Still, the two Ryzen 9 parts pull further apart in heavier inference workloads such as YOLO V3 and REAL-ESRGAN.

UL Procyon (higher score & lower ms is better)	AMD Ryzen 9 9950X3D2	AMD Ryzen 9 9950X3D	AMD Ryzen 7 9850X3D	AMD Ryzen 7 9800X3D
Overall AI Computer Vision Score	271	220	209	188
MobileNet V3	0.97 ms	0.94 ms	0.70 ms	0.61 ms
ResNet 50	3.76 ms	5.33 ms	5.95 ms	7.01 ms
Inception V4	13.90 ms	17.12 ms	19.34 ms	22.28 ms
DeepLab V3	19.26 ms	21.70 ms	20.40 ms	23.98 ms
YOLO V3	24.93 ms	35.27 ms	48.17 ms	56.07 ms
REAL-ESRGAN	1,593.81 ms	2,037.51 ms	2,348.97 ms	2,728.62 ms

PCMark10

PCMark 10 evaluates CPU performance by simulating real-world office productivity tasks like word processing, web browsing, video conferencing, and spreadsheet calculations. The benchmark combines workloads that reflect the demands of modern workplaces, providing a comprehensive assessment of how a CPU handles day-to-day applications.

PCMark 10 compresses the gap more than any of the heavier compute-focused tests. Interestingly, the non-D2 Ryzen 9 9950X3D actually posts the highest overall score at 10,849, edging out the 9950X3D2 by about 1.8%. Even so, both Ryzen 9 chips remain ahead of the Ryzen 7 9850X3D and 9800X3D, showing that everyday productivity performance is broadly strong across the stack with only small differences at the top.

PCMark10 (higher is better)	AMD Ryzen 9 9950X3D2	AMD Ryzen 9 9950X3D	AMD Ryzen 7 9850X3D	AMD Ryzen 7 9800X3D
Overall Score	10,650	10,849	10,461	10,250

SPECworkstation 4.4.0

SPECworkstation 4 specializes in benchmarks designed to test all key aspects of workstation performance. It uses over 30 workloads to test CPU, graphics, I/O, and memory bandwidth. The workloads fall into broader categories, including Media and Entertainment, Financial Services, Product Development, Energy, Life Sciences, and General Operations. We will list each broad-category result instead of the individual workloads. The results are averages of all individual workloads in each category.

In SPECworkstation 4.4.0, the 9950X3D2 leads most categories, but the non-D2 9950X3D remains firmly in second and well ahead of the Ryzen 7 parts in most professional workloads. In AI & Machine Learning, the 9950X3D2 scores 3.96, about 20% ahead of the 9950X3D, while the 9950X3D still leads the 9850X3D by roughly 12%. Some categories tighten considerably, such as Media & Entertainment and Life Sciences, but the overall pattern still puts the two Ryzen 9 chips ahead.

SPECworkstation 4.4.0 (higher is better)	AMD Ryzen 9 9950X3D2	AMD Ryzen 9 9950X3D	AMD Ryzen 7 9850X3D	AMD Ryzen 7 9800X3D
AI & Machine Learning	3.96	3.30	2.95	2.92
Energy	3.22	2.66	2.20	2.13
Financial Services	2.63	2.48	1.42	1.42
Life Sciences	2.62	2.71	2.11	2.15
Media & Entertainment	3.39	3.34	2.56	2.57
Product Design	2.75	2.43	2.14	2.08
Productivity & Development	1.39	1.28	1.14	1.12

Conclusion

The AMD Ryzen 9 9950X3D2 is not just an iteration; it is the point where AMD fully resolves the trade-offs that defined earlier X3D designs. By eliminating the asymmetric cache layout and extending 3D V-Cache across both CCDs, AMD has transformed what was once a situational advantage into a consistent, system-wide benefit. Every core now has equal access to a massive 192MB L3 pool, removing scheduling penalties and delivering the predictability high-end workloads demand.

The 9950X3D2 led in nearly every benchmark. Whether in heavily threaded compute like y-cruncher, rendering in Cinebench, or compression in 7-Zip, the 9950X3D2 repeatedly edges ahead of the 9950X3D. The gains span across nearly every category, reinforcing that this refinement meaningfully improves sustained performance rather than chasing peak numbers.

At the platform level, it also represents the ceiling of what AM5 can currently deliver. With drop-in compatibility, it gives existing users a clear upgrade path to the most balanced high-end desktop CPU AMD has produced to date. The higher 200W TDP and cooling requirements are the only real trade-offs, but they are proportional to the level of performance it offers.

Ultimately, the 9950X3D2 earns its place not by redefining the category, but by perfecting it. It takes the hybrid identity of X3D processors, part gaming chip, part workstation CPU, and removes the friction between those roles. For users who want top-tier gaming performance without sacrificing multithreaded capability, or vice versa, this is the first X3D processor to truly deliver on both fronts.

AMD Ryzen 9 9950X3D2 – Product Page

The post AMD Ryzen 9 9950X3D2 Dual Edition Review: 3D V-Cache on Both CCDs appeared first on StorageReview.com.

NetApp announced an expanded collaboration with Google Cloud, formalized through a four-year enterprise agreement to accelerate the deployment of NetApp storage within Google Distributed Cloud (GDC) Air-Gapped environments. Delivered with World Wide Technology (WWT), the offering targets sovereign cloud use cases that require strict data residency, security, and operational isolation.

The joint solution integrates NetApp’s data platform with Google Distributed Cloud’s full-stack private cloud architecture. The result is an air-gapped environment that supports sensitive and classified workloads while maintaining compliance with national sovereignty requirements. NetApp positions its storage systems as secure-by-design, enabling organizations to deploy controlled infrastructure that supports modern applications and AI workflows without external connectivity.

NetApp integrates its AFF all-flash systems, StorageGRID object storage, and Trident Kubernetes storage orchestration into the GDC stack. Together, these components form what the company calls an intelligent data infrastructure. Within GDC, this architecture supports zero-trust security models, local data storage, customer-managed encryption keys, and full operational control. The platform enables organizations to extend cloud capabilities to on-premises or edge environments while maintaining isolation, or to operate in fully disconnected, air-gapped configurations.

The collaboration is primarily aimed at government and regulated industries, where data-handling requirements limit the use of traditional public cloud. NetApp leadership highlighted that these environments require infrastructure capable of handling classified data while supporting modernization initiatives. By integrating with GDC, NetApp enables enterprise-grade AI and analytics capabilities within accredited environments, allowing agencies to derive insights and automate processes without compromising compliance or sovereignty.

Google Distributed Cloud is designed to extend Google Cloud services to customer-controlled locations, including on-premises data centers and edge sites. Google noted that public-sector organizations face growing pressure to extract value from data while complying with strict regulatory frameworks. GDC addresses this by enabling the deployment of cloud-native services and advanced AI in sovereign and disconnected environments.

As part of this effort, Google has expanded the availability of its AI capabilities for regulated use cases. Gemini models are now supported in GDC environments, enabling generative AI functions such as automation, content generation, discovery, and summarization directly on-premises. These capabilities can run in fully disconnected deployments, allowing organizations to leverage advanced AI while maintaining strict security and compliance boundaries.

The NetApp and Google Cloud partnership reflects a broader trend of bringing cloud and AI capabilities into controlled environments. By combining enterprise storage with sovereign cloud infrastructure, the companies are targeting organizations that require both advanced data services and strict operational isolation.

The post NetApp Expands Google Cloud Collaboration for Sovereign, Air-Gapped Deployments appeared first on StorageReview.com.

Comino recently sent us the latest version of the Comino Grando for review, configured with eight NVIDIA RTX PRO 6000 Blackwell cards, each with 96GB of VRAM, for a total of 768GB of GPU memory. We reviewed the Comino back in in 2024, configured with 6x RTX 4090s, offering 144GB of total GPU memory, as well as a version with NVIDIA H100’s. This latest update marks a substantial generational leap in both raw memory capacity and the range of workloads the platform can address. 

The Grando is a purpose-built 4U platform designed to resolve the critical conflict between high-density GPU compute and thermal management. While standard air-cooled chassis crumble under the sustained 600W+ TDP demands of modern professional cards, the Grando takes a fundamentally different approach, built from the ground up around a liquid-cooled architecture capable of dissipating a massive 6.5kW of continuous heat. This is not a retrofit or an afterthought; the entire chassis, from its inverted motherboard layout to its color-coded quick-disconnect manifold system, has been engineered around the cooling loop.

The result is a platform that can sustain eight full-TDP professional GPUs in a single 4U chassis, running 24/7 in ambient environments of 3-38°C, without thermal throttling, without the acoustic assault of high-RPM air cooling, and without compromising serviceability. For organizations deploying AI inference, machine learning training, or high-performance simulation workloads at scale, the Grando offers something genuinely rare: a server that does not ask you to choose between density, thermals, and reliability.

Comino Grando Specifications

The table below shows the physical specifications and supported hardware configurations for the Comino Grando platform.

Specification / Feature	Comino Grando
Comino Grando Server & Rackable Workstation
Cooling Capacity	6.5kW (Maximum 6 500 W @ 20°C intake air T)
Motherboards	Up to EATX & EBB
GPUs (Server)	Up to 8; NVIDIA: RTX A6000, RTX 6000 ADA, RTX PRO 6000, A40, L40, L40S, A100, H100, H200
GPUs (Rackable Workstation)	Up to 6; NVIDIA: 3090, 4090, 5080, 5090, RTX A6000, RTX 6000 ADA, RTX PRO 6000, A40, L40, L40S, A100, H100, H200; AMD: W7800, W7900
CPUs	Up to 2; Single Socket: Intel Xeon W-2400/2500 & 3400/3500, Intel Xeon Scalable 4th Gen, 5th Gen, XEON 6, AMD Threadripper PRO 5000WX, 7000WX, 9000WX, AMD EPYC 9004/9005 Dual Socket: Intel Xeon Scalable 4th Gen, & 5th Gen, XEON 6, AMD EPYC 9004/9005
RAM	Up to 2TB
M2 drives	Up to 8x NVME
Storage	Back panel hot swap cages: up to 4x hot swap SSDs (4x 7mm or 2x 15mm) and up to 4 more (4x 7mm or 2x 15mm) instead of 4th PSU; Internal 3.5″ cage up to 4x 3.5″ or 4x 2.5″ 15mm or 12x 2.5″ 7mm; Internal 2.5″ slots: up to 4x 2.5″ SSD 7mm
Power Supply & Operating Voltage	Up to 4x 2000W Hot Swap CRPS @ 180-264V Up to 4x 1000W Hot Swap CRPS @ 90-140V Redundancy modes: 4+0, 3+1, 2+2
Noise level	39dB-70dB
Lan	Up to 2x 10 Gbit/s on the MoBo and up to 400Gbit/s in PCIe
OS	Ubuntu / Windows 11 (Pro/Home) / Windows Server
Physical & Cooling Specifications
Liquid cooling	CPU with VRM and GPU with GDDR and VRM
Reservoir	Comino custom 450ml with integrated pumps
Fans	3x Ultra High Flow 6200RPM (high noise level) or 3x High Flow 3000RPM (low noise level)
Installation	19″ rack-mountable or standalone as a Workstation
Required rack space	4U
Size	439 x 681 x 177mm (without handles and protruding parts)
Weight	4 GPUs: 49kg (net), 67kg (gross) 6 GPUs: 52kg (net), 70kg (gross) 8 GPUs: 55kg (net), 72kg (gross)
Operating & storage temperature range	Storage: -5..50°C / 23..122°F Operating: 3..38°C / 38..100°F
Comino Monitoring System (CMS)
Overview	Controller Board with Sensors & Software for Real-Time Monitoring
Key Advantages	Cooling System & CPU/GPU Monitoring, Web Interface, Cooling System Log, Centralized Monitoring for Workgroups
Sensors & Connected Devices	Temperature (air and coolant), % Humidity, Voltage, Coolant flow, Reservoir coolant level, Fans, Pumps, Motherboard, Display, and buttons
Integration Possibilities	Establish monitoring via a REST API and push sensor data to monitoring software (e.g., Zabbix, Grafana) or databases (e.g., InfluxDB).
CMS Technical Requirements
OS	Windows 11/10 Ubuntu 22.04/20.4 (Dependency for Ubuntu: the target system must have nvidia-smi and sensors utilities installed)
Web Browsers	Mozilla Firefox, Google Chrome, Chromium, Apple Safari, Microsoft Edge (Attention: Internet Explorer 11 is not supported)
Hard disk drive	300MB
Controller firmware version	1.0.6 or newer
Controller PCB version	2.xx.xx

Design, Build, and GPU Density

Chassis Layout and Deployment

The Grando Server is a masterclass in space optimization, measuring 17.3 x 26.8 x 6.97 inches (4U). Unlike traditional servers, it places the motherboard’s rear at the front of the chassis, inverting the conventional internal layout. This ensures that air-cooled components, such as RAM modules and VRMs, receive the coldest possible intake air before it reaches the liquid-cooling radiator at the rear.

The chassis itself is built to the same exacting standard, featuring solid steel construction with a matte black powder-coat finish applied inside and out. This deliberate choice extends to the tubing, cables, radiator, and PCB solder mask, reflecting a clear intention for a clean, professional aesthetic throughout. Furthermore, the system supports versatile deployment, functioning seamlessly as either a 19-inch rack-mountable unit or a standalone desktop unit. Depending on the configuration, it weighs between 148 and 159 lbs.

GPU Cold Plates and Water Blocks

The proprietary copper water blocks form the core of the Grando’s density, cooling not only the GPU die but also the other components like memory and voltage regulators. Each GPU ships as an off-the-shelf card, on which Comino mounts a custom cold-plate assembly. In practice, this thin-profile design reduces each card to a single-slot footprint, allowing six or even eight professional GPUs to sit side by side within a single 4U chassis. Our review unit shipped with eight NVIDIA RTX PRO 6000 Blackwell cards, each with a TDP of 600W, resulting in a total cooling requirement of 4,800W under full load.

Achieving the Comino’s 8 single slot GPU density would be nearly impossible with air cooling, since stock NVIDIA RTX PRO 6000 cards each occupy two slots and require substantial airflow. In contrast, these custom-cooled cards occupy just one slot each. The cold plates are built solidly, adding noticeable weight to each card, but that weight reflects the quality and cooling performance required at this level.

Each pair of GPUs is plumbed through a dedicated sub-manifold that consolidates both cards into a single inlet and outlet connection to the main coolant manifold. This paired approach simplifies the overall loop architecture, reduces the number of connections at the main manifold, and allows a technician to disconnect a single pair of quick-disconnect couplings to remove two cards at once, further streamlining maintenance.

Water Distribution and Manifold

At the center of the system sits a large water distribution manifold that supplies cool liquid to each GPU and CPU cold plate and provides the return path to the radiator. All connections between the manifold and the GPU’s and CPU use Comino’s “TheQ” Quick Disconnect Couplings. These stainless steel dripless fittings are color-coded with red and blue rings to clearly identify the hot and cold sides of the loop, removing any ambiguity during installation or servicing.

They leave minimal residue on the mating surface when disconnected, allowing technicians to remove or replace individual GPUs or the CPU without draining the 450ml reservoir or the rest of the loop. In this way, the Grando brings the maintenance simplicity of air-cooled systems to a high-performance liquid-cooled platform.

CPU Cooling and Memory

The CPU and its voltage regulators also benefit from a dedicated cold plate connected directly to the coolant loop, preventing the processor from becoming a bottleneck during intense multi-GPU workloads. Our review unit shipped with an AMD Turin/Genoa board featuring a single AMD EPYC 9474F 48-core processor. The cold plate mirrors the quality of the card cold-plates, machined from solid copper and secured with stainless-steel hardware.

Flanking the CPU on both sides are eight fully populated DRAM slots that support configurations up to 2TB of RAM. Our review unit came equipped with 512GB of DDR5 RAM. A support bar spans above the GPU and CPU area of the chassis, perpendicular to them, securing sensitive components like the GPU’s and maintaining chassis rigidity during transport.

Radiator and Fans

Cooling is handled by a large triple 140mm radiator mounted at the rear of the chassis, paired with three high-speed 140mm fans capable of reaching 6,200 RPM and moving up to 1,000 m³/h of airflow. The dense fin stack provided by the thick radiator underscore the thermal headroom designed into the platform, which is built to dissipate up to 6.5kW of sustained heat in our configuration.

What is perhaps most surprising is that despite that workload and those fan speeds, the unit manages to stay within a tolerable noise envelope, with sound levels sitting 70+ dB at full tilt. That is loud by workstation standards but notably restrained for a system dissipating the thermal output of a small electric furnace, which speaks to how effectively the Comino’s liquid loop transfers heat away from the components.

Front Panel and Telemetry Display

On the front panel, an LED display provides a live readout of key telemetry data, including pump status, ambient air temperature, coolant temperature, and fan speed. Users navigate the menu using illuminated buttons on the cooling module, with short presses to scroll through available data. A long press on the PB2 button opens additional menu branches, including Commands, Service settings, and an Event Log. In addition, the front I/O panel includes a VGA port for display output, alongside a serial port, multiple USB ports, and network connections for peripheral and device connectivity.

Power and Storage Architecture

Power Delivery and Redundancy

Supporting this level of compute requires equally robust power delivery. The Grando supports up to four hot-swap 1000W or 2000W CRPS modules in a redundant configuration, delivering up to 8.0kW at 180–264V. With support for 4+0, 3+1, and 2+2 redundancy modes, the system can tolerate PSU failures while maintaining continuous operation for 24/7 AI and HPC workloads.

Our review unit shipped with four Great Wall 2000W 80 Plus Platinum hot-swap power supplies, forming the full 8.0kW configuration.

Power delivery to each GPU runs through a centralized 12-pin power distribution board mounted between the GPU array and the main cable run. The Grando uses this distribution board to consolidate incoming power feeds and then branch them to each GPU in an organized, space-efficient manner.

PCIe, Storage, and Networking

The Grando comfortably supports six GPUs without compromising slot bandwidth, and the chassis scales to a full eight-card configuration for maximum density. The Comino’s ASRock Rack GENOAD8X-2T/BCM motherboard provides seven x16 and one x8 PCIe Gen 5 slots, meaning seven of the eight GPUs run at full x16 bandwidth with the eighth card operating at x8. This is a trade-off between the number of PCIe lanes a single-socket CPU can support and Comino’s reluctance to add the size, cost, and complexity of a PCIe switch plate. Moving to a dual-socket motherboard would provide more PCIe lanes but offer even fewer slots, since the 2nd socket would occupy space otherwise used by PCIe slots in the space-constrained form factor.

Running eight GPUs in a single-socket system consumes the lion’s share of available PCIe lanes, and that comes with trade-offs. Our review unit, based on AMD Genoa, has 128 PCIe Gen 5 lanes available in total. With eight GPUs consuming 120 of those lanes, the remaining 8 lanes are split x4 to each M.2 SSD slot, so it is not possible to simultaneously run eight GPUs and a full complement of NVMe drives in the rear of the chassis connected via the two MCIO connectors. In our full 8-GPU configuration, only 2 M.2 slots were available for storage. Administrators who need additional NVMe capacity alongside maximum GPU density should be aware that adding rear hot-swap NVMe storage via the back-panel cages will consume additional PCIe lanes and disable some GPU capacity in their system.

With that said, storage is equally modular and expansive, though the configuration does affect the PCIe lane budget for GPUs, which is worth planning around for the intended use case. The rear panel of our review unit features a 2.5″ drive cage that supports up to four 2.5-inch SSDs in either 4x 7mm or 2x 15mm configurations, with an optional second set of up to four available in place of the fourth PSU slot. Because our review unit required all four power-supply bays to support the full 8-GPU configuration, we had access only to the first of the two hot-swap bays. Internally, the chassis can support a 3.5-inch cage that accommodates up to four 3.5-inch drives, four 2.5-inch 15mm drives, or up to twelve 2.5-inch 7mm drives, plus four additional internal 2.5-inch 7mm SSD slots if configured.

For networking, two onboard RJ45 10 Gb/s ports powered by the Broadcom BCM57416 are standard on the motherboard, alongside a dedicated Gigabit Ethernet IPMI management port. Administrators can further increase bandwidth by installing PCIe NICs that support up to 400 Gb/s for high-bandwidth fabric connectivity, though note that additional PCIe NICs occupy GPU slots, reducing the maximum number of GPUs the system can host.

Remote Management and System Intelligence

To safeguard the hardware and optimize performance, the system includes the Comino Monitoring System (CMS). A separate, autonomous controller board drives the CMS and serves as the server’s “brain,” independent of the main operating system. In practice, this controller reads a comprehensive array of sensors that track air and coolant temperatures, humidity levels, coolant flow rates, and reservoir levels in real time. Crucially, this autonomous design enables the CMS to perform self-diagnosis and trigger emergency shutdowns upon detecting a leak or a pump failure, protecting the expensive internal hardware from damage.

A web-based GUI handles day-to-day management, providing administrators with clear visibility into cooling performance, uptime, and real-time energy consumption for the CPU and GPUs. For enterprise-scale deployments, the CMS also connects to centralized monitoring tools via REST APIs, such as Zabbix, Grafana, and InfluxDB. Together, these capabilities help administrators maintain a 3-year interservice period and keep the server running at peak efficiency without thermal throttling, even in high-ambient environments.

Beyond AI: Creative and Engineering Applications

While our testing focused on AI inference workloads, the Grando serves an equally practical role for creative professionals and engineers who need substantial local GPU compute. The 768GB of aggregate VRAM across eight RTX PRO 6000 cards unlocks capabilities that conventional workstation configurations cannot match.

FX artists and motion graphics professionals can render complex scenes with massive texture sets entirely in VRAM, eliminating the disk-swapping bottlenecks that plague productions using 8K footage or high-polygon environments. CAD engineers running computational fluid dynamics or structural simulations can tackle assemblies of unprecedented complexity without partitioning their models into multiple runs. Video editors working with multi-stream 8K RAW timelines, colorists applying ML-based noise reduction at full resolution, and 3D artists rendering path-traced finals locally rather than waiting for cloud farm availability all benefit from this density of GPU memory and compute.

The Grando does not require a full eight-GPU configuration. Comino offers the platform in four-GPU, six-GPU, and eight-GPU configurations, with all variants available for immediate shipment. Smaller studios, independent creators, and engineering teams can right-size their investment to current needs while retaining a clear upgrade path as workloads grow.

Platform Trade-offs: Density vs. Expandability

The Grando’s compact design delivers exceptional GPU density and thermal management within a standard 4U footprint, but that density involves architectural trade-offs worth understanding before deployment.

The chassis accommodates motherboards with EATX and EEB form factors, but not extended server boards found in traditional dual-socket platforms. This limits the total number of PCIe lanes available for peripherals beyond the GPU array. In our eight-GPU configuration, the AMD EPYC processor’s 128 PCIe Gen 5 lanes are almost entirely consumed by the GPUs, leaving little bandwidth for additional NVMe storage or high-speed networking beyond the onboard 10GbE ports.

This contrasts with the eight-GPU platforms we have reviewed from Dell, HPE, and Supermicro. Those systems use larger chassis, dual-socket configurations, and PCIe switch topologies to support significantly more peripheral connectivity. They typically accommodate four to eight additional NICs or DPUs alongside the full GPU complement, plus eight or more hot-swap NVMe bays, making them well-suited for distributed inference workloads that require high-bandwidth fabric interconnects.

However, that expanded capability comes at a substantial cost. Power draws exceed 8kW. Thermal loads require dedicated data center cooling infrastructure. Noise floors preclude deployment outside purpose-built machine rooms. And lead times frequently stretch six to eighteen months due to persistent supply constraints on enterprise GPU platforms.

The Grando occupies a different position. For organizations that prioritize rapid deployment, manageable operating environments, and inference or creative workloads over large-scale distributed training, the trade-offs are often favorable. Teams that need their hardware now, in an environment they can actually work with, may find the Grando’s approach to density more practical than waiting in a queue for a platform they cannot realistically deploy once it arrives.

Comino Grando Performance Testing Results

System Configuration

• Chassis: Comino Grando
• Motherboard: ASRock Rack GENOAD8X-2T/BCM
• CPU: AMD EPYC 9474F 48C
• Memory: 512GB DDR5
• GPU: 8 x NVIDIA RTX PRO 6000
• Storage: M.2 SSD

Claude Code Serving – MiniMax M2.5

Beyond traditional raw LLM inference benchmarks, we wanted to evaluate how well this hardware performs in an agentic coding workflow, specifically by serving multiple concurrent Claude Code sessions using a locally hosted model. This use case maps directly to development team productivity: how many engineers can simultaneously use an AI coding assistant served from a single node before the experience degrades?

To test this, we built a benchmark harness that generates a dataset of moderately difficult coding problems (such as implementing an LRU cache, building a CLI todo application, writing a markdown converter, and constructing a REST API) and runs each Claude Code session in a separate Docker container against the local vLLM server. A transparent proxy sits between the sessions and the inference endpoint, capturing per-request metrics for each Claude Code instance. The model used was MiniMax M2.5, served via vLLM on the system’s eight NVIDIA RTX PRO 6000 GPUs. While not the top-ranked coding model on public leaderboards, M2.5 is a capable model that many users, including our developer friends, run locally.

For a baseline reference point, we use Anthropic’s Claude Opus 4.6 average output throughput via OpenRouter.ai, one of the most popular routing services for production API access. That baseline comes in at approximately 37 tokens per second per API request.

We measured two key metrics: the average output tokens per second per Claude Code session (what each developer experiences) and the aggregate output tokens per second across all sessions (the total work the server produces).

Based on the results, a single concurrent Claude Code session delivers 67.3 tok/s per user and an aggregate output of 64.7 tok/s. At two sessions, per-instance throughput drops modestly to 57.4 tok/s, while aggregate output climbs to 95.1 tok/s as vLLM’s batching begins to amortize overhead. Four concurrent sessions maintain 49.2 tok/s per user, still a highly responsive experience for interactive coding workflows, while aggregate throughput reaches 177.2 tok/s. Eight sessions represent the sweet spot for aggregate output, peaking at 206.7 tok/s total, while per-instance throughput settles at 38.7 tok/s, a level that remains comfortable for real-time code generation and iteration.

At 16 concurrent sessions, the system exhibits the classic batching trade-off: per-instance throughput drops to 31.1 tok/s, and aggregate output falls to 105.8 tok/s. This suggests that, at this concurrency level, the 230B MiniMax M2.5 model is pushing the limits of what eight GPUs can sustain without introducing meaningful latency for each user. The aggregate dip from 8 to 16 sessions reflects the memory-bandwidth demands of a large MoE architecture under heavy simultaneous decode load, rather than a scheduling inefficiency.

For organizations evaluating self-hosted AI infrastructure for developer tooling, the Grando makes a strong case. Running a frontier-class 230B model, it can comfortably serve up to eight simultaneous Claude Code sessions at throughput levels that feel genuinely interactive, with per-user speeds exceeding 38 tok/s at peak aggregate output. Teams of four to eight engineers can operate at near-optimal throughput without perceptible degradation in responsiveness.

The liquid-cooled architecture also makes this level of compute practical in environments where traditional GPU servers cannot operate. The system runs quietly enough to sit in a startup office, a small machine room, or a dedicated corner of an open workspace. Air-cooled systems with similar GPU density typically reach 90 dB or higher, which is loud enough to require dedicated data center space or, at a minimum, a closed server closet with serious acoustic treatment. The Grando can coexist with the team that uses it. Combined with full data locality, no per-token API costs, and complete control over model selection, it offers a self-hosted path that scales with a growing development team without requiring datacenter infrastructure or lockstep cost increases.

vLLM Online Serving – LLM Inference Performance

vLLM is one of the most popular high-throughput inference and serving engines for LLMs. The vLLM online serving benchmark evaluates the real-world serving performance of this inference engine under concurrent requests. It simulates production workloads by sending requests to a running vLLM server, with configurable parameters such as request rate, input and output lengths, and the number of concurrent clients. The benchmark measures key metrics, including throughput (tokens per second), time to first token, and time per output token (TPOT), helping users understand how vLLM performs under different load conditions.

We tested inference performance across a comprehensive suite of models spanning various architectures, parameter scales, and quantization strategies to evaluate throughput under different concurrency profiles.

Summary Of Results

Model	Precision	Equal (256/256)	Prefill-Heavy (8k/1k)	Decode-Heavy (1k/8k)
Comino Grando w/ 8× RTX PRO 6000 Blackwell — vLLM Inference Results (tok/s, peak at BS=256)
GPT-OSS 20B	ep_dp1	17,280	32,061	11,187
GPT-OSS 120B	ep_dp1	11,726	21,636	7,570
Llama 3.1 8B Instruct	FP8	12,109	20,137	7,353
Llama 3.1 8B Instruct	FP4	11,954	20,206	7,239
Llama 3.1 8B Instruct	BF16	11,752	17,346	6,155
Qwen3 Coder 30B A3B	FP8	10,985	16,659	4,907
Qwen3 Coder 30B A3B	BF16	10,588	16,680	4,829
Mistral Small 3.1 24B	BF16	8,925	11,846	4,975
MiniMax M2.5 (230B)	ep_dp1	5,753	7,357*	2,555
All values in tok/s, peak throughput at BS=256. *MiniMax M2.5 prefill-heavy peaked at BS=128 (7,357 tok/s); BS=256 was 7,141 tok/s.

GPT-OSS 120B and 20B

The GPT-OSS model family was tested in both 120B and 20B configurations on the Comino Grando.

GPT-OSS 120B

Under equal workload (256/256), the 120B model delivers 268.85 tok/s at BS=1, reaches 6,666.23 tok/s at BS=64, and peaks at 11,726.04 tok/s at BS=256. Prefill-heavy (8k/1k) starts at 1,375.69 tok/s, climbs to 16,374.19 tok/s at BS=64 and 17,944.55 tok/s at BS=128, and peaks at 21,636.41 tok/s at BS=256. Decode-heavy (1k/8k) grows from 196.28 tok/s at BS=1 to 7,569.97 tok/s at BS=256, with latency well-controlled at lower concurrency levels.

GPT-OSS 20B

The 20B model delivers 334.80 tok/s at BS=1 under equal workload, reaches 10,303.56 tok/s at BS=64, and peaks at 17,280.12 tok/s at BS=256. Prefill-heavy starts at 2,007.90 tok/s, climbs to 24,990.46 tok/s at BS=64 and 26,866.25 tok/s at BS=128, peaking at 32,060.72 tok/s at BS=256, the highest absolute prefill throughput recorded across both model sizes. Decode-heavy grows from 286.08 tok/s at BS=1 to 11,187.36 tok/s at BS=256, delivering roughly 1.5× the decode throughput of the 120B at peak concurrency while maintaining tighter latency throughout.

Qwen3 Coder 30B A3B Instruct and FP8 Instruct

The Qwen3-Coder-30B-A3B-Instruct model was tested with both BF16 and FP8 precision.

Qwen3-Coder-30B-A3B-Instruct (BF16)

Under an equal workload (256/256), the BF16 model delivers 1,902.32 tok/s at BS=8, reaches 6,683.58 tok/s at BS=64, and peaks at 10,587.56 tok/s at BS=256. Prefill-heavy (8k/1k) starts at 1,256.03 tok/s at BS=1, climbs to 14,400.57 tok/s at BS=64 and 15,308.35 tok/s at BS=128, and peaks at 16,679.52 tok/s at BS=256. Decode-heavy (1k/8k) grows from 169.19 tok/s at BS=1 to 4,828.82 tok/s at BS=256, with latency well-controlled at lower concurrency levels.

Qwen3-Coder-30B-A3B-Instruct (FP8)

The FP8 model delivers throughput comparable to BF16 across most scenarios, with equal workload reaching 6,478.54 tok/s at BS=64 and peaking at 10,984.61 tok/s at BS=256, a slight improvement over BF16 at peak concurrency. Prefill-heavy starts at 987.48 tok/s at BS=1, climbs to 14,036.46 tok/s at BS=64 and 15,156.69 tok/s at BS=128, and peaks at 16,658.98 tok/s at BS=256. Decode-heavy grows from 130.70 tok/s at BS=1 to 4,906.51 tok/s at BS=256, marginally outpacing BF16 at peak concurrency while the two configurations remain closely matched throughout the rest of the concurrency range.

Mistral Small 3.1 24B Instruct 2503

Under an equal workload (256/256), the model delivers 1,598.79 tok/s at BS=8, reaches 4,713.84 tok/s at BS=64, and scales strongly to 8,925.12 tok/s at BS=256. Prefill-heavy (8k/1k) starts at 897.84 tok/s at BS=1, climbs to 9,632.58 tok/s at BS=64 and 11,488.13 tok/s at BS=128, peaking at 11,846.15 tok/s at BS=256. Decode-heavy (1k/8k) grows from 124.98 tok/s at BS=1 to 2,653.82 tok/s at BS=64, then accelerates noticeably at higher concurrency levels, reaching 4,262.53 tok/s at BS=128 and peaking at 4,975.06 tok/s at BS=256, reflecting the model’s ability to sustain strong decode throughput as concurrency scales.

Llama 3.1 8B Instruct

The Llama-3.1-8B-Instruct model was tested across three precision configurations on the Comino, providing a clear view of how quantization affects throughput for this model size.

Llama 3.1 8B Instruct BF16

Under an equal workload (256/256), the BF16 model delivers 2,776.42 tok/s at BS=8, reaches 7,369.01 tok/s at BS=64, and peaks at 11,751.56 tok/s at BS=256. Prefill-heavy (8k/1k) starts at 1,645.29 tok/s at BS=1, climbs to 14,990.47 tok/s at BS=64 and 17,140.71 tok/s at BS=128, and peaks at 17,345.80 tok/s at BS=256. Decode-heavy (1k/8k) grows from 234.78 tok/s at BS=1 to 6,154.73 tok/s at BS=256.

Llama 3.1 8B Instruct FP8

FP8 quantization delivers a meaningful uplift across all scenarios. The equal workload reaches 7,530.39 tok/s at BS=64 and peaks at 12,108.98 tok/s at BS=256. Prefill-heavy climbs to 16,546.53 tok/s at BS=64 and 19,306.49 tok/s at BS=128, peaking at 20,137.35 tok/s at BS=256, roughly a 16% gain over BF16 at peak concurrency. Decode-heavy peaks at 7,353.40 tok/s at BS=256, approximately 19% ahead of BF16.

Llama 3.1 8B Instruct FP4

FP4 delivers throughput that is closely competitive with FP8 at higher concurrency levels, though it falls slightly behind at lower batch sizes. The equal workload peaks at 11,954.40 tok/s at BS=256, and prefill-heavy reaches its highest point at 20,205.57 tok/s at BS=256, narrowly edging out FP8 at peak concurrency. Decode-heavy peaks at 7,239.29 tok/s at BS=256, remaining within a few percent of FP8 throughout, making FP4 a compelling option when memory efficiency is a priority without a meaningful sacrifice in throughput.

MiniMax M2.5

The MiniMax-M2.5 230B, tested on the Comino Grando, was the largest and most demanding model we used.

Under an equal workload (256/256), the model starts at 16.35 tok/s at BS=1, reaches 2,751.25 tok/s at BS=64, and scales strongly at higher concurrency, peaking at 5,753.24 tok/s at BS=256. Prefill-heavy (8k/1k) starts at 606.97 tok/s at BS=1, climbs steadily to 5,351.02 tok/s at BS=32 and 6,557.92 tok/s at BS=64, reaching its peak at 7,357.26 tok/s at BS=128 before slightly tapering to 7,140.74 tok/s at BS=256, suggesting the model approaches saturation in prefill throughput beyond BS=128. Decode-heavy (1k/8k) grows consistently from 82.21 tok/s at BS=1 to 1,485.28 tok/s at BS=64, peaking at 2,554.87 tok/s at BS=256, reflecting the expected memory bandwidth demands of a 230B MoE architecture under sustained decode workloads.

Conclusion

The Comino Grando is best understood as a system purpose-built to unlock the full potential of eight NVIDIA RTX PRO 6000 GPUs. Every major design decision, from the inverted motherboard layout to the cooling loop and integrated monitoring stack, is intended to ensure those GPUs can operate continuously at full 600W TDP without thermal or power constraints.

What makes the Grando compelling is not any single feature in isolation but the way the entire system coheres. The liquid cooling is not a bolt-on addition; it is the architecture. The power delivery is redundant, hot-swappable, and scaled to the 4,800W load of eight 600W cards with headroom to spare. The monitoring system goes beyond reporting temperatures; it autonomously protects the hardware when something goes wrong. Nothing here feels like an afterthought.

The performance numbers reinforce that cohesion. Across a diverse suite of models, from Llama 3.1 8B to the 230B MiniMax M2.5, the Grando delivered throughput figures that hold up well for a self-hosted platform. Claude Code concurrency testing put a finer point on the practical value: eight engineers can run simultaneous agentic coding sessions against a locally hosted 230B model at interactive speeds, with per-user throughput exceeding 38 tok/s at peak aggregate output. Teams of four to eight can operate at near-optimal throughput without perceptible degradation.

The value of this configuration extends beyond AI inference. With 96GB of VRAM per GPU and dense multi-GPU scaling, the platform is equally well suited for high-end creative and engineering workloads, including VFX rendering, large-scale simulation, and complex CAD pipelines. The system scales down to four-GPU and two-GPU configurations, making this level of performance accessible to smaller studios and teams that still require workstation-class density.

Where the Grando differs most from the enterprise eight-GPU platforms we have reviewed is in deployment practicality. Those systems offer more PCIe lane headroom, more NIC slots, and deeper storage connectivity, but they also require dedicated data center infrastructure, draw well over 8kW, and have lead times that can stretch beyond a year. The Grando trades some of that peripheral expandability for a system that runs quietly enough to share a room with its users, dissipates less heat into the surrounding environment, and ships now. For organizations that prioritize rapid deployment and manageable operating environments over maximum fabric connectivity, the trade-off is favorable.

Product Page – Comino Grando
Comino Configurator – Page

The post Comino Grando RTX PRO 6000 Review: 768GB of VRAM in a Liquid-Cooled 4U Chassis appeared first on StorageReview.com.

At the AI in Finance Summit, Broadcom introduced VMware Tanzu Platform agent foundations, positioning it as a secure-by-default runtime for building and operating autonomous AI applications on VMware Cloud Foundation (VCF). The release extends Tanzu’s established code-to-production model to AI agents, targeting enterprise teams seeking to move from isolated AI experiments to governed, production-scale deployments.

Moving AI Agents into Enterprise Operations

As AI agents assume execution and decision-making roles, operational requirements shift toward governance, security, and integration with enterprise systems. Many organizations still run AI workloads in isolated environments that lack access to core data and standardized controls.

Tanzu Platform agent foundations address this gap by providing a pre-engineered platform-as-a-service layer for agent workloads, built directly on VCF. This enables platform engineering teams to manage AI services alongside traditional applications with familiar tooling and processes, without requiring deep specialization in AI infrastructure.

Deny-by-Default Agent Runtime

The agentic runtime introduces a set of controls to constrain agent behavior and reduce operational risk.

The software supply chain is managed using trusted Buildpacks rather than user-defined Dockerfiles. Containers are automatically built, patched, and verified, reducing exposure to embedded vulnerabilities or malicious components.

Secrets management is enforced at the structural level, preventing agents from accessing credentials outside their scope. This isolation is reinforced by VMware vDefend, which extends protections across infrastructure services and external SaaS integrations, limiting lateral movement.

Networking uses a zero-trust model. Agents operate within predefined resource and connectivity boundaries and have no default access to internal systems or models. Access is granted explicitly via secure service bindings, ensuring agents interact only with authorized data sources and services.

Developer Onboarding and Integrated Data Services

The platform includes pre-built agent templates to accelerate onboarding. Developers can provision agents with governed access to models, Model Context Protocol servers, and curated marketplace services defined by IT.

Data services are integrated into the platform, including Tanzu for Postgres with pgvector, as well as caching, streaming, and data flow services. Support for Spring AI memory services enables stateful agent behavior that aligns with enterprise application patterns.

Operational Scaling on VMware Cloud Foundation

Tanzu Platform agent foundations integrate with VCF infrastructure APIs to abstract away resource provisioning and lifecycle management. This ensures that agent workloads and their dependencies receive the required compute, storage, and networking resources without direct interaction with the infrastructure.

Elastic scaling allows environments to scale up or down based on workload demand, supporting both short-lived and persistent agents while optimizing cost and utilization.

High availability is achieved through multiple layers of redundancy and automated remediation. The platform continuously monitors and self-heals the underlying infrastructure to maintain service continuity for mission-critical autonomous applications.

An integrated AI gateway provides centralized control of model and tool access. It manages availability, usage policies, cost controls, and safety filtering for both public and private models on VCF.

According to Purnima Padmanabhan, General Manager of the Tanzu Division at Broadcom, the rapid agentic application development is driving collaboration with customers to accelerate innovation. She highlights that Tanzu Platform agent foundations enable the rapid deployment of agentic ideas on modern private clouds, specifically using VMware Cloud Foundation 9.

With agent foundations, Broadcom is aligning Tanzu Platform with emerging enterprise AI requirements, with a focus on governance, security, and operational consistency. The approach builds on existing VMware infrastructure investments and introduces a standardized runtime for agent-based applications, making AI deployment more predictable and manageable at scale.

The post Broadcom Extends VMware Tanzu Platform with Agent Foundations for Enterprise AI appeared first on StorageReview.com.

Wasabi Technologies has reached a definitive agreement to acquire the Lyve Cloud business from Seagate Technology. As part of the transaction, Seagate will receive an equity stake in Wasabi, officially becoming a shareholder. While specific financial details remain undisclosed, the move marks a significant consolidation in the pure-play cloud storage market.

David Friend, co-founder and CEO of Wasabi, noted that the acquisition bolsters the company’s position as a leader in independent cloud storage. The integration of Lyve Cloud brings a dedicated enterprise customer base into Wasabi’s ecosystem. These customers will transition to Wasabi’s global data center infrastructure, which features specialized security tools such as Covert Copy and integrated AI capabilities. The provider intends to maintain high levels of technical support and partner integration for the incoming Lyve Cloud users.

For Seagate, the divestiture serves a specific strategic purpose. Gianluca Romano, Seagate’s CFO, indicated that the sale allows the company to refocus resources on its core mass-capacity storage hardware business. As the demand for high-capacity drives continues to climb, Seagate aims to prioritize manufacturing and innovation in hard drive technology. By transitioning the cloud service to Wasabi, Seagate ensures that a specialized provider services its existing cloud customers while the manufacturer maintains an indirect interest through its new equity position.

Engineering for Enterprise Scale

The proliferation of AI initiatives, large-scale analytics, and extensive video workloads is currently driving demand for enterprise-grade storage. As organizations manage data volumes reaching the petabyte scale, the total cost of ownership and vendor complexity become critical factors in infrastructure design. Many firms are moving away from traditional hyperscalers in favor of providers that offer predictable pricing models and robust security without the egress fees often associated with legacy cloud platforms.

Lyve Cloud established itself as a viable enterprise platform by prioritizing compliance and security features. By merging these assets with Wasabi’s established channel reach and execution strategy, the combined entity provides a streamlined alternative for professional IT environments. The acquisition aims to deliver consistent performance at scale while addressing the economic challenges of long-term data retention.

Ecosystem Integration and Data Protection

The consolidation of these two platforms simplifies the data protection and backup landscape for administrators. Both Wasabi and Lyve Cloud maintain deep integrations and certifications with leading backup software providers, including Veeam, Rubrik, and Commvault. This overlap ensures that existing automated workflows and S3-compatible API calls remain functional during and after the transition.

For channel partners and system integrators, the acquisition reduces the overhead of managing multiple independent S3-compatible storage vendors. By unifying the service under a single banner, Wasabi enhances its ability to support mission-critical backup and recovery workloads. This move strengthens the broad ecosystem of independent storage solutions, providing technical teams with a reliable, cost-effective target for enterprise data offloading.

The post Wasabi Technologies to Acquire Seagate Lyve Cloud Business appeared first on StorageReview.com.