Huawei has released its next-gen Ascend chip and SuperPod roadmap, which comprises Ascend 960DT/PR, Ascend 970, and Ascend 980 accelerators. Huawei Pushes Its AI Accelerator Roadmap Ahead With Next-Gen Ascend Chips Arriving Earlier Than Anticipated At Huawei Connect 2026, the company disclosed its latest Ascend and Atlas roadmaps, which feature various next-generation chips and technologies. The main point is that Huawei is going to push its AI infrastructure roadmap ahead to address the growing industry demands, and there are four upcoming chips to look at. The Ascend Chip Portfolio Starting with the chips, Ascend will first introduce its next-gen Ascend […]

Read full article at https://wccftech.com/huawei-ascend-960dt-960pr-2027-970-2028980-2028-superpods-ai/

NVIDIA showcased its energy efficiency optimizations that have been featured on Vera Rubin platforms for increased token throughput. NVIDIA Discloses Some Hefty Power Efficiency Gains For Its Vera Rubin Platforms at AI Infra Summit AI demand continues to pose a serious burden on energy infrastructure, and to address this challenge, NVIDIA is working on a broad range of solutions as a part of its full-stack AI factory platform, which spans Vera Rubin systems, Dynamo inference software, NeMo libraries, and a broad range of networking solutions including NVLink for scale-up computing, Spectrum-X for Ethernet and ConnectX SuperNICs for connecting 1000s of […]

Read full article at https://wccftech.com/nvidia-dsx-maxlps-drives-more-token-throughput-per-megawatt/

MediaTek just wrapped up the Dimensity 9600 Pro announcement, its fastest chipset to date, and also the first 2nm silicon that is expected to power a wave of Android flagship smartphones. Naturally, a chipset unveiling isn’t complete without some performance slides, and what better way to show the Dimensity 9600 Pro’s capabilities than by showcasing some single-core and multi-core numbers. However, while these figures show that the SoC is faster than Apple’s A20 Pro in one area, MediaTek has resorted to deceptive tactics by reducing the Dimensity 9500 scores just to uplift the Dimensity 9600 Pro. Single-core and multi-core scores of Dimensity […]

Read full article at https://wccftech.com/dimensity-9600-pro-beats-a20-pro-multicore-dimensity-9500-scores-lowered/

DIGITIMES observes that major memory suppliers posted record-high revenue in the second quarter of 2026, with yearover-year growth remaining in the triple digits. Driven by AI demand, SSDs and LPDDR are emerging as key competitive
battlegrounds.

The SiP (System in Package) belonging to the Apple Watch went through a series of mediocre and iterative changes, meaning that these smartwatches had to offer different major selling points to warrant an upgrade. This year, however, the S11 found in the Apple Watch Series 12 and Apple Watch Ultra 4 has finally broken that cycle, and it’s all thanks to the A20 Pro found in the latest iPhone 18 Pro and iPhone Duo. The S11 is derived from the A20 Pro, with a series of specification upgrades such as memory improvements and more that help to improve the on-device AI […]

Read full article at https://wccftech.com/apple-watch-s11-chipset-upgrade-a20-pro-architecture/

The A20 Pro wasn’t just designed around a packaging change to enable Apple’s newest 2nm SoC to deliver raw performance gains, but it was also to improve the on-device AI experience, which meant the company’s engineers had to work some magic on increasing memory bandwidth without unnecessarily raising costs by switching to the LPDDR6 standard. The result is a chipset that edges past Xiaomi’s XRING 03, which, by the way, offers LPDDR6 RAM support but loses to the A20 Pro in memory bandwidth. At 115.2GB/s, the A20 Pro manages to beat the XRING 03, which tops out at 113.8GB/s; here’s how […]

Read full article at https://wccftech.com/apple-a20-pro-memory-bandwidth-beats-xiaomi-xring-03-lpddr5x-ram/

Qualcomm was already facing a torrent of struggles with its declining chipset business due to the DRAM crisis, and to make matters worse, it’s going to pay TSMC a hefty sum for utilizing its 2nm lithography for the upcoming Snapdragon 8 Elite Gen 6 Pro and Snapdragon 8 Elite Gen 6. The only solace for the company to reduce its manufacturing costs is teaming up with Samsung, but the latest report states that both entities have failed to enter into an agreement due to pricing differences. Exynos 2700 is proof that Samsung has stabilized its advanced 2nm yields, but cannot […]

Read full article at https://wccftech.com/samsung-passes-technical-hurdles-manufacturing-qualcomm-2nm-chips-hits-price-agreement-snag/

It goes without saying that the A20 Pro is truly a wonder when it comes to smartphone chipsets, with the latest look at its single-core and multi-core scores cementing its reputation in the industry. However, we’re still learning about the SoC’s impressive feats, with the latest one being that its dual 16-core Neural Engine is said to have a higher peak compute than its 7-core GPU, assuming those workloads are specifically tailor-made for the NPU. Running smaller 3B on-device AI models and prompt processing would be specific tasks for the A20 Pro’s dual 16-core Neural Engine, offering immense peak performance gains On […]

Read full article at https://wccftech.com/a20-pro-dual-16-core-neural-engine-peak-compute-throughput-higher-than-7-core-gpu-specialized-workloads/

A new manufacturing process and a package redesign gave Apple the proper direction to develop the A20 Pro, and with these changes come massive improvements showcased in Geekbench 6. Not only has the chipset achieved a new record in the single-core result, but its multi-core scores rival those of Apple’s M-series SoCs like the M3. The latest Geekbench 6 scores show that the A20 Pro is approaching 5,000 points in single-core workloads What’s interesting about the latest leak is that the A20 Pro wasn’t tested in Geekbench 7, but the older Geekbench 6 suite, yet the results are still impressive, […]

Read full article at https://wccftech.com/apples-a20-pro-shatters-geekbench-6-records-beats-m5-max/

DIGITIMES identifies three main directions for memory development in cloud AI accelerators: processing near memory, 3D Stacked SRAM, and high-bandwidth flash.

The iPhone 18 Pro, iPhone 18 Pro Max, and iPhone Duo will be rocking Apple’s new 2nm chipset, the A20 Pro, making it the first time in the technology giant’s history that it has introduced a silicon fabricated on a next-generation lithography. There are a ton of improvements incorporated into the SoC, including a packaging redesign, so let us get into the details. Apple claims the A20 Pro features a desktop-class CPU, retaining a 6-core configuration while delivering decent performance gains The new A20 Pro features two new super cores that are 20 percent faster than the A19 Pro’s performance […]

Read full article at https://wccftech.com/apple-a20-pro-iphone-first-2nm-soc-gpu-cpu-neural-engine/

The entire iPhone lineup, which is expected to be unveiled in less than two hours, is expected to be equipped with Apple’s second 2nm SoC after the M6, the A20 Pro. Just before the official announcement, a multitude of rumored specifications have come through. In short, the new silicon will likely be a beast in nearly all departments, so let us check out all the details. New A20 Pro to also feature 36MB of SLC cache, bringing a massive performance boost A previous front die shot leak revealed that the A20 Pro would retain a 6-core CPU configuration like its immediate predecessor, the […]

Read full article at https://wccftech.com/apple-a20-pro-leaked-specs-4-9ghz-cpu-7-core-gpu-m3-bandwidth/

NVIDIA’s newest DLSS 5 has been going through the testing blender to see the level of Neural Rendering changes the company’s latest upscaling iteration brings to the table. Naturally, the unofficial version isn’t exactly kind to modern-day GPUs, as various modded games deliver crippled performance. However, DLSS 5 is exceptionally unforgiving of Apple Silicon like the M5 Pro, delivering insane image quality improvements but also major trade-offs like single-digit framerates and 10 times the input latency compared to a desktop RTX 5050. M5 Pro’s lackluster showing with DLSS 5 enabled is due to a lack of dedicated FP8 hardware units, bringing […]

Read full article at https://wccftech.com/dlss-5-tested-apple-m5-pro-experimental-layer-boosts-image-quality-latency-penalty/

The first design glimpse of Qualcomm’s upcoming flagship SoC, the Snapdragon 8 Elite Gen 6 Pro, revealed that the chipset has a Heat Pass Block similar to Samsung’s Exynos 2600. Now, we get to see this solution live because a real-life chipset die has been pictured, showing a significantly bigger area than the Snapdragon 8 Elite Gen 5, indicating that we could bear witness to exceptional thermals in the future. Snapdragon is finally prioritizing thermals with the Snapdragon 8 Elite Gen 6 Pro, with a new cooling solution The traditional PoP (Package-on-Package) implementation is expected to be retired with the arrival […]

Read full article at https://wccftech.com/snapdragon-8-elite-gen-6-pro-die-leak-heatsink-dram-cooling/

The upcoming “Surprise and Shine” event scheduled for September 9 will showcase Apple’s first-ever 2nm chipset for the iPhone, the A20 Pro. We’ve already had our initial glimpse at the SoC’s packaging, showing that Apple will move to WMCM (Wafer-Level Multi-Chip Module) technology, ditching the older InFO_PoP solution. Now, we’ve stumbled upon our first front-side die shot of the silicon, showing a higher GPU core count and other improvements, so let us begin. A20 Pro could be Apple’s first iPhone SoC with a 7-core GPU, thanks to the new packaging change, bringing exceptional graphics performance to the table A previous look […]

Read full article at https://wccftech.com/apple-a20-pro-die-shot-record-gpu-cores-wider-memory-bus/

Advanced lithography and next-generation manufacturing processes meant that a smartphone’s chipset would be the costliest component running inside these mobile devices, that’s until the AI boom arrived. According to a tipster’s latest estimations, a DRAM and NAND flash pairing has become 340 percent more expensive and costs more than a top-end SoC. With a 12GB RAM + 256GB NAND flash combo costing approximately $310, smartphone manufacturers will struggle immensely to generate healthier margins Despite smartphone SoCs moving to TSMC’s 2nm “N2” and “N2P” process this year, starting with Apple’s A20 Pro, it still won’t be as expensive as DRAM and […]

Read full article at https://wccftech.com/dram-nand-flash-costs-exceed-soc-bom/

The M6 might be Apple’s first 2nm chipset thanks to its long-standing partnership with TSMC, but there’s a massive difference between having bragging rights over adopting a newer lithography and how well it’s going to be adopted. For the new Apple Silicon, a new report states that its importance is limited to proving that TSMC’s 2nm node is ready for the mass market. As for increasing the share of this manufacturing process, that’s a job for the A20 Pro. An ongoing DRAM shortage risks the M6 Mac mini’s adoption, thanks to its 50 percent price increase compared to the M4 Mac […]

Read full article at https://wccftech.com/apple-m6-validates-tsmc-2nm-process-a20-pro-launch-holds-more-weight/

The M6 Pro and M6 Max were getting skipped in favor of the M7 Pro and M7 Max, but shortly after the M6 and Mac mini refresh announcement, an update claims that Apple is still developing both higher-end SoCs, and they are said to arrive with advanced packaging technologies like SoIC-MH and WMCM (Wafer-Level Multi-Chip Module), bringing a host of benefits to the table. The iPhone 18 packaging is coming to the M6 Pro and M6 Max, assuming the report checks out An article published by Commercial Times states that the base M6 retains the use of Apple’s SoIC-MH technology, […]

Read full article at https://wccftech.com/apple-m6-pro-m6-max-soic-mh-wmcm-packaging-report/

NVIDIA has just announced its custom HBM solution called NVHBM, which will be used in future GPUs, offering higher bandwidth & lower power. NVHBM Is A Custom HBM Solution Designed By NVIDIA & Amazon's Annapurna Labs To Speed Up DRAM Capabilities For Future Agentic & Physical AI Workloads The NVIDIA NVHBM DRAM solution is an extension of NVLink Fusion, offering higher memory performance and efficiency to XPUs at a time when memory bandwidth is becoming critical as model sizes increase tremendously into the trillion-parameter era. NVIDIA says that its NVHBM solution offers better bandwidth and efficiency than standard HBM solutions […]

Read full article at https://wccftech.com/nvidia-develops-custom-nvhbm-memory-for-ai-more-bandwidth-lower-power-than-hbm4e/