Topic
Explore English articles about AI Semiconductor.
Micron says HBM3E uses roughly three times more wafer capacity than DDR5 at the same process node, and that this input ratio will rise further in future generations. Here's how expanding bandwidth demands are straining DRAM supply and thermal design.
NVIDIA has announced NVHBM, which moves HBM control circuitry to the base die. The company claims gains in bandwidth, power, and compute area, but absolute specifications, mass-production timing, and adopting products remain undisclosed.
Fujitsu's 144-core MONAKA 3D-stacks N2P compute cores with N5 SRAM and I/O. Ahead of 2027 shipment, estimated performance and stacking still need real-world verification.
TrendForce forecasts that DRAM and NAND will account for 68% of major cloud providers' capital expenditure in 2027. We examine the assumptions behind this estimate, the role of long-term contracts, and the impact of reduced memory configurations.
AMD's Technical Information Portal product categories and a Linux HSMP patch shed light on Olympic Ridge and Medusa1/2 configurations across AM5 and FP10. Here's what's confirmed and what remains unannounced.
Intel's Crescent Island pairs 32 Xe cores with up to 480GB of LPDDR5X in a 350W air-cooled PCIe card aimed at long-context inference. Memory bandwidth and real-world performance remain undisclosed.
Windows 11のプレビュー版内部から、GPUやAI向けに統合メモリを手動配分する隠し機能が見つかった。RTX Sparkなど大容量統合メモリPCの登場を前に、OS側のメモリ管理はどう変わるのかを読み解く。
Microsoft Paint records a server-issued GUID in both the AI-generated image and its C2PA data, even for local generation. Photos was found to use a similar code path, but no personal identification or account matching has been confirmed.
Arm's newly disclosed AGI CPU splits more than 100 billion transistors across two compute chiplets, fitting 136 cores into a 300W envelope. Here's a look at the design goals and what stands between $2 billion in demand and mass production.
NVIDIA's Groq 3 LPX, now in mass production, splits AI inference work between GPUs and LPUs to reach 3,431 output tokens per second. Here's what that number means, where the benchmark falls short, and what's behind the $17 billion deal.
