Arm Targets Server AI Bottlenecks with Custom Architecture Benchmarks
Arm claims its upcoming AGI CPU design nearly doubles rack-level throughput against AMD's flagship Turin chip in complex agentic workloads.
Silicon, foundries, GPUs — the substrate of everything.
156 stories · page 2 of 7
Arm claims its upcoming AGI CPU design nearly doubles rack-level throughput against AMD's flagship Turin chip in complex agentic workloads.
Intel Foundry Services is reportedly securing new customers, raising questions about whether previous cost-cutting initiatives might hinder its long-term capacity and technological readiness for broader industry demands.
A $4 billion industrial park in Texas aims to transplant Taiwan's specialized chemicals, precision tooling, and packaging suppliers directly alongside domestic fab expansions.
Applied Materials' new multi-billion dollar EPIC Center represents a necessary shift in semiconductor manufacturing, moving from siloed development to integrated co-optimization of materials, equipment, and process architecture.
SK hynix is moving away from generic Arm architectures for I/O management, introducing bespoke H-TPU cores to handle the massive throughput of next-generation enterprise storage.
A new distributed computing framework from Apple aims to solve the AI compute shortage by networking idle Mac hardware into a virtualized, high-performance GPU cluster.
Intel's new GPU D-Series family, powered by the Crescent Island AI inference accelerator, promises substantial FP8 compute and high memory bandwidth, targeting AI workloads with specific performance metrics.
Seven years after abandoning the FinFET race, GlobalFoundries pivots back to advanced scaling with a 7nm-class FD-SOI node targeting high-efficiency AI and RF applications.
Researchers have demonstrated a method using controlled oxygen introduction to regulate transition metal dichalcogenide growth, addressing critical nucleation flaws for future sub-nanometer logic nodes.
A leaked factory notice cited by Tom's Hardware says Nvidia will stop producing the GeForce RTX 5090 and shift its GB202 silicon to data‑center and professional GPUs, making the RTX 5080 the new top‑end consumer card.
A new five-year, $2 billion agreement brings GlobalFoundries into TSMC's advanced packaging ecosystem, utilizing mature US fabs to relieve AI hardware bottlenecks.
AMD's upcoming EPYC Verano AI host CPUs are slated to adopt an all-new SB1 socket, pairing 72 Zen 6 cores with a 24-channel LPDDR5X memory subsystem.
The CHIPS Act pivots toward long-term R&D with a massive investment in IBM-backed Anderon, aiming to bridge the gap between lab-scale qubits and industrial manufacturing.
EV charging firm Xeal is set to deploy 100,000 Nvidia GPUs across its U.S. network, transforming idle charging infrastructure into a distributed edge inference compute platform.
Qualcomm has stopped actively blocking Linux support on its Snapdragon X2 Elite silicon, releasing documentation and enabling driver development for third-party operating systems.
Elon Musk has confirmed discussions with TSMC regarding a potential Terafab chipmaking collaboration, while Intel remains another named partner tied to the project.
A Californian CEO faces federal charges for allegedly orchestrating a $300 million scheme to illegally export advanced Nvidia AI servers to China, circumventing critical US export controls designed to restrict access to high-performance computing hardware.
Huawei and DeepSeek have released a comprehensive suite of open-source programming tools for the Ascend 950 AI chips, aiming to simplify development and foster an alternative ecosystem to Nvidia's CUDA.
Gears of War: E-Day's pioneering integration of Unreal Engine 5's MegaLights technology showcases unprecedented demands on GPU hardware, particularly for ray-traced lighting, signaling a significant shift in graphics rendering paradigms and future GPU architectural priorities.
According to The Register, US datacenter construction is accelerating, but advanced packaging limits could cap 2027 deployments in the low 20s of gigawatts.
A newly published patent reveals Intel's method for embedding MicroLEDs into glass substrates using through-glass vias, a technique that could support both aesthetic functions and future optical interconnects.
Qualcomm unveiled two new flagship smartphone SoCs that can run a 30‑billion‑parameter mixture‑of‑experts model locally, marking a leap in on‑device AI performance and reshaping the mobile compute landscape.
Nvidia's RTX Mega Geometry 2.0 SDK introduces dynamic VRAM streaming for ray-tracing assets, solving memory bottlenecks in complex rendering pipelines.
The massive expansion of the Colossus supercomputer cluster highlights the shift in AI scaling bottlenecks from silicon availability to utility-scale power generation and liquid-cooling infrastructure.