On a single 8-GPU AMD Instinct MI355X node, AMD served Kimi Linear 48B-A3B with context lengths from 1024 tokens to 64Mi (67,108,864), using vLLM with FP8 KV cache and tensor parallelism 8, recording …
AMD donated technology to more than 800 universities, research institutions and nonprofit organizations from 2020 through 2025, benefiting an estimated 104.8 million people, and pledged $150 million i…
D-Matrix presented its Raptor 3D-DRAM accelerator for generative inference at Hot Chips 2026, claiming it addresses the bandwidth and capacity limits of SRAM and HBM by stacking compute directly on DR…
Skytech Gaming's Azure 3 PC, featuring an AMD Ryzen 7 9850X3D and Radeon RX 9070 XT with 16GB VRAM, is available on Amazon for $400 off, offering a high-performance 1440p or 1600p gaming machine. The …
Anthropic released a technical postmortem on April 23 addressing user reports that Claude seemed less capable, attributing the perceived decline to three separate causes: a reduced reasoning budget, a…
AMD announced on May 21, 2026, that it will invest over $10 billion (NT$315.58 billion) in Taiwan's semiconductor ecosystem, focusing on advanced chip packaging technology developed with TSMC. The mul…
Compute grew 106x since NVIDIA's P100 in 2016, while memory bandwidth grew only 10.9x, a 10x decline in bytes-per-FLOP that has left AI accelerators bandwidth-starved, according to a chip inventory of…
Analog Devices reported record fiscal third-quarter revenue of $4.02 billion, up 40% year over year, driven by data center optical and power business growth of over 100%, which now accounts for about …
At the 2018 International Symposium on Computer Architecture, John Hennessy and David Patterson delivered their Turing Lecture 'A New Golden Age for Computer Architecture,' noting that single-threaded…
A user reports running large language models on a 128GB Bosgame M5 Strix Halo mini PC purchased used for 1800€ on eBay, achieving 30 tokens per second decode with Qwen 3.8 27B and 50 tokens per second…
A developer created a fork of llama.cpp optimized for AMD GFX906 GPUs (Mi50, Mi60, Radeon VII, GCN HIP), and another developer noted they had previously finetuned llama.cpp for one card, mentioning th…
Samsung Electronics reported its HBM4 memory production yield reached 80% in early August, up from below 60% in February, four months ahead of schedule, according to TrendForce and Korean media. The y…
VLLM's disaggregated serving using AMD's MORI-IO connector achieved 2.5x higher goodput on the same 8-GPU node compared to standard collocated serving, according to an April 2026 blog post. The benchm…
Anthropic has hired Amir Salek, the engineer who founded and led Google's custom silicon division from 2013 to 2022, including its Tensor Processing Unit line, as part of its move to design its own AI…
Anthropic announced on August 5, 2026, the formation of an in-house custom silicon team to design proprietary AI chips, aiming to cut inference costs and reduce dependence on Nvidia, AMD, and Google. …
Anthropic confirmed on August 5, 2026 that it is assembling an in-house custom silicon team to design AI chips tailored to its Claude models, hiring a veteran from Google's chip organization to lead t…
Dell's Pro Precision 5 16s AMD, featuring the Ryzen AI 9 HX PRO 475 with 12 cores, 24 threads, and a 60 TOPS NPU, starts at $2,118.70 and approximately $5,087 as configured with 64GB RAM and 1TB SSD. …
AMD reported record revenue of $11.54 billion for Q2 2026, with Data Center revenue of $6.72 billion up 107% year over year, while NVIDIA reported $82 billion in quarterly revenue with Data Center rev…
Alphabet's robotaxi subsidiary Waymo has developed a custom chip for its self-driving taxis, reducing its reliance on Nvidia and AMD. The chip, manufactured by TSMC on a 5-nanometer process, delivers …
Linux networking maintainers say AI-generated patches have left them 'completely overwhelmed,' with 632 fixes and 648 feature patches merged for Linux 7.3, and an estimated one-third to one-half of th…