Desktop GPU Shipments Grow by One Third, AI PCs Not the Main Driver


GitHub code from an AMD AI executive accidentally revealed Anthropic as a customer, signaling its effort to diversify chip supply beyond a single vendor, per SemiAnalysis.....
Sony PS6 rumored to adopt video frame interpolation, exclusive super-resolution, and AI ray tracing with AI frame generation, aiming for 4K 120fps smooth experience.....
AMD released the vLLM-ATOM plugin, aiming to fully tap into hardware potential without changing the existing workflow, significantly accelerating the inference of mainstream large language models such as DeepSeek-R1 and Kimi-K2. vLLM is an open-source framework optimized for throughput and GPU memory utilization in high-concurrency scenarios, focusing on request scheduling and cache management. The ATOM plugin further enhances this capability.
AMD launched the vLLM-ATOM plugin, optimizing large language model deployment on AMD hardware. It boosts inference performance for Chinese models like DeepSeek-R1 and Kimi-K2 without altering existing workflows. Tailored for Instinct GPUs, it leverages vLLM's high memory efficiency, enabling low-cost technical migration and smooth performance upgrades.....
OpenAI has partnered with five major companies, including AMD, Broadcom, Intel, Microsoft, and NVIDIA, to launch the Multi-Path Reliable Connection (MRC) protocol, aimed at addressing network latency and failure issues in large-scale AI training. The protocol has been open-sourced through the Open Compute Project (OCP) and is driving a shift from a three-tier architecture to a two-tier design, breaking single points of failure and improving training stability and efficiency.