News

AMD provided a full block diagram of the MI300X architecture, with each XCD containing 2 compute units disabled, for a total of 304 CUs in MI300X out of the full 320 CU design.
AMD's Instinct MI300X is a beast of an AI accelerator - built on the company's third-generation CDNA architecture and TSMC's advanced 5nm and 6nm processes, it features 19,456 stream processors ...
AMD Instinct MI300X GPUs are built on the AMD CDNA 3 architecture and are designed for AI - including generative AI - and high-performance computing workloads. The GPUs feature 19,456 stream ...
AMD’s Instinct MI300X relies on a chiplet layout that places a large block of memory right beside the compute silicon. This architecture is tailor-made for heavy agentic AI inference work.
AMD tested the MI300X with Genoa and Turin CPUs, showing good performance scaling. The MI300X's peak FP16 performance is 1,307.4 teraflops, generating 2,530.7 tokens per second in server mode.
AMD is set to release their Instinct MI325X in Q4, with MI400 series, powered by CDNA "Next" architecture, expected in 2026. I think AMD has a solid R&D roadmap for the fast-growing GPU market.
AMD Instinct MI300X GPUs are built on the AMD CDNA 3 architecture and are designed for AI - including generative AI - and high-performance computing workloads. The GPUs feature 19,456 stream ...
AMD announced the details as part of a newly disclosed plan to release a new data center GPU every year starting with the MI325X, which, like the MI300X, uses the company’s CDNA 3 architecture ...
The AMD design aims to compete with Nvidia’s Blackwell architecture. The MI325X chip uses the same architecture as the already-available MI300X, which AMD launched last year.
AMD's Instinct MI300X is a beast of an AI accelerator - built on the company's third-generation CDNA architecture and TSMC's advanced 5nm and 6nm processes, it features 19,456 stream processors ...
AMD Instinct MI300X GPUs are built on the AMD CDNA 3 architecture and are designed for AI - including generative AI - and high-performance computing workloads. The GPUs feature 19,456 stream ...