In addition to the presentation of the new AMD Ryzen Pro 7000 processors for business notebooks and workstations, Lisa Su’s company has organized for these days the announcement of MI300X, the most advanced GPU for artificial intelligence signed by the red team. The CEO of the US giant has unveiled the AI chip and officially launched the gauntlet towards NVIDIA, defining herself ready to respond to the company’s “largest and most strategic long-term growth opportunity”.
AMD unveils the MI300X accelerator
Defined by AMD itself as an “accelerator”, the MI300X chip exploits the CDNA architecture to work with large language models and other state-of-the-art AI solutions, enabling content generation and making it ever more optimized. With up to 192GB of memory supported, the MI300X chip nimbly outperforms NVIDIA’s H100 rival – at least on paper – and runs nimbly on models with over 40 billion parameters.
Furthermore, AMD has stated that it will offer aInfinity architecture combining eight of its MI300X accelerators in a single system. Lisa Su then added, in this regard, that it is the right solution for particularly large AI models.
The MI300X chip will be available this fall for early testing and will ship in larger volumes next year, but there’s no clear price tag yet. In all likelihood, however, AMD will play to the bottom against NVIDIA which, we recall, also offers the H100 GPU for over $30,000. In order to reach more and more developers, server manufacturers and AI solutions, Lisa Su and associates could reduce the list price and make generative AI applications less expensive, therefore more accessible to end users as well.