MI350 Series Marks AMD’s Push for a Larger Slice of AI

AMD datacenter GPU accelerator MI350 series delivers 4× AI compute and 35× inferencing gains, with CDNA 4 architecture and TSMC 3nm process.

AMD's newest datacenter GPU accelerators hit the shelves at a pivotal moment in the AI and high performance computing market, where demand is surging and NVIDIA continues to dominate. Rather than aiming for outright leadership, AMD is positioning the MI350X and MI355X GPUs as compelling alternatives, emphasizing strong performance, massive memory capacity, and open-standards infrastructure. AMD's Instinct MI350 series GPUs, including the MI350X and MI355X accelerators, are built on AMD's advanced CDNA 4 architecture and TSMC's 3nm process. MI350 series GPUs deliver up to 4× higher AI compute and a 35× leap in inferencing performance compared to the previous generation, while also upgrading the amount of memory by 50% and memory bandwidth by 51%.