AMD brings enterprise AI acceleration to existing data centres via MI350P PCIe GPUs

AMD has launched the Instinct MI350P PCIe GPUs, designed to assist organizations in running AI workloads on standard infrastructure. These GPUs allow enterprises to adopt AI without significant redesign of existing setups, facilitating easier integration of AI acceleration. The MI350P features a dual-slot PCIe design, fitting into air-cooled servers and supporting up to eight accelerator cards. With performance reaching 2,299 teraflops and peak performance of 4,600 teraflops using MXFP4 precision, they are suitable for small to large AI models in enterprise data centers.