AMD Benchmarks a 512-GPU MI355X Cluster in MLPerf 6.1 as NVIDIA Maintains Strong Position

MLCommons has released MLPerf Inference v6.1, the latest edition of its artificial intelligence benchmark suite. The release includes workflows designed to reflect recent trends in the AI sector and results from AMD, NVIDIA, and Intel.

AMD benchmarked a 512-GPU cluster based on its MI355X hardware. NVIDIA retained a strong position with results involving its Blackwell platform and its first Vera Rubin submission. Intel also submitted results using Xeon and Arc Pro hardware.

The results cover a range of AI benchmarks, including DeepSeek R1. MLCommons presents the suite as a showcase for current and latest hardware from the participating companies. The submissions provide a comparison of how AMD, NVIDIA, and Intel hardware performed across the workloads included in MLPerf Inference v6.1.