AMD reports faster MI455X inference results and adds MI450X GPU
AMD says new testing shows its MI455X GPU on the Helios rackscale system delivers 21,333 tokens per second per GPU on DeepSeek-R1 in an MLPerf Inference Offline configuration. AMD says that is up to 30% higher throughput than the latest published MLPerf offline result for Nvidia Vera Rubin. AMD is also introducing the MI450X, which has 288 GB of HBM4 memory per GPU and the same core compute, memory bandwidth, and connectivity as the MI455X. The MI455X keeps extra memory for larger models, longer contexts, or higher concurrency.