Both Nvidia and Intel’s Habana Labs reported performance gains with the MLPerf 2.1 benchmark in training machine learning models.