NVIDIA Blackwell Achieves 2.6x Higher Performance in MLPerf Training v5.0

NVIDIA Blackwell delivers up to 2.6x higher performance in MLPerf Training v5.0, achieving the fastest time to train across all seven benchmarks. This includes LLM pretraining, LLM fine-tuning, text-to-image generation, recommender systems, graph neural networks, natural language processing, and object detection. NVIDIA Blackwell incorporates architectural innovations, such as the second-generation Transformer Engine, faster and wider NVIDIA NVLink interconnects, and higher-bandwidth and higher-capacity HBM3e memory, to enable faster model training.