NVIDIA Hopper-based GPU tops inference benchmarks

The NVIDIA H100 Tensor Core GPUs has topped the MLPerf industry-standard AI benchmarks, scoring up to 4.5x better performance than previous-generation GPUs.

Also known as Hopper, the H100 raised the bar in per-accelerator performance across all six neural networks, leading in throughput and speed in separate server and offline scenarios.

NVIDIA A100 Tensor Core GPUs and the NVIDIA Jetson AGX Orin module for AI-powered robotics also led in inference performance across all MLPerf tests — image and speech recognition, natural language processing and recommender systems.

Since their debut on MLPerf in July 2020, A100 GPUs have advanced their performance by 6x, thanks to AI software improvements.

The NVIDIA Orin won more tests than any other low-power system-on-a-chip. It recorded a 50 percent gain in energy efficiency compared to its debut on MLPerf in April.

MLPerf is a consortium of AI leaders from academia, research labs and industry who aims to provide unbiased evaluations of training and inference performance for hardware, software and services conducted under prescribed conditions. The benchmarks help enterprises make more informed buying decisions.