MLCommons releases benchmarks for AI applications' speed and energy consumption, with Llama 2 and Stable Diffusion XL tested on Nvidia H100-powered servers.
MLCommons, an AI benchmarking group, has released new tests and results measuring the speed of responses from AI applications and systems. This includes Llama 2, a large language model with 70 billion parameters, and Stable Diffusion XL, a text-to-image generator. Servers powered by Nvidia's H100 chips from companies like Google, Supermicro, and Nvidia themselves performed well in raw performance. However, energy consumption is also a significant factor when deploying AI applications, with MLCommons having a separate benchmark category for measuring power consumption.
March 27, 2024
18 Articles