New benchmark tests speed of training ChatGPT-like chatbots

MLCommons, a group that creates benchmark tests for artificial intelligence (AI) technology, has unveiled a new test to determine the system speeds used for training chatbots like ChatGPT. Nvidia has come out on top, with its latest H100 chip enabling faster training. The MLPerf benchmark is based on the AI model GPT-3, which is used to train ChatGPT, the viral chatbot developed by OpenAI and backed by Microsoft. Only two chip firms submitted results for the benchmark, with Nvidia’s largest system submitting 3,584 H100 chips, resulting in a training time of 10.94 minutes. Intel’s Habana Labs ran the benchmark in 311.945 minutes with a smaller system equipped with 384 Gaudi2 chips. The results demonstrate the potential of Gaudi2, which will have a software update in September to boost speed. Intel said the industry needs a second supplier of chips for AI training, and the MLPerf results show Intel can fill that need.