Intel (INTC) unveiled a new AI chip on Tuesday aimed at competing with Nvidia’s (NVDA) popular AI processors as part of its Intel Vision event.
The company’s Gaudi 3 chip not only matches but surpasses Nvidia’s H100 AI processor in training and deploying generative AI models. The H100 is widely used across various tech giants, including Microsoft and Google, for artificial intelligence applications.
According to Intel, the Gaudi 3 boasts 40% greater power efficiency and 50% faster inferencing than Nvidia’s H100. Training involves the time taken to train an AI foundation model like OpenAI’s GPT-3, while inferencing refers to the actual usage of these models in applications.
Intel claims the Gaudi 3 is up to 1.7 times faster in training common large language models compared to the H100. Moreover, Intel asserts that its chip outpaces Nvidia’s higher-end Nvidia H200 by an average of 1.3 times in inferencing certain language models.
