Beijing Business Daily News (Reporter Yang Yuehan) - On the early morning of March 19th, at the NVIDIA GPU Technology Conference (GTC), NVIDIA CEO Huang Renxun announced the launch of a new generation of AI chip architecture, Blackwell, which has achieved comprehensive performance and power consumption surpassing the previous generation. The first Blackwell chip is called GB200 and will be launched later this year.
This new product is known by the outside world as the "strongest" AI chip. Taking the GB200 NVL rack scale system as an example, equipped with 36 Grace Blackwell superchips, compared to the H100 Tensor Core graphics processor with the same number of chips, the inference performance can be improved by up to 30 times, but the power consumption and cost are reduced by 1/25 of the previous generation.
Huang Renxun stated that many companies expect to adopt Blackwell, such as Amazon Web Services, Dell, Google, Meta, Microsoft, OpenAI, Oracle, Tesla, and xAI. Blackwell supports building real-time generative AI. Huang Renxun believes that Blackwell will become Nvidia's most successful product launch.