Shulou(Shulou.com)11/24 Report--

Thanks to CTOnews.com netizens for the delivery of clues about the past. CTOnews.com, November 9, according to Science and Technology Innovation Board Daily and Blue Whale Finance, the industry chain said that Nvidia has now developed the latest improved version of AI chips for China, including HGX H20, L20 PCle and L2 PCle. As of press time, Nvidia has not responded yet.

The three chips are all based on improvements to the Nvidia H100, and Nvidia is expected to announce them as soon as November 16, and domestic manufacturers will get samples as soon as these days, according to people familiar with the matter.

According to the public information of CTOnews.com inquiry, NVIDIA H100 Tensor Core GPU adopts the new Hopper architecture and integrates 80 billion transistors based on TSMC N4 process. Compared with the previous generation product, it can provide 9 times higher training speed for the multi-expert (MoE) model.

It is equipped with a fourth-generation Tensor Core and Transformer engine (FP8 precision), as well as highly scalable NVLink interconnection technology (up to 256H100 GPU, 9 times higher bandwidth and 900GB/s bandwidth speed than the previous generation of HDR Quantum InfiniBand networks).

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.