Nvidia Creates New Supercomputer Chip For Chinese language Market

Posted on

Nvidia has created a brand new knowledge heart GPU for the Chinese language market, designed to satisfy new US export management guidelines. In line with a report from Reuters (opens in new tab), the chip is named the A800, and can be utilized rather than the A100 for a similar duties.

(Picture credit score: Nvidia)

The A100, utilized in supercomputers by the likes of Oracle (opens in new tab), Tesla (opens in new tab) and Meta (opens in new tab), is a Tensor core GPU primarily based on the Ampere structure (as within the RTX 30-series). They’re closely used for AI, and deep studying duties equivalent to picture and speech recognition, in addition to high-performance computing. They characteristic as much as 80GB of reminiscence per chip, the world’s quickest reminiscence bandwidth at 2TB/s, and the power to hyperlink collectively via PCIe Gen 4 enabling hundreds of GPUs to work in unison.

Supply hyperlink

Leave a Reply

Your email address will not be published. Required fields are marked *