NVIDIA has unveiled the Grace microchip, its first processor designed specifically for data centers.
Grace is a processor for rapidly processing large amounts of data for use in training artificial intelligence models.
It is suitable for systems based on “giant AI” that control automated and robotic factories, as well as “smart” cities with hundreds of thousands of sensors, sensors, and other IoT devices.
As the company itself assures, the chip will be 10 times more productive than the most advanced competing chips on the market today when performing ultra-complex calculations. And its throughput will reach 900 Gbps, which is “30 times higher than today’s fastest servers.””
Data exchange with RAM at a speed of up to 500 GB/s and connection to the GPU is carried out via the NVLINK bus at a speed of up to 900 GB/s.
The Swiss National Supercomputing Center (CSCS) and Los Alamos National Laboratory of the United States will be the first organizations to build new supercomputers based on NVIDIA Grace processors.