Nvidia reveals Grace, a high-powered Arm-based server CPU that leverages its Ampere GPU tech
Nvidia is heading well down the path of marrying Arm-based CPUs with its graphics technology which was previously the domain of x86 processors from Intel and AMD. Among several announcements which includes a new partnership with MediaTek that will bring its RTX graphics tech to Chromebooks and other Arm-based PCs, the company has announced its own initiative. ‘Grace’ is the name that Nvidia has given to its new high-powered Arm-based server CPU that will be 10x faster than current servers when processing complex AI-based tasks including natural language processing.
Nvidia’s Grace CPU, which gets its name from Grace Hopper, a US computer programming pioneer and Navy rear admiral, uses Nvidia’s NVLink interconnect technology to create a massive 900 GB/s connection between it and Nvidia’s Ampere GPU technology. Nvidia says this creates a 30x higher aggregate bandwidth compared with today’s server technology. It also uses a LPDDR5x memory subsystem that delivers twice the bandwidth and 10x better energy efficiency compared to DDR4 memory.
When Nvidia Grace ships in 2023, it will be supported by an Nvidia HPC (High Performance Computing) software development kit and a full suite of CUDA and CUDA-X libraries that can be used to accelerate more than 2,000 GPU applications. This will also benefit scientists and researchers as well as data centers. While Nvidia’s planned US$40 billion acquisition of Arm may be facing potential roadblocks, the company is committed to Arm technology in its own processors, which are also finding a home in many of today’s high-tech cars.