NVIDIA’s Arm-based Grace chips gather momentum in Taiwan

The momentum is gathering for the NVIDIA Grace CPU Superchip and Grace Hopper Superchip. Taiwan’s leading computer makers are set to introduce dozens of server models powered by the Arm-based chips for a wide range of workloads, including AI, digital twins, cloud graphics, gaming, and high performance computing (HPC).  

ASUS, Foxconn Industrial Internet, GIGABYTE, QCT, Supermicro, and Wiwynn are expected to roll out Grace-powered systems in the first half of 2023.

“These new systems from our partners, powered by our Grace Superchips, will bring the power of accelerated computing to new markets and industries globally,” said Ian Buck, Vice President of  Hyperscale and HPC at NVIDIA.

Designed to power AI factories, the two NVIDIA Grace Superchip technologies will enable a broad range of compute-intensive workloads across a multitude of system architectures. 

The Grace Hopper Superchip is a single superchip module with a direct chip-to-chip connection between the Grace CPU and the Hopper GPU. With NVLINK-C2C connectivitiy, Grace will transfer data to Hopper 15x faster than any other CPU can and will increase the working data size of Hopper up to 2TB. It is built to accelerate the AI, HPC, cloud and hyperscale workloads.

The Grace CPU Superchip consists of two Grace CPU chips connected coherently over NVLink-C2C. It has 144 CPU cores, 1TB/s of memory bandwidth and twice the energy efficiency of existing servers, consuming only 500W.

The Grace server portfolio is optimised for NVIDIA’s rich computing software stacks, including NVIDIA HPC, NVIDIA AI, Omniverse, and NVIDIA RTX. 

NVIDIA is extending its NVIDIA-Certified Systems programme to servers using the NVIDIA Grace CPU Superchip and Grace Hopper Superchip. The first certifications are expected soon after partner systems ship.