If there’s any doubt about NVIDIA’s relationship with Arm following the unsuccessful acquisition, the newly-announced NVIDIA Grace CPU Superchip provides a clear and resounding answer.
Based on Arm Neoverse, the discrete data centre CPU is designed for AI infrastructure and high performance computing, providing the highest performance and twice the memory bandwidth and energy-efficiency compared to today’s leading server chips.
It comprises not just one but two powerful CPU chips connected over NVLink-C2C for high-speed, low-latency and chip-to-chip interconnect.
Packing 144 Arm cores in a single socket, the chip delivers an estimated performance of 740 on the SPECrate2017_int_base benchmark — more than 1.5x higher compared to the dual-CPU shipping with the DGX A100 today.
It will run all of NVIDIA’s computing software stacks, including NVIDIA RTX, NVIDIA HPC, NVIDIA AI, and Omniverse.
“A new type of data centre has emerged — AI factories that process and refine mountains of data to produce intelligence. The Grace CPU Superchip offers the highest performance, memory bandwidth and NVIDIA software platforms in one chip and will shine as the CPU of the world’s AI infrastructure,” said Jensen Huang, Founder and CEO of NVIDIA.
NVIDIA is working with HPC, supercomputing, hyperscale and cloud customers on the Grace CPU Superchip, which is expected to be available in the first half of 2023.