Do you remember when Nvidia announced the Grace CPU? The company combined two to create the Grace CPU Superchip, a 144-core server processor for HPC, AI, data analytics, scientific computing, and large-scale computing applications.
According to Nvidia, upcoming Arm Neoverse-based processors will feature 2 Grace processors interconnected using NVLink-C2C technology, allowing it to deliver up to 1TB/s of memory bandwidth. Each of the processors integrated into the Superchip has 72 Arm v9 cores, for a total of 144 cores. Additionally, Nvidia Grace CPU Superchip will support Nvidia’s computer software stacks, including Nvidia RTX/HPC/AI and Omniverse.
The Grace CPU Superchip has an estimated performance rating of 740 on the SPECrate 2017_int_base benchmark, about 50% better than Nvidia’s dual-processor DGX A100 systems.
The Superchip Grace CPU supports LPDDR5x ECC memory, ideal for creating a more power-efficient system. When the Grace CPU Superchip is combined with LPDDR5x, the combined power consumption of the two components is approximately 500W.
Nvidia is already working with HPC, supercomputing, hyper-scale and cloud partners to launch the Grace CPU Superchip. The company plans to make it available alongside the Grace Hopper superchip in the first half of next year.
KitGuru says: Nvidia’s approach to creating the Grace CPU Superchip reminds us of Apple’s design for the M1 Ultra. Simply put, the two companies took two of their chips and joined them together using a high-speed bus so that the system would see it as a single processor. Maybe we’ll see other Arm processor manufacturers try something similar in the years to come.
Become a patron!