NVIDIA Announces New Data Center Processors and Entire Solutions to Faster Computing With Grace™
GPU manufacturer Nvidia has announced its first CPU at the Nvidia GPU Technology Conference (GTC). This will be a Data Center CPU built for serious artificial intelligence (AI) and high-performance computing tasks. the NVIDIA Grace™ CPU is designed to address the computing requirements for the world’s most advanced applications — including NLP, AI supercomputing, and many more.
NVIDIA has designed Grace on Arm System IP, the CPU coupled with the GPU and DPU, Grace gives us complete power for computing, and the ability to excel data centers at giant-scale AI computing, DL, and High-performance computing.
The New NVIDIA Grace™ CPU Combined With Other Solutions Bring Amazing Performance Lift To The Data Center Computing
The CPU design generation has been named Grace after the computer programming pioneer – Grace Hopper. Nvidia said that the new Grace CPUs when coupled with NVIDIA GPUs, can deliver 10 times faster AI performance than current-gen Nvidia DGX-based systems with Intel x86 CPUs.
Nvidia’s Grace CPUs feature the fourth-generation NVIDIA NVLink® interconnect technology. To clarify, this goes to a whopping 900 GB/s connection speed with NVIDIA GPUs to enable 30x higher average bandwidth compared to the servers platforms of the current generation. Also, this features the LPDDR5x memory subsystem that gives 2x bandwidth and 10x power efficiency compared to the current DDR4 memories.
The new Grace architecture enables single memory address space. As a result, combining system and HBM GPU memory to unify cache coherence with the shared resource data. this will simplify the programmability of the applications further.
NVIDIA Grace will have the support of the NVIDIA HPC software development kit and the full suite of CUDA® and CUDA-X™ libraries, which accelerate more than 2,000 GPU applications. This will be necessary for ML, DL, HPC based use and will help the researchers and engineers working on complex and heavy workloads.
Data Centers That Are Preparing To Use The NVIDIAs New Grace™ CPU Based Systems in Future
The Swiss National Supercomputing Centre (CSCS) and the U.S. Department of Energy’s Los Alamos National Laboratory have decided to build their supercomputers on Grace based platform in support of national scientific research efforts. CSCS and Los Alamos National Laboratory both are planning to build their supercomputers on Grace-powered System. This is expected to happen by 2023.
Availability for the NVIDIA Grace™ CPU is expected in the early first quarter of 2023.