Nvidia says Grace will deliver 10x the performance of today’s fastest servers.
The processor will be available in 2023.
Nvidia is making Arm-based CPUs again. The Santa Clara, California-based GPU giant announced its first data centre CPU, Nvidia Grace, last night at GDC 2021. Without divulging too many technical details, Nvidia said Grace will deliver 10x the performance of today’s fastest servers for AI and high-performance computing workloads. The processor will be available in 2023.
Grace isn’t designed for your typical run-of-the-mill data centres though. It will instead serve a “niche segment of computing” and target workloads such as “training next-generation NLP models that have more than 1 trillion parametres.” In other words, Grace – which is named after computing pioneer Grace Hopper – is meant to power supercomputers.
Meet the NVIDIA Grace #CPU, leveraging the flexibility of @Arm’s data center architecture and designed from the ground up to accelerate the largest HPC and AI workloads. #GTC21 https://t.co/PHDaxrfzQv pic.twitter.com/uck0akde3a
“Leading-edge AI and data science are pushing today’s computer architecture beyond its limits – processing unthinkable amounts of data,” Nvidia founder and CEO Jensen Huang said in a statement. “Using licensed Arm IP, Nvidia has designed Grace as a CPU specifically for giant-scale AI and HPC. Coupled with the GPU and DPU, Grace gives us the third foundational technology for computing, and the ability to re-architect the data centre to advance AI. Nvidia is now a three-chip company.”
The Swiss National Supercomputing Centre (CSCS) and the U.S. Department of Energy’s Los Alamos National Laboratory are first to announce plans to build Grace-powered supercomputers. These supercomputers will go online in 2023 and will be used for national scientific research.
The announcement is significant for many reasons. One if of course the fact that Nvidia is making CPUs again. The other — and this is the most significant aspect — is the timing of the announcement. It comes only months after Nvidia confirmed that it was acquiring Arm from SoftBank for $40 billion to create “the world’s premier computing company for the age of AI.” Grace is the first step in this new journey even as Arm continues to have a field day being part of some of the biggest announcements the tech industry has seen over the last year.
“As the world’s most widely licensed processor architecture, Arm drives innovation in incredible new ways every day,” Arm CEO Simon Segars said. “Nvidia’s introduction of the Grace data centre CPU illustrates clearly how Arm’s licensing model enables an important invention, one that will further support the incredible work of AI researchers and scientists everywhere.”
Although Nvidia isn’t sharing a lot of information about Grace just yet, it has given a sneak peak of what to expect. Nvidia says it is using its fourth generation NVLink interconnect technology with a record 900 GB/s connection between the CPU and GPU. This will enable up to 30x higher aggregate bandwidth compared to today’s leading servers.
Grace will also use LPDDR5x memory subsystem for twice the bandwidth and 10x better energy efficiency compared with DDR4 memory.