Nvidia is generating Arm-based CPUs once again. The Santa Clara, California-based GPU giant announced its 1st information centre CPU, Nvidia Grace, final evening at GDC 2021. Without divulging as well quite a few technical specifics, Nvidia stated Grace will provide 10x the functionality of today’s quickest servers for AI and higher-functionality computing workloads. The processor will be readily available in 2023.
Grace is not made for your standard run-of-the-mill information centres even though. It will as an alternative 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 just after computing pioneer Grace Hopper – is meant to energy supercomputers.
Meet the NVIDIA Grace #CPU, leveraging the flexibility of @Arm’s information center architecture and made from the ground up to accelerate the biggest HPC and AI workloads. #GTC21 https://t.co/PHDaxrfzQv pic.twitter.com/uck0akde3a
— NVIDIA GTC (@NVIDIAGTC) April 12, 2021
“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 stated 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 1st to announce plans to create Grace-powered supercomputers. These supercomputers will go on-line in 2023 and will be utilised for national scientific analysis.
| Rocket Lake-S deep dive: Intel on broken embargoes, benchmark scores, walled gardens and energy of Computer
The announcement is substantial for quite a few causes. One if of course the truth that Nvidia is generating CPUs once again. The other — and this is the most substantial aspect — is the timing of the announcement. It comes only months just after Nvidia confirmed that it was acquiring Arm from SoftBank for $40 billion to build “the world’s premier computing company for the age of AI.” Grace is the 1st step in this new journey even as Arm continues to have a field day getting element of some of the greatest announcements the tech market has noticed more than the final year.
“As the world’s most widely licensed processor architecture, Arm drives innovation in incredible new ways every day,” Arm CEO Simon Segars stated. “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.”
Nvidia getting into the CPU race once again also sets the alarm bells ringing for Intel, who by the way only lately announced its 3rd Gen Intel Xeon Scalable processor for information centres touting almost 50 % functionality jump and constructed-in AI. You can study more about that right here.
Intel shares dive just after Nvidia announces plans to make its 1st chip for servers and information centers https://t.co/DyneXDAzB0 pic.twitter.com/C1OwpFZpKb
— CNBC Now (@CNBCnow) April 12, 2021
Nvidia Grace: sneak peek
Although Nvidia is not sharing a lot of facts about Grace just but, it has offered a sneak peak of what to count on. Nvidia says it is employing its fourth generation NVLink interconnect technologies with a record 900 GB/s connection amongst the CPU and GPU. This will allow up to 30x greater aggregate bandwidth compared to today’s top servers.
Grace will also use LPDDR5x memory subsystem for twice the bandwidth and 10x greater power efficiency compared with DDR4 memory.