- ‘Grace’ CPU delivers 10x performance leap for methods coaching big AI fashions, using energy-efficient Arm cores
- Swiss Supercomputing Center and US Department of Energy’s Los Alamos National Laboratory First to Build NVIDIA CPU-Powered Supercomputers
SANTA CLARA, Calif., April 12, 2021 (GLOBE NEWSWIRE) — GTC — NVIDIA right this moment introduced its first knowledge heart CPU, an Arm-based processor that may ship 10x the efficiency of right this moment’s quickest servers on essentially the most complicated AI and excessive efficiency computing workloads.
The results of greater than 10,000 engineering years of labor, the NVIDIA Grace™ CPU is designed to deal with the computing necessities for the world’s most superior purposes — together with natural language processing, recommender systems and AI supercomputing — that analyze monumental datasets requiring each ultra-fast compute efficiency and large reminiscence. It combines energy-efficient Arm CPU cores with an modern low-power reminiscence subsystem to ship excessive efficiency with nice effectivity.
“Leading-edge AI and data science are pushing today’s computer architecture beyond its limits – processing unthinkable amounts of data,” stated Jensen Huang, founder and CEO of NVIDIA. “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 center to advance AI. NVIDIA is now a three-chip company.”
Grace is a extremely specialised processor concentrating on workloads akin to coaching next-generation NLP fashions which have greater than 1 trillion parameters. When tightly coupled with NVIDIA GPUs, a Grace CPU-based system will ship 10x quicker efficiency than right this moment’s state-of-the-art NVIDIA DGX™-based methods, which run on x86 CPUs.
While the overwhelming majority of knowledge facilities are anticipated to be served by present CPUs, Grace — named for Grace Hopper, the U.S. computer-programming pioneer — will serve a distinct segment section of computing.
The Swiss National Supercomputing Centre (CSCS) and the U.S. Department of Energy’s Los Alamos National Laboratory are the primary to announce plans to construct Grace-powered supercomputers in help of nationwide scientific analysis efforts.
NVIDIA is introducing Grace as the quantity of knowledge and dimension of AI fashions are rising exponentially. Today’s largest AI fashions embrace billions of parameters and are doubling each two-and-a-half months. Training them requires a brand new CPU that may be tightly coupled with a GPU to eradicate system bottlenecks.
NVIDIA constructed Grace by leveraging the unimaginable flexibility of Arm’s knowledge heart structure. By introducing a brand new server-class CPU, NVIDIA is advancing the purpose of know-how range in AI and HPC communities, the place alternative is essential to delivering the innovation wanted to resolve the world’s most urgent issues.
“As the world’s most widely licensed processor architecture, Arm drives innovation in incredible new ways every day,” stated Arm CEO Simon Segars. “NVIDIA’s introduction of the Grace data center 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.”
Grace’s First Adopters Push Limits of Science and AI
CSCS and Los Alamos National Laboratory each plan to convey Grace-powered supercomputers, constructed by Hewlett Packard Enterprise, on-line in 2023.
”NVIDIA’s novel Grace CPU permits us to converge AI applied sciences and traditional supercomputing for fixing a number of the hardest issues in computational science,” stated CSCS Director Prof. Thomas Schulthess. “We are excited to make the new NVIDIA CPU available for our users in Switzerland and globally for processing and analyzing massive and complex scientific datasets.”
“With an innovative balance of memory bandwidth and capacity, this next-generation system will shape our institution’s computing strategy,” stated Thom Mason, director of the Los Alamos National Laboratory. “Thanks to NVIDIA’s new Grace CPU, we’ll be able to deliver advanced scientific research using high-fidelity 3D simulations and analytics with datasets that are larger than previously possible.”
Delivering Breakthrough Performance
Underlying Grace’s efficiency is fourth-generation NVIDIA NVLink® interconnect know-how, which gives a report 900 GB/s connection between Grace and NVIDIA GPUs to allow 30x greater mixture bandwidth in comparison with right this moment’s main servers.
Grace can even make the most of an modern LPDDR5x reminiscence subsystem that may ship twice the bandwidth and 10x higher vitality effectivity in contrast with DDR4 reminiscence. In addition, the brand new structure gives unified cache coherence with a single reminiscence tackle house, combining system and HBM GPU reminiscence to simplify programmability.
Grace shall be supported by the NVIDIA HPC software development kit and the total suite of CUDA® and CUDA-X™ libraries, which speed up greater than 2,000 GPU purposes, dashing discoveries for scientists and researchers engaged on the world’s most necessary challenges.
Availability is predicted at first of 2023.
NVIDIA’s (NASDAQ: NVDA) invention of the GPU in 1999 sparked the expansion of the PC gaming market and has redefined fashionable laptop graphics, excessive efficiency computing and synthetic intelligence. The firm’s pioneering work in accelerated computing and AI is reshaping trillion-dollar industries, akin to transportation, healthcare and manufacturing, and fueling the expansion of many others. More data at https://nvidianews.nvidia.com/.
For additional data, contact:
Senior Director, Enterprise Communications
Certain statements on this press launch together with, however not restricted to, statements as to: the options, efficiency, advantages, affect and availability of the Grace CPU; CSCS’s and Los Alamos National Laboratory’s plans to construct Grace-powered supercomputers and convey them on-line by 2023; the expansion of AI fashions; knowledge facilities being served by CPUs; Grace serving a distinct segment section of computing; modern AI and knowledge science pushing right this moment’s laptop structure past its limits — processing unthinkable quantities of knowledge; the coaching of AI fashions requiring a brand new CPU; NVIDIA advancing the purpose of know-how range in AI and HPC communities; the pliability of Arm’s open structure driving innovation; and Arm enabling invention and supporting the work of AI researchers and scientists are forward-looking statements which can be topic to dangers and uncertainties that would trigger outcomes to be materially completely different than expectations. Important components that would trigger precise outcomes to vary materially embrace: world financial situations; our reliance on third events to fabricate, assemble, bundle and take a look at our merchandise; the affect of technological growth and competitors; growth of recent merchandise and applied sciences or enhancements to our present product and applied sciences; market acceptance of our merchandise or our companions’ merchandise; design, manufacturing or software program defects; modifications in client preferences or calls for; modifications in business requirements and interfaces; surprising lack of efficiency of our merchandise or applied sciences when built-in into methods; in addition to different components detailed now and again in the newest stories NVIDIA information with the Securities and Exchange Commission, or SEC, together with, however not restricted to, its annual report on Form 10-Okay and quarterly stories on Form 10-Q. Copies of stories filed with the SEC are posted on the corporate’s web site and can be found from NVIDIA with out cost. These forward-looking statements will not be ensures of future efficiency and converse solely as of the date hereof, and, besides as required by legislation, NVIDIA disclaims any obligation to replace these forward-looking statements to mirror future occasions or circumstances.
© 2021 NVIDIA Corporation. All rights reserved. NVIDIA, the NVIDIA emblem, CUDA, CUDA-X AI, DGX, NVIDIA Grace, and NVLink are logos and/or registered logos of NVIDIA Corporation within the U.S. and different nations. Other firm and product names could also be logos of the respective firms with which they’re related. Features, pricing, availability and specs are topic to alter with out discover.
A photograph accompanying this announcement is offered at https://www.globenewswire.com/NewsRoom/AttachmentNg/b3cfff26-6195-4f21-a7cc-55f7f16af7a6