NVIDIA has laid out its plans for the next generation of computing, with AI poised to play a major role. Company founder and CEO Jensen Huang recently delivered a keynote from his California kitchen, due to COVID-19, discussing NVIDIA's data center vision as well as new AI hardware and software announcements.
The company is now shipping the third generation of its NVIDIA DGX AI system, the A100, touted as the world's first 5-petaflops server. Each DGX A100 may be divided into up to 56 independently running applications. A single server may also be scaled up to race through tasks like AI training, and also scaled out for AI deployment and inference.
The system is already being used by the U.S. Department of Energy's Argonne National Laboratory to cluster AI and computing power in the fight against COVID-19. The University of Florida and the German Research Center for Artificial Intelligence are also using the system. The solution will be available to NVIDIA's cloud and partner server makers under the HGX A100 moniker.
To put the power of the system in perspective, Huang explained that a data center powered by five DGX A100 systems for AI training and inference can run on just 28 kw of power and cost $1 million. This type of system can handle the workload of a typical data center using 50 DGX-1 systems and 600 CPU systems, consuming 630 kw of power at a cost of more than $11 million.
NVIDIA also rolled out the next-generation DGX SuperPOD, offering 700 petaflops of AI performance. Huang said the offering, which is powered by 140 DGX A100 systems with networking technology from Mellanox, acquired by NVIDIA last month, is the equivalent of one of the 20 fastest computers on the planet.
“The data center is the new computing unit,” said Huang, emphasizing that NVIDIA is focusing on accelerating performance gains from silicon to the connection of CPUs and GPUs and ultimately across entire data centers.
The company is putting its money where its mouth is, and is expanding its own data center using four DGX SuperPODs. The move will add 2.8 exaflops of AI computing power for a total of 4.6 exaflops to NVIDIA's SATURNV internal supercomputer and will make it the world's fastest AI supercomputer.
NVIDIA is also working with researchers and scientists to use GPUs and AI to treat, mitigate, contain and track COVID-19 globally. According to Huang, Oxford Nanopore Technologies has used the company's solutions to sequence the virus genome in just seven hours. Oak Ridge National Laboratory and Scripps Research Institute have screened one billion potential drug combinations in a day, and Plotly is using NVIDIA for real-time infection rate tracing. Structura Biotechnology, the University of Texas at Austin and the National Institutes of Health are also using the technology to reconstruct the 3D structure of the COVID-19 spike protein.
“Researchers and scientists applying NVIDIA accelerated computing to save lives is the perfect example of our company’s purpose — we build computers to solve problems normal computers cannot,” said Huang.
To learn more about how AI and supercomputing are transforming software stacks, data centers and the future of work globally, TMC is hosting a Future of Work Expo from June 22-25, 2021 at the Miami Beach Convention Center. The event will examine how AI and supercomputing are changing the workforce and the global information economy as we know it, with applications in the data center, contact center, customer service, sales and marketing and beyond.
Edited by
Maurice Nagle
