Nvidia Joins War Against COVID-19 With AI, GPU Supercomputing
It’s genuinely not clear to me whether bringing exaflops of processing power to bear on a single virus can yield material improvements in killing it, but we seem increasingly likely to find out the answer as time goes by. Folding@Home usage is growing, Rosetta@Home has new projects (including smartphone and Raspberry Pi support), and now a task force from Nvidia has joined the COVID-19 High-Performance Computing Consortium.
The goal of the COVID-19 HPCC is to provide scientists with access to 400 petaflops worth of supercomputer divided between 30 systems, as the single fastest supercomputer, IBM’s Summit, can sustain roughly 147 petaflops of performance.
Nvidia has a unique position in the supercomputing world. While they don’t manufacture an HPC-specific CPU architecture the way that Intel and AMD do, they more or less invented the AI and GPGPU markets out of whole cloth starting about 12 years ago, with the launch of the original G80. Nvidia’s effort will be led by Ian Buck, VP and general manager of Accelerated Computing.
“The COVID-19 HPC Consortium is the Apollo Program of our time,” Buck said. “Not a race to the moon, this is a race for humanity. The rocket ships are GPU supercomputers, and their fuel is scientific knowledge. NVIDIA is going to help by making these rockets travel as fast as they can.”
Nvidia will assist in three ways: It will apply AI principles to help the HPCC analyze more data, more quickly; it will “accelerate science” (it’s implied that having a great deal of expertise at the company will improve the rate at which these workloads can be processed); and the company will aid in optimizing computing at scale. Nvidia notes that improvements from optimization are massive when talking about 400 petaflop systems — a 10-20 percent gain in computational throughput is itself equal to the performance of one of the larger supercomputers on Earth.
Other partners in the HPCC include Amazon Web Services, AMD, Google Cloud, Hewlett Packard Enterprise, IBM, Microsoft, and now Nvidia. Researchers who want to submit COVID-19-related research proposals for evaluation can do so via this portal. A list of active projects is available at the bottom of this page.
It’s going to be interesting to see if anything comes from this massive assault on the virus. It wouldn’t surprise me if nothing does, at least not right away — while an awful lot of computational horsepower is being hurled at SARS-CoV-2, it’s not as if anyone had a lot of time to plan for this or deploy hardware in the most advantageous manner. Folding@Home has faced difficulties ramping up the availability of work units, for example, simply because the network has grown so much, so quickly.
We may have to wait for some time before we know if any of the work has been useful. Even if breakthroughs come too late to help for COVID-19, they could be instrumental in helping us understand coronaviruses more generally or dealing with a future pandemic should one occur.
- At Stanford’s AI Conference, Harnessing Tech to Fight COVID-19
- Here’s How New York City Will Split Ventilators to Treat COVID-19
- YouTube Says It Will Remove 5G Misinformation After People Burn Cell Towers