The Stanford-conceived distributed computing project Folding@Home, which uses donated compute power from people’s home PCs to crunch biomedical problems, has ballooned in power since it announced it will be tackling COVID-19 related problems a few weeks ago. The project stats page now shows a total compute performance of over 1.5 ExaFLOPS (aka 1,500,000,000,000,000,000 floating point operations per second), which dwarfs the computing ability of even the most powerful mainframe supercomputers on the planet.
To put that very large number in perspective, consider IBM’s recent announcement that the company is working with the U.S. Department of Energy and others to pool the power of several supercomputers for coronavirus-related protein crunching. The initiative harnesses 16 of the world’s most powerful supercomputers, including those operated by MIT and NASA, to perform 330 PetaFLOPS worth of COVID-19 calculations—that’s still just 0.33 ExoFLOPS, less than a quarter of what Folding@Home is now capable of.
The fact that the project is able to generate such a colossal amount of compute power simply by harnessing the idle CPU (processor) and GPU (video card) time of people’s personal computers is a testament to how much public interest the endeavour has garnered. According to HotHardware, which runs their own team of contributors to the Folding@Home project, only 98 PetaFLOPS of work was being done by Folding@Home users as of the beginning of March. That means the project has seen a more than fifteenfold increase in its donated resources since they announced their pivot to coronavirus-related research.
This meteoric rise in Folding@Home’s capabilities has also been buoyed by the support of some heavy hitters in the PC gaming space, from companies like Ubisoft and NVIDIA to the likes of Linus Tech Tips and reddit’s unfortunately-named-but-very-large PC gaming community. This has allowed the project to tap into a group of people who own a disproportionately large amount of computing power, as PC gaming generally requires a reasonably powerful graphics card that in most cases happens to be very good at performing the kind of calculations Folding@Home requires.
The ultimate goal behind harnessing all this compute power is to help develop a vaccine for COVID-19 by modeling how the virus’s proteins move around. Dr. Greg Bowman, the director of Folding@Home, tweeted an example of how exactly this works. The protein depicted in his example is an Ebola protein, but the same principle applies; what they’re looking for is an instance in which the protein moves in a way that reveals a “pocket” that may allow a vaccine to bind with the protein and begin attacking it.
Folding@Home has produced a wealth of peer-reviewed scientific research over its 20 year history, even without the immense firepower now at their disposal, so the chances of the project having a meaningful role to play in finding a coronavirus vaccine seem high. Regardless of whether this particular initiative pays off, however, the project’s current capabilities are a welcome reminder of the power of human solidarity. It’s important to recognize that a bunch of regular folks collectively donating their regular home computer hardware to a good cause has resulted in a distributed computing network over ten times as powerful as the world’s most powerful supercomputer. There might even be a metaphor for how to approach other problems threatening humanity in there somewhere.