Folding@Home Now More Powerful Than All the Supercomputers on Earth
The past six weeks have seen a lot of unprecedented activity across the globe, most of it bad. One bright spot in the gloom — though of course, a rather conditional one — has been the sudden explosion in computational power being flung at analyzing and understanding COVID-19.
It wasn’t that long ago we were reporting on the Folding@Home network’s record-breaking 1.5 exaFLOPS of performance. Now, the system has broken 2.5 exaflops. At that speed, it’s reportedly faster than the sustained performance of all Top500 supercomputers combined.
With our collective power, we are now at ~2.4 exaFLOPS (faster than the top 500 supercomputers combined)! We complement supercomputers like IBM Summit, which runs short calculations using 1000s of GPUs at once, by spreading longer calculations around the world in smaller chunks! pic.twitter.com/fdUaXOcdFJ
— Folding@home (@foldingathome) April 13, 2020
There’s a pretty significant asterisk to this, however. Folding@Home is still ramping up the amount of work they can assign to the network. If you don’t feel like this problem has improved much in recent weeks, it’s because the new machines have been signing on as fast as additional work can be farmed out.
On April 13, the team announced that Folding@Home would now test simulations of how existing antiviral drugs like Remdesivir may interfere with the normal replication of COVID-19 within the body. The goal for the project is to evaluate all of the FDA-approved NTP (Nucleotide TriPhosphate) drugs currently on the market to see which may be the most effective at killing coronavirus. The project is also continually working to distribute more work units to all available systems, so if you’ve noticed your system being idle lately, you’ve got two choices: Wait for some of the newer work units coming down the line, or move to a project like Rosetta@Home, which is generally smaller and could also probably use some help.
I’m genuinely curious to see what comes of all the computer time dedicated to solving COVID-19. We’ve never spent this much horsepower attacking a single virus in such a focused way, and what we learn here might be vital to long-term viral research on this and other projects. Never in human history — not with the human genome, or AIDS, or SARS, or even cancer research — has so much horsepower been so unilaterally focused on a single goal. I don’t know if it’ll pay dividends, but if it does, we’ll have learned something about just how much compute horsepower it takes to drive these kinds of breakthroughs.