Thanks to an influx of volunteers donating computer power to support COVID-19 research, the Folding@home project is now, in effect, the world’s most powerful supercomputer. But how much computer power is necessary to make a difference? After comparing notes with a couple of others, I think I have some answers.
The ideal donor
The ideal donor is someone with a modern gaming rig, with a high-end CPU and GPU, a couple of years old or newer. A machine like that can chew through tens of thousands of credits in a single day, just from you letting it run while you’re asleep and working.
That’s where the heavy lifting comes from. But I think it’s safe to say most of us don’t have more than one system like that to contribute. Some may not have one at all.
Foldingathome with a modest PC
I keep my PCs a very long time, and in recent years, I’ve started just buying high-quality office PCs used and using them until their wheels fall off. Every PC in my house is modest. For some of them, calling them modest is probably a compliment.
Putting a modest PC to work may not do much to help you climb the leaderboard, but that doesn’t mean there isn’t work for it to do.
My 4-core AMD Phenom CPU can chew up 9-10 work units on a good day, if they’re small. That’s just through CPU power. That’s not bad for an outmoded machine I built about 9 years ago. An Intel-based machine of comparable age ought to be able to do about as well.
I have a Celeron G1610 laying around. In 2014 I needed a machine for a project and that was the cheapest thing I could find. That CPU can get through 1-4 work units in a day, depending on their size.
An ancient Intel Core 2 Duo E8400 from 2008 can get through 1-2 work units in a day. That’s getting to the point where I wonder if it’s worth it.
After Jason Bucata, one of my Twitter followers, mentioned that GPU power is more valuable than CPU power, I bought an Nvidia GTX 1050 and slapped it in the Celeron. It did 10,000 credits less than 24 hours, but then ran out of work. My CPUs haven’t been sitting idle. CPU work is more plentiful.
Is it worth running Foldingathome on older PCs?
I can get high quality third-generation Intel i5 desktops for under $100 on Ebay these days. And one of those would absolutely thrash everything I’ve got, including my AMD Phenom. Why not buy one of those, then transfer the guts from the old junkers into that?
The thought crossed my mind. I opted for an easier upgrade, swapping the Celeron chip out for an i5. It cost $25. It should be a nice boost once it gets here.
But I already have these other computers. Their contributions look quaint, but they’re making those contributions right now. An i5 can’t do anything in the mail.
If you want to load Foldingathome on your home PC and let it use your cycles when you’re not using it, that makes a difference. The newer the computer is, the better it will do. But if you’ve got anything better than a Core 2 Duo in the garage or basement and it has an operating system on it, go ahead and load the client on it. They’ll put it to use.
Here’s one more thing to consider. COVID-19 is getting priority, but other diseases are still getting research. You could put a Core 2 Duo or similar machine on cancer research, while the big rigs are working on the more immediate need. That way other medical research doesn’t have to grind to a halt.
Optimizing an older PC for Foldingathome
Sometimes Foldingathome will recognize your older GPU, but many older GPUs can’t do modern folding work because they can’t do math at a high enough precision. Having the GPU enabled ties up a CPU core even if you’re not using it. So if your older PC has a GPU and Foldingathome recognizes it but won’t use it, you’re better off disabling it.
So if you go into Configure > Slots and click on the GPU slot and dlick remove, you free that CPU core so Foldingathome can run on all available cores. When you only have two cores, and they’re fairly old, being able to use both of them is the key to being able to get 1-2 work units done in a day.
That said, if Foldingathome is recognizing your GPU, and GPU work is available, your GPU will outperform that CPU core. If you can use a GPU, go ahead.