Folding Forum

I've recently taken a renewed interest in Folding due to COVID-19, as I'm sure many others have too, and noticed quickly that I wasn't able to get WUs reliably. It's a lot better than it was (thanks FaH team!) but there are of course still times when no WUs are downloaded.

I tested installing BOINC, running Rosetta@home, and Folding@home together on the same machine. Rosetta@home currently only uses the CPU, not the GPU. Rosetta@home has not had the influx that Folding@home has had, so it is currently assigning WUs reliably. This means that when Folding@home has not assigned WUs the CPU, at least, is not idle and is doing work for Rosetta@home.

I set Folding@home to run at a slightly higher priority (Configure > Advanced > Folding Core Priority > Slightly higher). When it downloads a WU for the GPU there's really no impact and it computes as fast as it did before. When it downloads a WU for the CPU it is using around 60-70% of the cores and computes slightly slower than it would otherwise, but it's still quite quick.

I also experimented with BOINC's 'suspend when CPU activity above X%' setting, but actually found this to be unreliable so just left it to compute all the time. Rosetta@home and Folding@home are sharing quite happily.

In setting it up this way, I found I am less inclined to try to force Folding@home to download WUs since the PC is now always being productive for an important cause.

P.S. Once again, a huge thank you to the team for dealing with this influx and letting us mere mortals help out, and a huge thank you to the volunteers on the forum for answering all of our questions.

If you do not get FAH work units for your GPU currently then BOINC also has GPU projects like https://www.gpugrid.net/ But they don't have COVID-19 projects as far as I know but only other medical research.

foldy wrote:If you do not get FAH work units for your GPU currently then BOINC also has GPU projects like https://www.gpugrid.net/ But they don't have COVID-19 projects as far as I know but only other medical research.

That's good to know.

Do you know if that'll play nicely with FaH if both try to do GPU WUs?

CJB1981 wrote:Do you know if that'll play nicely with FaH if both try to do GPU WUs?

I decided to test this and can confirm they do work together side-by-side, so no issues.

CJB1981 wrote:

CJB1981 wrote:Do you know if that'll play nicely with FaH if both try to do GPU WUs?

I decided to test this and can confirm they do work together side-by-side, so no issues.

I strongly suggest you do not do both at same time on GPU, if you want to do BOINC just do that but F@H depends on trajectories so you are slowing down the project unnecessarily by slowing down your work units like this which could have been done by another person. If you want do F@H on your CPU and BOINC on GPU or other way around but don't do both on the same device.

kiore wrote:If you want do F@H on your CPU and BOINC on GPU or other way around but don't do both on the same device.

I'll just do FaH then.

I do understand the frustration of having resources sitting idle, am experiencing this myself, but this is improving and will hopefully not be an issue soon.
Am not against doing BOINC stuff but splitting resources on a CPU is workable, on a GPU that just too complex and risks failure or at least slowdown of both projects.

Anyone using BOINC on a Mac?
I'd like to set it up as a fallback, but I'm having trouble finding the F@H binaries to add to the exclusion list as I'd like to give F@H priority on my resources.
FYI, I also setup GPUGRID.NET on Boinc so I've got a GPU fallback now as well.

kiore wrote:Am not against doing BOINC stuff but splitting resources on a CPU is workable, on a GPU that just too complex and risks failure or at least slowdown of both projects.

I've switched GPU to FaH only based on the previous post, so I've taken that onboard. Initially I had only split work on the CPU and that does seem to be working well.

BOINC has an option to suspend computation if another process on the computer takes up more than x% of CPU time, x is configurable.
Since the Rosetta COVID project is (currently?) only delivering CPU WUs, this could be a workaround that doesn't cost any computing time, and prioritizes F@H.

EDIT: Just tried my own suggestion - doesn't seem to work, looks like BOINC is not honouring its own setting in the latest version.

uyaem wrote:EDIT: Just tried my own suggestion - doesn't seem to work, looks like BOINC is not honouring its own setting in the latest version.

That's what I found too, which was disappointing to say the least. That's why I set FaH to run at a slightly higher priority.

You guys can always install Foldit and fold proteins yourself! It doesn't take any power away from FAH.

Just do GPUGrid. Let others do Folding. Otherwise, you are just competing for the same resource rather than adding to the total results.
Until they add more capacity, there is no point in more crunchers piling on the heap. In fact, it might tie up the servers and reduce output a little.

Is there a way to set up BOINC to run its projects when there is no FAH work?

Welcome to the F@H Forum RTS708,

Please note that F@H and BOINC are two different set of software and there's no communication between them. You have few options.

Option 1
Select either F@H or BOINC and stick with it. This is the simplest option.

Option 2
If you are using CPU and GPU, you can decided if the CPU is used for BOINC and GPU used for F@H. That way, you can run both applications but different components of your system depending on what works best. It might take a bit of time for you to get the right mix.

Option 3
Install F@H and BOINC on your system but you manually decide when to run each application depending on the situation.