Folding at home is running out of work because of the overwhelming response. I've found that sometimes restarting the FAH client/container will trigger a new job even after being stuck in a loop for hours
Rosetta@home (BOINC) still has a ton of work items and is CPU based, so it's easy to run on most anything. Some are COVID-19, some are MERS, some are things that have nothing to do with coronaviruses.
I've been running BOINC, and since none of the tasks required GPU, as you mentioned, I was able to scale to a number of devices that the client could be compiled on.
If you have 5-8 hours to devote to a round, put everything you have into it. It's quick to set up and is certainly worth the cycles.
Linus (from Linus Tech Tips) said [0] he is working with the Folding@Home people to set up an additional high-capacity server to alleviate the load troubles.
My CPU client occasionally gets new work units, but it's been idle for most of the week.
Currently, it looks like one of the main issues is not having enough server capacity. If anyone has extra resources or expertise to contribute, see https://foldingforum.org/viewtopic.php?p=314382.
I had a couple hours earlier today of CPU. Currenlty running on 1 Compute node (out of 3) (OpenCL) on my GPU. Which Windows reports as 3% of the GPU. But it's 99% of the actual Compute module.
Mac users, note that Folding@Home can't use GPUs at all. I considered Boot Camp but their docs (and Boinc's) seem focused on Nvidia and don't appear to have kept up on "ATI" (AMD) video card instructions.
Folding@Home is runnign GPU stuff on Windows without any configuration on my AMD RX 580. Although it's not using much of it. Presuably because OpenCL only runs of the Compute units which are relatively small part of the GPU.
No, all GPUs released in the last 10 years have a unified shader architecture. Vertex, fragment, and compute shaders all run on the same units, and they are the big part of the GPU.
"Compute units" is AMD terminology for a grouping of ALUs and stuff. Nvidia calls them "stream multiprocessors".
Example: Vega 64 has 64 CUs, each CU has 64 "shader processors", so there's 64*64 == 4096 "shader processors" total. This is the big part.
Yea I install the Asus GPU-Z and i was better and gave that it was using 90% of the whole GPU.
Windows just doesn't know how to count GPU usage apperntly.
Is there a way to have some more fine-grained control about how Rosetta and Folding use resources? For example, I want them not to run in sleep mode, and certainly not when the laptop's lid is closed. Or, better perhaps, to throttle above certain temperatures.
It's very configurable. I let BOINC go to town for up to an hour whenever I leave my computer alone, then the computer can go to sleep. It's not running continuously but the progress adds up steadily over time.
Would still like to have a "suspend when lid is closed" or "suspend when temperature is > 70°C" option, but I guess that would require a lot more hardware specialization than the authors of the tool could reasonably be expected to do.
I can’t understand the statistics for Folding@Home.
If there are X million PCs donating Y% of their time to the project, and they cost $Z each, I was wondering what XYZ was.
That is to say, if discovering protein structures is a critical problem needed to solve Covid19, how much would it cost to just replace this distributed computation with some dedicated machinery?
I wonder why Amazon, Rackspace, and other major hosting providers that are sitting on a ton of hardware haven't contributed resources to efforts like this.
"Rackspace is setting aside $10M in free OpenStack Public Cloud hosting resources over the next 6 months for organizations participating in COVID-19 relief efforts"
Seems like if you have a huge cluster of computing and free electricity and want to help out it might be more useful to mine bitcoins and donate them to covid-19 charities.
These computational chemistry problems make it easy-ish to eliminate candidate drugs that would be impossible in practice. Think of it as narrowing the search space, or reducing the size of the haystack rather than trying to find the cure/needle.
I once caught wind that my professor gave extra credit in distributed systems class for BOINC points. I ran this constantly for years and then he didn't even give the extra credit.
At least with folding at home they have set the covid19 research as top priority for all of the clients (regardless of the clients preference).
Note - as far as I know there are no covid19 CPU based works units, so you will continue to get normal work units if you have CPU slots available. This is worth mentioning, because by default the client adds your CPU to the available slots. On linux you have to go into advanced settings to enable the GPU at all, so it is something to watch out for.
The Folding@Home GPU client works way better on Linux with NVIDIA/CUDA than it does with AMD. I could not get a 4x Radeon VII rig with ROCm to utilise more than a single GPU at any time due to OpenCL errors. Also, the folding@home Ubuntu client has three-year-old issues with not installing properly [1] so you have to workaround that, too.
If you have the time to contribute reliability patches to the Linux clients, I'm sure it would be appreciated as well.
Do you know what CPU WUs/projects are available? I have not been assigned any WU in days so I am not sure if there are any new ones which have been released recently.
But current the covid19 page (last updated mar-10) has the following statement:
"All projects are using the new GPU-accelerated Core22 based on the open source OpenMM biomolecular simulation engine."
I love these projects and have donated time in the past to different projects. I'll be promoting this to all my friends -- there are so many ways to help and this feels like one of the easier ones.
Folding at home is apparently really hard to install and run on windows without admin permissions. Big bummer. I tried for 2 hours!
(the beefy pc is for gaming)
I have an admin, but it's not the "active" account. It's used only for installing stuff and general maintenance of your computer.
This is not only the proper way to use an admin account, but it's also recommended.
The fact that Folding@home supports only running the software when the admin account is logged in out of the box (there might be a solution to my problem, I just hadn't found it), it's disappointing and problematic.
>>The fact that Folding@home supports only running the software when the admin account is logged in out of the box (there might be a solution to my problem, I just hadn't found it), it's disappointing and problematic.
Well, no, that's not entirely true. You can run F@H as a service and then it can run without anyone logged in. The problem then is that it cannot use the GPU for work, but that's not a problem with F@H but with how Windows manages access to resources for background services without a user logged in.
Now I'm confused. Administrative privilages have nothing to do with using the GPU. A service process cannot use the GPU, regardless of who owns it, that has nothing to do with F@H, that's just how Windows is designed.
According to the organisation behind F@H , the application requires administrative priviladges to make sure you have the permissions to use the application on the machine. I guess they had enough complaints about people running it when they don't(students, employees etc), so requiring administrative priviladges is the easiest way.
Your explanation though makes sense, even though the solution to the problem it's terrible. I presumed it was for the GPU for weird reasons (bad software, CUDA, somethng like that), but what you described makes the most sense.
Apparently, I managed to get it running today with some magic using icacls and file ownership. The web control does not work, but I'm able to run the Advanced Control Panel and start everything from there.
Inexplicably F@H is not available as an _official_ docker container, and their install procedure for Linux compares unfavorably with wisdom tooth extraction.
How so? I just looked, and they have both .deb and .rpm packages available.
I once spent two-and-a-half hours having two wisdom teeth removed, and afterwards I got an infection in the empty sockets. I'd rather do all of that again, than having to use Docker for anything.
I just sent up an LXC container in 10 minutes to do GPU folding. Docker is much easier than LCX to configure, especially for GPU stuff. Really isn't that difficult. It's a single deb package.
I get that F@H does a lot of computation with my CPU/GPU if I am running the client. What do those cycles do? It tries a bunch of random stuff to calculate... what? What is it trying? It sounds a lot like Bitcoin mining, but I'm not sure what it's trying to do.
