I wish the Active Projects page show supported compute architectures CPU, GPU or both

[danwat1234]

https://apps.foldingathome.org/psummary
The reason why I wish this page would show the architecture compatibility of each project is, is to see if there are any projects in which CPU cores can be used efficiently i.e. no GPU support? Whereas I believe many projects support both processor architectures and so spare CPU cycles are computing way less efficiently (more units of electricity) than if a GPU was doing the same work. And there is probably a surplus of CPU compute available on the network that CPU-only projects already have plenty of power so the rest goes to 'busy work adjacent' in which a GPU could do instead.

Also, if there are projects that are CPU-only, is there a way to just allow those projects on the machine? There are allowable project categories such as Cancer but not specific project IDs, on the GUI at least.
It would be nice if supported architectures on the Active Projects page would be in the table and able to enable only specific project IDs on the machine. Has someone maintained a list?

BOINC projects including 'umbrella' type such as WCG it is typically easy to see which architectures are supported on each and each sub project, why not Folding?

[bikeaddict]

The Core column of that page shows whether the project runs on CPU or GPU. Cores 0x22 and 0x23 are GPU. Cores 0xa8 and 0xa9 are CPU. There is another page that shows how many available WUs there for each core.

https://apps.foldingathome.org/serverstats

FAHControl can be used to configure a client to run slots for only GPU, only CPU or for both. There are some projects that aren't able to run on GPU due to limitations of the processing required.

[danwat1234]

Thank you! That tells me that zero projects can take advantage of both GPUs and CPUs besides a CPU core needed to feed the GPU? So, no CPU work is enabled for GPU so it is all efficient?
Is it allowed to list multiple core types in the table in case a project supports both?
Edit: It appears multiple projects have the same description but I do not see any with the same contact with different compatible cores which could be a way to list the same project being compatible with multiple cores.

[calxalot]

There are currently no hybrid cores that support a WU using CPU+GPU.

Using a CPU thread to feed a GPU core doesn't count, of course.

[danwat1234]

Thanks me neither i meant a project that supports formatting a work unit for crunching with either a C or a G but not simultaneously. Ok well that answers my question then. CPU cores despite their low PPD are doing necessary work that is not coded to work with GPUs, sweet.

[calxalot]

The latest GROMACS can run in hybrid mode, but is still in testing.
My understanding is that performance can vary wildly.

[danwat1234]

If I could expand the thread a bit...
If CPU work is necessary, then why is PPID so low? I know the flops is low. Obviously the system works but I just think it could descentivize people from using their CPU cores. They could configure their clients to use less cores for some reason or another, or probably more often, not set up a CPU-only box for folding because the low credit makes it seem as though the work is not important although I see over a million CPU jobs in the queue.
Perhaps there has been times when CPU work does pile up but not enough so that you need to alter the credit system. It regulates itself somehow. Edit;
Thoughts; Both in terms of volunteers not modifying default configuration of client, and systems having a good GPU tend to have a decent CPU, but curiously, scientists haven't submitted projects with ginormous CPU work unit quantities. One could come along and then with current credit system it would sort of be second priority to the GPU projects. Unless GPU demand goes up so more computers go on the network, CPU power will not go up for this new CPU project

[danwat1234]

This is a continuing problem. https://apps.foldingathome.org/serverstats . ~40K total GPU work units available whereas 0xa8 (CPU) work units is 1.2 million with low hourly assignment. It seem to peg there every time more are submitted which may mean there are even more researchers want to submit.
The credit disparity between GPU work and CPU work may be limiting research. Next year I may help with ~24 5950x boxes but it will takes some time Doordashing that money.
Sidenote, I wish Folding would support work unit caching like BOINC so work can continue in case of sporadic WIFI. Some research projects could be compatible?