If you were following the greatest BOINC team site in the world, which everyone should be ;) you'd know that all you need to do is set exclusions in cc_config for FahCore_21.exe and 22. This works for GPU, which is all that I'm doing.

	<exclusive_app>FahCore_21.exe</exclusive_app>
	<exclusive_app>FahCore_22.exe</exclusive_app>

Set your preferred project to a resource share of zero and you're good to go.

Thanks, but it also suspends all BOINC CPU work. That is too much for me, but it is a clever idea.
Also, for Linux it is just FahCore_21 and FahCore_22, without the ".exe".

Interesting idea about the Exclusive App!

I'm pretty sure I'd want to use the GPU version though.

https://boinc.berkeley.edu/wiki/Client_configuration

<exclusive_gpu_app>important.exe</exclusive_gpu_app>

I'll think about setting this up, but don't know if it'd work the way I want it to, on my PCs that have 3 GPUs.

Interesting idea about the Exclusive App!

I'm pretty sure I'd want to use the GPU version though.

https://boinc.berkeley.edu/wiki/Client_configuration

<exclusive_gpu_app>important.exe</exclusive_gpu_app>

I'll think about setting this up, but don't know if it'd work the way I want it to, on my PCs that have 3 GPUs.

Thanks for pointing this parameter out, I never noticed it. I've been running separate clients for CPU and GPU for years, and rarely run projects that have GPU work on the CPU, so what I posted works perfectly for me. Actually I run separate clients for every GPU as well so I have better control over what's running. I tend to forget most folks' setups aren't like mine. This <exclusive_gpu_app> is a nice option for mixed work.

---

Team USA forum | Team USA page
Join us and #crunchforcures. We are now also folding:join team ID 236370!

One possible solution:

This is a low level heavily scripted option but might make a useful project if you have time or the inclination.

Step 1
Use nvidia-smi to check for low GPU temperature....eg less than 12W. This will indicate no activity in Folding@home.(nvidia-smi can output selected figures such as power, I have used this in the past)

Step 2
On low temp use boinccmd to set gpugrid as active and retrieve a wu.

Step 3
pause 300 seconds, then use boinccmd to set gpugrid status: fetch no more tasks

Step 4
use either boinccmd or nvidia-smi to check for activity in gpugrid. When gpugrid task complete, fetch folding@home task. Havent scripted folding@home so cant comment on commandline control of fetching tasks there.(this could render this script pointless)

Step 5
repeat

I can submit a script when free in a day or so if needed.

It looked to me like FAH is accessed via telnet, which might be a pain. Haven't fully researched it all. Also, I'm attached to all 55 projects, with GPU work allowed for about 5 some of which also have CPU work I want, so I'm not interested in suspending projects :)

I am planning SARS-CoV-2 work.

When will this work be available?

I just got a long run marked PABLO. Is this COVID-19 research?

I just got a long run marked PABLO. Is this COVID-19 research?

Never mind- it errored out after 4 seconds.

I just got a long run marked PABLO. Is this COVID-19 research?

Never mind- it errored out after 4 seconds.

Probably a virus...

I just got a long run marked PABLO. Is this COVID-19 research?

Never mind- it errored out after 4 seconds.

Probably a virus...

No doubt the OC@2GHz virus. 🐞💻🐞

The real conundrum is the preparedness of the public health services and/or governments which is put to the test by this world wide pandemic.

Here, in Italy, the situation is dramatic

Here, in Italy, the situation is dramatic

You have probably heard of remdesivir.
https://www.npr.org/sections/health-shots/2020/03/21/819099156/might-the-experimental-drug-remdesivir-work-against-covid-19

It was used on an elderly patient in Italy with apparent success, though the trials are ongoing.

I just got a long run marked PABLO. Is this COVID-19 research?

Never mind- it errored out after 4 seconds.

Interesting, 64 tasks released without Wrapper (using ACEMD2 v9.23 Windows). Did someone forget to use ACEMD3?

Interesting, 64 tasks released without Wrapper (using ACEMD2 v9.23 Windows). Did someone forget to use ACEMD3?

Please tell this old newbie how you knew that, I'd be much obliged.

Interesting, 64 tasks released without Wrapper (using ACEMD2 v9.23 Windows). Did someone forget to use ACEMD3?

Please tell this old newbie how you knew that, I'd be much obliged.

Last year in April, GPUgrid embarked on an upgrade of their WUs so they would be compatible with CUDA 10 and Turing based GPUs. The new design of the WUs used "Wrappers" to make the WUs more portable and easier to create for the Project team.
The last of the old WUs issued by GPUgrid were ACEMD2 with a version number of v9.xx
The release of the new WUs was in October last year. The new WUs have been referred to as ACEMD3 with a version number of 2.xx

For your reference, the list of GPUgrid apps are found here:
http://www.gpugrid.net/apps.php

If you check the Server Status page you can see "Short ..." and "Long ..." WUs. These are the legacy ACEMD2 WUs that wont run on Turing based GPUs and havent been issued since October last year (Windows only).

We are currently running WUs listed as "New Version of ACEMD".

At the bottom of the Server Status page you can see a detailed list of the ACEMD2 WUs which lists the current batch of "old" WUs.

All have failed.

Thanks very much Rod4X4! I seem to have a lot of catching up to do on history here. I really appreciate the mentoring.

Thanks very much Rod4X4! I seem to have a lot of catching up to do on history here. I really appreciate the mentoring.

Your welcome

Is the FAHClient_22 use CUDA?
I coulndn't find a clear answer for that on their forum. (It is planned as far I could find anything about it.)
I get power consumption with that app as much I get with acemd3.
I've never experienced that an OpenCL app could draw as much power as a CUDA app does.
Is there a way to force work assignment besides restarting the FAHClient service? (like pushing the "update" button in BOINC manager.)
It is very hard to get any work from FAH...

Good questions! I don't have the answers though, sorry.

Regarding getting work in F@H... Whenever it gets stuck and isn't getting work, I have been using FAHControl (the Advanced Control program), to tell the PC or slot(s) to Pause, then tell the PC or slot(s) to Fold. I think this resets backoff timers, and might even allow it to hit different servers when trying to find work.

Regarding keeping the GPUs busy when F@H it out of work... I have setup my PCs with a BOINC GPU Exclusion for FahCore_22.exe .... so that, if none of the GPUs in the PC are working on F@H GPU tasks, it will at least allow BOINC to put those GPUs to work on BOINC projects.

.. in cc_config.xml ..

<exclusive_gpu_app>FahCore_22.exe</exclusive_gpu_app>

I'm not an expert, and I don't even know if this is the best approach towards keeping my PC GPUs busy (especially when 2 of my PCs have 3 GPUs in them, and I want all 3 to be crunching!)

Is the FAHClient_22 use CUDA?

No, it doesn't yet. But they plan for it.

And curious you should ask. I am in a rather fraught discussion of that now.
Maybe you can enlighten some of them on the benefits better than I can.
https://foldingforum.org/viewtopic.php?f=17&t=34173&p=324695#p324695