Is there likely to be any mileage in asking at git-hub for BOINC not to download GPU tasks when GPU computing is suspended. Hasn't been an issue for some time because the only tasks from WCG were MCM1's. Not there are some Open Pandemics ones using the GPU is has become an issue especially as when I first noticed it, the WCG website was down meaning I couldn't change my preferences that way. I had a quick look over there and couldn't see any outstanding requests for this but it was a quick look and I could easily have missed it. Even if I have suspend GPU processing while computer is in use, it still makes it sluggish when I am typing, particularly if typing in a terminal on Manic in a VM where just tapping on a key will often produce a letter repeated three or four times.

Also the behaviour seems to suggest that suspend GPU computing when computer is in use (minutes for keyboard/mouse use set to 3 minutes) might not be working or not working as intended.

The BOINC client only does what it is told. You need to look at your settings for the project in the client, the project website and which venue the host is located to be sure you actually didn't ask for gpu work.

The BOINC client only does what it is told. You need to look at your settings for the project in the client, the project website and which venue the host is located to be sure you actually didn't ask for gpu work.

I know I asked for it. I have now been able to go into the project website to prevent it but at the time the problem surfaced their website was down! To me it would make sense to not ask for GPU work if computing on GPU is suspended via the activity tab.

I don't run Windows, but if I remember correctly, the Suspend activity for gpu work . . . . or cpu work for that matter extends for only 60 minutes. Then reverts to active again.

The only way to make that permanent on a project is to suspend the project or suspend currently cached gpu tasks for the project and set NNT.

That is my understanding of how it currently works. Didn't know about the suspend only working for 60 minutes - will check that. I don't do the Windows bit either. this is with 7.25.0 on Ubuntu 23.04.

I think it's the 'snooze' action from right-clicking the system tray icon that's limited to 60 minutes, and I think that the 'suspend' action from the activity menu in the full display is permanent until consciously changed.

That's what I assumed the OP was posting about. I always tend to assume most questions come from Windows users.

Using the Activity menu in the Manager should be permanent as Richard states.

I don't run Windows, but if I remember correctly, the Suspend activity for gpu work . . . . or cpu work for that matter extends for only 60 minutes. Then reverts to active again.

The only way to make that permanent on a project is to suspend the project or suspend currently cached gpu tasks for the project and set NNT.

Snooze stops crunching only for 60 minutes. If you change crunching to not follow preferences but to suspend GPU/CPU crunching that suspension is until you change it back to based on preferences or always.
Suspending any task for a project disables the work requests for that project. You don't need to set NNT for that project.

Suspending any task for a project disables the work requests for that project. You don't need to set NNT for that project.

The problem with that method is that suspending a task suspends all work requests for that project - it's not just work for the matching device. I think Dave was trying to suspend GPU requests, while still allowing CPU requests - particularly relevant for a multi-application project like WCG, which offers both types of task. Differential NNT is only available via a project website - when accessible.

Edit - the only local solution for this problem I can think of is to use a 'device exclusion' option in cc_config.xml while the project website is broken.

<exclude_gpu>
Don't use the given GPU for the given project. <device_num> specifies the number of the GPU to exclude (0..63). If not given, exclude all GPUs of the given type. <type> is required if your computer has more than one type of GPU; otherwise it can be omitted. <app> specifies the short name of an application (i.e. the <name> element within the <app> element in client_state.xml). If specified, only tasks for that app are excluded. You may include multiple <exclude_gpu> elements. If you change GPU exclusions, you must restart the BOINC client for these changes to take effect. If you want to exclude the GPU use for all projects, look at the <ignore_ati_dev>, <ignore_nvidia_dev> and <ignore_intel_dev> options further down. Requires a client restart.

<exclude_gpu>
   <url>project_URL</url>
   [<device_num>N</device_num>]
   [<type>NVIDIA|ATI|intel_gpu</type>]
   [<app>appname</app>]
</exclude_gpu>

Thanks Richard for that summary and how to stop GPU requests when website not available. I still think it would make sense of suspending GPU crunching from the activity tab to stop requests for new GPU work.

Maybe if your cache was full of GPU tasks, but not when there's still tasks missing according to your work request. A suspended GPU only tells BOINC that there is a reasonable chance it will get resumed at a later time, and at that time, BOINC will want to have enough work in cache for it. Only by telling BOINC that you do not want to use the GPU, do you tell it not to download work for it.