I am currently running 5 programs. 4 of them are GPU only and the 5th is CPU only. The problem I have is that the CPU only program doesn't run on the CPU when one of the other 4 applications is running on the GPU. I have tried setting the resource share of the CPU project to 100% and the other projects to 1% and that has not helped. I tried setting the "switch between applications" to 1 minute, figuring that the CPU would run once every 5 minutes for a single minute.....nope....nothing ran on the CPU all the way up to the 15 minute point. I tried setting the "switch between applications" setting to 1,000,000 minutes figuring it would run the CPU based on not meeting the deadline date.....nope.....nothing. So, how can I force the CPU to run 100% of the time on the only CPU project I have while running the GPU 100% of the time but switching between applications?

Mike

How many CPUs do you have?
How many GPUs?
Which projects are you attached to and do you have work from?
Which BOINC version?
Which OS?

---

I have 8 threads available on the CPU for the Aqua D-wave multi-threaded app. I have 2 Nvidia GTX 480's available for the GPUs. The GPUs ALWAYS run no matter what settings I use for BOINC. They switch the order of running based on the "switch between application" settings but they always work.

BOINC version 6.10.18

Vista Ultimate 64-Bit

edit: I want to add than I have no CPU time restrictions and it is set to run while the computer is in use. I also didn't have this problem until AFTER I set the other apps to only give me GPU work units. When I had CPU workunits from the other apps along with the GPU work units, the CPU work units cycled correctly. Now that there are only CPU work units from a single app, the CPU doesn't run them.

BOINC v6.10.18 is a bit elderly for AQUA - there was an AQUA-specific bug-fix in v6.10.19 - and another in v6.10.57, one beyond the currently-recommended v6.10.56

Not worth hunting out .57 - different problem fixed - but .56 would be worth installing.

There may be an issue here with the somewhat extreme combination of 2 x GTX 480, AQUA, and Vista - I doubt that the developers can afford such a rig to test on.

Depending what project you run on the Fermis (you didn't answer that question): I'm seeing that some projects declare a high CPU requirement for Fermi support, especially with Vista/Win7. If the project application configuration ends up deciding that a GPU job requires more than 0.5 CPUs, then two of them are going to take you above 1 CPU reserved. Under those conditions, BOINC would run 7 single-threaded CPU jobs: I don't know how it would cope with an 8-CPU call from AQUA.

I think the best way to proceed would be to start with an upgrade to v6.10.56, and observe how that behaves. If there's still a problem, I think you've just volunteered yourself as guinea-pig in chief: we may have to call on you to enable some debug logging, and post the resulting messages, to aid diagnosis. Is that OK? But do the simple upgrade before anything else.

OK. I give the upgrade a try first and see if that solves it. For the record, the Fermi apps I'm running are Seti@home, Seti@home beta, Collatz conjecture, and GPU Grid. Although, GPU Grid is currently dry so you can count that one out of this scenario.

Good news! 6.10.57 solved the problem. Hopefully it continues to work and isn't unstable as stated it might be. Thanks for the help. I didn't know before that BOINC updates are sometimes project-specific. I always thought it was just a scheduling framework and the apps had nothing to do with it.

OK. I give the upgrade a try first and see if that solves it. For the record, the Fermi apps I'm running are Seti@home, Seti@home beta, Collatz conjecture, and GPU Grid. Although, GPU Grid is currently dry so you can count that one out of this scenario.

SETI (and of course SETI Beta) was the high-demand project I was thinking of. In the 'status' column of BOINC Manager (Advanced view, Tasks tab), for a GPU task it should show "Running - x.xx CPUs and 1.00 GPU", or words to that effect. What's the x.xx?

(may have to scroll / expand column to read it)

SETI (and of course SETI Beta) was the high-demand project I was thinking of. In the 'status' column of BOINC Manager (Advanced view, Tasks tab), for a GPU task it should show "Running - x.xx CPUs and 1.00 GPU", or words to that effect. What's the x.xx?

(may have to scroll / expand column to read it)

The x.xx is the percentage of a thread that is required to convert/send/receive data to and from the GPU for processing.

SETI (and of course SETI Beta) was the high-demand project I was thinking of. In the 'status' column of BOINC Manager (Advanced view, Tasks tab), for a GPU task it should show "Running - x.xx CPUs and 1.00 GPU", or words to that effect. What's the x.xx?

(may have to scroll / expand column to read it)

The x.xx is the percentage of a thread that is required to convert/send/receive data to and from the GPU for processing.

Sorry, inadequately phrased question. I know what it's for: I was asking what value the BOINC server had calculated as appropriate on your particular system (it's not a fixed value, but calculated dynamically from, e.g., the relative speeds reported by your host for its CPU and GPU).

Edit - and before we get into another misunderstanding, could you confirm whether you run the stock application, as issued by the SETI project from Berkeley: or do you install applications yourself, using the 'anonymous platform' mechanism and an app_info.xml file?

In the first case, the x.xx I'm after will be calculated and issued by the server: in the second, BOINC will simply read back the value you yourself have put into the file. Which is less meaningful, but easier to cope with.