Yes, there is a problem. I have BOINC setup to use my CPUs even though I am at the computer also using them, and there is no interference with my usage. I also have BOINC setup to use my GPUs even though I am at the computer also using them, and there is no interference with my usage while processing Poem@Home tasks or GPUGrid.net tasks. But there is a huge interference, in the form of massive display lag, while processing WCG HCC tasks, for some users, and there shouldn't be. The GPU app is supposed to share the GPU with you, when you are active, just as the CPU app shares the CPU with you, when you are active.


%GPU use has nothing to do with the problem. Other project GPU tasks use upwards of 99% of the GPU, and have no problems with lag. This project uses ~98% GPU, and has major problems. Note also that it has nothing to do with CPU availability, either. I'm still convinced it has to do with priority, or frequency.

I'm available to test any possible solutions that WCG would like to try, including beta apps on a beta channel. Please fix this.

Regards,
Jacob

You are correct, the lag has nothing to do with GPU usage, even at a usage of 50% this app could reduce your framerates drastically, the problem is with a default setting in this app has been configured in a way it will make sure it will prevent your screen from refreshing often. There is a command you can insert into your app info to reduce the frequency and polling rates so that you can regain your refresh rates.

It seems the app has been configured to settings optimal for, say, a Radeon HD 7870s power level, so 7900 cards can get by with minimal refresh rate lag, anything below that will have major lag.

Then perhaps my observations are totally wrong and if so my apologies.
For myself I see no major issues so no need to continue in this thread.
Best of luck.

I think it does have something to do with the problem, but it may not be a direct correlation. From what I hear on the Folding forum, the %GPU gives an indication of use of some resource on the chip (not sure what), but what other resources are actually being used depends on the application. For example, games and videos will use different parts of the chip than distributed computing applications. And even the DC applications will vary, depending on fixed point, floating point and a lot of other factors. So what is 99% on one application may be different than what is 99% on another.

And comparing CPU and GPU percentages is really comparing apples and oranges, since the CPU can be preempted for other tasks much more easily than a GPU can. If you can change the priorities and get an improvement, that is great, but I would not be too surprised if the HCC developers have already done what they can in that regard.

I think at the user end, our real level of control is to choose either Nvidia or AMD for a given project, and ideally use two cards, one for display and one for computing. It is clear to me that AMD is better for this project, and you probably won't be changing that advantage whatever you do. (For Folding, I would give the opposite advice.)

Just wondering. Can this be an Open-cl support problem with the Nvidia cards? I think Nvidia is known to have a really bad Open-cl support compared with ATI\AMD card?