My 560 Ti it appears have massive difficulties handling the lightest task (youtube, large images etc), so even very light surfing with the GPUGRID task running is like walking in mud, in slow motion. The AMD 6950 had no such problems when running milkyway@home (could even have CIV5 in 1200x1000 windowed mode with 15-ish fps in-game and no slowing when moving between tasks or surfing, running HD video etc), is this a common problem with low-ish end GPUs or does it affect all nvidia GPUs? If I run a 690 on my main rig will it still be so impossible to do anything else while GPUGRID is running?

Also, is it simple to run a 690/680 in the first PCI-E slot and another nvidia card in the second (perhaps a low-end card), and then only connect the low-end card to the screen? Would free up many PCI-E slots, I could run low-end or AMD card in second slot (which is only x4 in my server) and high-end cards in the first PCI-E slots, perhaps with "use all GPUs" turned off so only the high-end card in PCI-E 0 slot runs GPUGRID.

To answer a few of your questions:

I crunch with a GTX 670 and regularly play graphically lower-end games like Left For Dead, Payday: The Heist and similar stuff with around 30-ish fps while crunching, however, games like BF3 and other newer titles are unplayable at high settings. So, I would infer that it depends on your card as to how well you can play games while crunching. I like to think that crunching takes a percentage of a card's "work ability", for lack of better term, and that taking, for example, 95% of a GTX 690 away for crunching and 95% of a GTX 460 would leave you with more work ability left in the 690 than the 460, which would allow better gameplay.


Sorry for the run on sentence and bad explanation, but basically 5% of a 690 would give more fps in a game than 5% of an older card.

I know what you mean, and I am sorry but I forgot to mention "load" is 99%, on the WUs that only take 38-40% I can do whatever I want, but the AMD ran at 98% (I assumed it was a software peculiarity of saying 100%?), so does it mean the 560 Ti is a rare model that uses most of its actual ability in GPUGRID and that most others do not run at 99%?

My GPU load is around 90-95% on average, with the exception of those low-usage tasks.

I really doubt the 560 TI is a "rare model" as you described, and I'm pretty sure all the cards have varying GPU load while crunching, however I can't tell you for sure, as I have not crunched with anything other than the 670.

I'm pretty sure the 560 TI is not an exception, though. Hopefully someone can post stories of their cards, so we can get more insight.

Whether the system stays responsive under GPU load entirely depends on the software, that's why you've seen such differences between nVidia and AMD on different projects. One could easily configure Milkyway so that an AMD running it becomes unresponsive as well. Making GPU-Grid more responsive is not as easy, though.

You might want to try option out of the long-run tasks and see whether the shorter ones fare better (smaller molecules to simulate -> can divide the work into smaller slices, making it easier to squeeze the ocassional screen refresh between the crunched numbers).

MrS

---

Scanning for our furry friends since Jan 2002