Got these new and cheap, and ran well for ~6 weeks. Crysis 2, SC2, everything great. One morning, woke up and saw that every time the wallpaper switched, the driver crashed. Also couldn't open NVIDIA control panel with a third monitor plugged in. Opens with two, but not three (even when hot-swapped), and it did not matter which card the third was plugged into. Upgraded to beta drivers (the control panel issue being mentioned in this beta release), problem solved.
Problem resurfaced 2 weeks later, this time no control panel problem, but relentless driver crashing with wallpaper switching, especially bad if the third monitor was plugged in, and SLI bridge in place. Taking the third away and removing the bridge helped a lot, but it was still crashing occasionally. At its worst, I would even get win3.1-ish green coloring in some spots on screen, the windows bars especially. I upgraded to newest official, and again, problem solved.
But now it's back. What's dumb though is I can play games fine like Crysis 2 and SC2 fine, but I know if I leave it alone too long with the wallpaper set at 10 second intervals, it'll eventually starts crashing again. The only reason it isn't right now is my desktop was off for about 10 days because of weather issues. I'm about to fire sale these things on ebay and get some real 460's.
Previous EVGA 8800GT's in SLI ran well, and the PSU I believe is around 850-950W. Can't get the exact model number on it because of how my case is, but it's one of those longer than usual Ultra X3's, except back when they still used sheathed cable. Got it about 3 years ago. And as I type, the idle temps are 32C and 38C.
Anyone have any suggestions? I don't want to just band-aid this by having static wallpaper.
Well that PSU is old and they do play up especially as time goes on. You cant write the PSU out of the picture just yet I am afraid. Get it tested. Better to find out if there any dramas especially as other components can be damaged.
I think I have ruled out the PSU. I checked voltages using a multimeter both when it was working fine and just now when it started acting up again, the 12v and 5v rails were both times 12.1 and 5.1.
At this point I think my "second" card has something wrong with it's SLI capabilities.
As stated I have 3 displays. The leftmost is connected to the card in question, the other two to the one I think is fine. When I run furmark to stress test them, I get flawless results on the two "good" displays. Maxes the card out from what EVGA precision tells me. When I put the furmark on the left screen, the other GPU seems to struggle to help, only getting to about 12-20% usage, the other GPU still maxing at 100%.
When I split the furmark image over the "bad" screen and one of the good ones, the image on the left side struggles, a lot. Jitters like crazy, and any windows programs on that screen also start going nuts, while on the other screen everything looks fine. EVGA precision shows that what little help the "bad" card offered when the image was displayed on that screen by itself before is gone entirely, with the GPU at 0-2% usage.
As usual, pulling the display of the "bad" card causes any flickering/driver crashing to cease. Was playing SC2 and I started to get jittery, but immediately pulled the third monitor mid-match and everything was fine from then on. Are there any better ways to test if this is an SLI problem? I think I'm ready to call this a lost cause and go with "it's a bad card"...
Once you pull assorted Windows to the top of the game, is is not running the game effectively in Windowed mode, i.e., not really truly in SLI mode at all? (I thought I read that, and it would explain your low GPU useage....)
This does not rule out one of your cards having 2D Windows issues, however...if you swap the cards around, does even the jittering switch to the other side? If so, one of the cards would indeed seem to have issues...