Recently upgraded from GTX570 SLI to GTX970 SLI and a 34" 3440x1440 LG screen - which is nice. Spec as per sig. However, I've been plagued by crashes since the upgrade - so, what I've done so far --> - Made sure everything is powered, connected and seated correctly - Changed the power cable connectors - PSU has 4 x 6+2 pin - Returned the system to non-overclocked - Used MSI Afterburner to return the two (factory overclocked) cards to standard nvidia specs - Swapped the cards around - Swapped the SLI connector - Wiped the previous drivers, and installed latest - and after this didn't work .... - Complete re-install of Windows Issue manifests as - Random freeze crashes after between 20mins or up to 3hrs system on time. No identifiable trigger. When the screen comes back out of power saving mode, or screen saver, there are graphics errors - little flecks all over the screen. The flecks can be resolved by changing the resolution, and then back again - but I lose perfect graphics, every supposedly sharp line is misaligned - difficult to describe, but it's not "right". Interestingly, most games play OK - but WOWP crashes the system completely. Additionally interestingly, when I run either of just one card - WOWP black screens, but the system doesn't crash completely. All that said, the system still crashes over time - so it's not just a WOWP issue. I'm currently running just one card to see if I can emulate the crash, and if not, will swap to the other and try that one - my suspicion is that there's some some of memory error on one of the cards - but I'm not sure how to test for that. Any pearls of wisdom chaps?
You're doing what I would have suggested by running one card at a time to see if it crashes.. One things for sure one of them isn't right. One thing I would do though is return both to the default factory overclock. Then test them.
I haven't had a need to test VRAM so i can't vouch for how useful or effective the following programs are, but they may precluded the need to take a wait and see approach. CUDA GPU memtest And Video Memory stress Test
I was looking into that monitor and apparently there are issues with Nvidia and Sleep on this screen, so best to disable that power saving etc.
Done that, no issues reported Done that, nice to know it's the cause of the glitches, but it's not the cause of the crashes Ok - will do - thanks
Nothing particularly helpful Event 41 : Kernel-Power "The system has rebooted without cleanly shutting down first. This error could be caused if the system stopped responding, crashed, or lost power unexpectedly." Maybe yet another failing PSU - I'll hunt out the trusty (and well used) PSU tester
It crashes on both cards when they're used on their own, sooooo - it might not be the cards ..... or it might be both, but that's unlikely. Going try it with the old GTX570. Sent from my iPhone using Tapatalk
Ok, hooked up the PSU tester, and my 12V supply to the main board is goosed, and this from a less than 2year old Antec 1200 High Current - bet I can't find the sodding warranty card anywhere either. Sent from my iPhone using Tapatalk
Sounds like you have found the problem, the chances of it being both card would be fairly slim. That said I have had 3 dodgy mobos one after the other all of which had the same problem. Dodgy batch at it turned out.
Well, further testing (let's be truthful here - correct use of the tester, and plugging in the separate 4pin 12V cable) has revealed that it's NOT the fault of the PSU - so the issue remains. Sent from my iPhone using Tapatalk
Right, apparently the new (post 900 series) drivers under volt the second card when operated in SLI - this is apparently by design, NVIDIA are not offering a fix for it. Importantly, this change has been rolled back to cover the 700 series as well, and so if you've got 700s in SLI - and have been experiencing issues, roll your drivers back. Frustrating, because now I've to to learn how to extract, adapt and flash the graphics BIOS in order to fix the built in undervolt. Sent from my iPhone using Tapatalk
Update - I've used nvmonitor to overvolt and overclock card 2 (to maintain the overvolt, and stop it reverting back to idle voltage) and set a task to run at startup. Both cards now permanently at 1v (or within 0.05v) - and are matched - previously they were 1v and 0.875v (idle voltage) even when both were under load. Going to run Furmark all day today. Sent from my iPhone using Tapatalk
I've used Afterburner, but it's difficult to isolate an overvolt to just one card. You also can't overwrite the programming which apparently deliberately imbalances the voltage supply. Not yet had chance to properly flash the BIOS's - probably have to wait till the weekend. Sent from my iPhone using Tapatalk