GTX 1070 driver crashes when idle

[Mightyena]

Hi guys,

I have an EVGA GTX 1070 SC that's about a year old now, and it's displaying some odd problems.
Basically I will get intermittent driver crashes, but only when the GPU is idling. Under load/in games it is rock solid....

What happens:
When idling (so far its happened when on the desktop and browsing on Firefox), the screen will go black for a few seconds, and then return with the message

Display driver NVIDIA Windows Kernel Mode Driver, Version 388.43 stopped working and has successfully recovered

This happens over several driver versions. I've tried 388.43, 376.33 and 387.92

Happens every so often, as shown by this error log. I can't figure out what's causing it, or if its a hardware problem...


What I've tried:
I've tried removing drivers with DDU, and reinstalling the three versions above. This seems to stop it for a couple of weeks, but then it comes back (of course the problem is pretty intermittent so I could be wrong there).
Also tried reseating GPU. Haven't tried it in a different system, since it can sometimes be weeks between crashes, but tried my old GTX 680 in this PC, and there were no crashes for the 3 days I had it in there.



My system:
Intel Core i5-4690K (originally overclocked to 4GHz, but the problem persists at stock)
ASUS Z97-AR
16GB (2x8GB) G.Skill Basic 1600MHz RAM (running at 1800MHz, but problem persists at stock)
Seasonic M12II SS-620GM
Crucial MX300 SSD
WD Blue 3TB HDD
EVGA GTX 1070 SC (no OC)
Windows 7 Professional 64-bit

Does anyone have a fix for this?

Thanks

 

[panathas]

It could also be the PSU or the motherboard (PCIe slot). Try to borrow another PSU from a friend and use the 2nd PCIe slot from your board. If nothing works then your GPU probably has some damaged components and you need to RMA it. It seems to have voltage problems in idle situations which may be caused by a failing component on the VRM of your card which causes all those stability issues.

EDIT. You said you have that GPU for a year. Was it working perfectly before or did it had issues from the beginning?

 

[Mightyena]

Hi, so sorry I haven't replied, I missed the notification. I can try my old 680 in there, but cause of how infrequently the crashes happen, I'd have to have it in for a couple of months to tell anything. I've got another 550W PSU around somewhere too, so I'll try that as well.

Yeah I've had it for around a year. Bought the card in January last year, and it ran absolutely fine until about October, then the crashes started. No hardware changes have been made to the system since it was built

 

[panathas]

Testing with your old GPU may not help you because what you are facing may be software or hardware issue. The 680 is a completely different card so it may not help you. Only if you managed to get your hands on another 1070 you may be able to find something more meaningful. On the other hand if the 680 caused the same issue you'd know that something else is causing this, but if you can't reproduce it, it doesn't necessarily mean that the 1070 is faulty.

Unfortunately this issue is rare and on top of that it only occurs during idle times so troubleshooting it takes a lot of time. If it was during load, you might be able to fix it by lowering the clocks or upping the voltage a little bit but during idle the GPU behaves more automatically and you have little control over it.

I think that you best bet is to try to test it with another PSU. Good luck.

 

[Mightyena]

Short of RMAing it and hoping EVGA will actually replace it, getting another 1070 to test with will be pretty difficult.

Thanks, I'll have a go with the other PSU and see if anything changes. If not then as long as it only happens very infrequently and isn't actually damaging anything I think my best option is pretty much to put up with it...

 

[panathas]

You could try to RMA it but since this is a rare issue they may not able to replicate it and they will probably return it to you, saying that it's fine. That said you have nothing to lose. Send it back and hope for the best, you never know, their tests are more thorough and you may get lucky after all. Don't just keep a card that may be faulty. The crashes may get more frequent and by the time you decide to RAM it, the card may already be out of warranty.

However before you go that route try to test it with another PSU. Also if you try to remember all the times that this issue occurred you may be to figure out a pattern which will make all your troubleshooting efforts much easier. One last thing you should do, is download GPU-Z and monitor your GPU with it. You may be able to spot one or two abnormalities which will help you understand your GPU's behaviour better. Just search online for your GPU's specs in order to have something to compare to. Good luck.