GTX 960 keeps dying..

[KgD1001]

Hello! A few days ago I posted on few other forums about my problem with new GTX 960. http://www.tomshardware.co.uk/answers/id-2581724/gtx-960-driver-error.html
Many people suggested to RMA the card. I did that yesterday ant today they told me that they tested everything and it works just perfectly for them with no issues. So they gave me the same card back. Now I'm completely lost and have no idea what is wrong. The same problem continues... While doing nothing, everything works fine, some games also run fine. But when it comes to high load (demanding games or stress tests) the screen just turns black and comes back to life with the same error:

No such problems with my old GTX 560...
Something's wrong with my hardware? Does anybody have any ideas? Thank you in advance!

My specs:
Gigabyte GeForce GTX 960 OC
AMD FX-4350 QUAD 4.20GHz
DDR3 Kingston HyperX Beast 8GB (2x4GB) 1600MHz CL9 1.65V
ASUS M5A97 LE R2.0 Motherboard
1TB HDD
Corsair CX600 600W PSU
Windows 8.1 64-bit

 

[RaulF]

Did you try a fresh install? OS and Nvidia driver? I would start with a fresh install of the driver. Use some sort of driver cleaner after the uninstall to make sure everything got removed. Also if the card is OC from the factory i think MSI afterburner will let you down clock it, try to bring the clock to factory specs and see if it still persist.

Or maybe another card in your system if you still have one. Also if you have friend maybe try your card on his system or another computer if you have it.

Install MSI afterburner and see what your temps are like during demanding apps.

PSU seems fine.

 

[tviceman]

Reinstall the drivers, make sure you check the option to do a fresh install.

 

[Shehriazad]

Wasn't there very recently a kernel crash issue if you use Nvidia Experience?

Do you by any chance use Nvidia Experience or even update via that tool? If yes...that could be it.

I remember that not too long ago people were complaining about driver kernel crashes while using Experience. (Was it like 2 weeks ago? Not sure) If you have Experience installed, kick that off...and uninstall the driver with a tool like DDU. Then do a fresh install of either the most recent, or the driver before that.

 

[KgD1001]

Nothing really helped. As I mentioned, no problems with my old GTX 560, it works perfectly fine. I also tried my GTX 960 on my friend's PC which has the same PSU and motherboard and everything works a little bit better. For example, on my PC FurMark's stress test crashes in just 2 seconds, while on my friend's PC it can probably run forever. It only crashes when we try stress test with 8X MSAA anti-aliasing on. The same error occurs but instead of error code 3, we got error code 7. We also ran some games with no issues.

 

[Shehriazad]

Well...if it does happen on other machines...then something clearly must be wrong. Also by RMA do you mean you just brought it back to the Shop and they were like "Yea, everythings' okay"?

Because that's not proper RMA...Have you mentioned that this happens on other PCs you tested the card with as well?
If the Vendor is being useless...push for an actual RMA and get it sent back to the manufacturers company/the actual vendor...but beware...that usually takes weeks to months until you get back your (hopefully) repaired card or a refund.

 

[KgD1001]

I brought it to the vendor yesterday. Told everything about this issue and they told me that they will take a look. Today I got the card back and they told me that they tested it with FurMark and 3DMark with full load and no problems were detected. If there were any problems, the card would have been sent back to manufacturers but... They also told me that my PSU might be the main problem, but it's kinda strange because it has no issues with GTX 560. Well I guess I should try it on different PSU then.

 

[Leyawiin]

If two different places tested it and it came back O.K. its your system. I know its a p.i.t.a., but you might want to do a fresh install of Windows if wiping the old drivers (use DDU) doesn't help.

 

[Ketchup]

I never saw you take the suggestions to install the latest drivers. Use the clean install option.

 

[KgD1001]

No good signs after reinstalling OS either..

 

[Ketchup]

What drivers are you using?

 

[Mem]

Since you have OC version ,try underclocking it slightly to see if that improves stability.

I had to do that many years ago with my Nvidia OC card(7800gtx),reducing speed by 10MHz fixed my issue back then.

 

[Shehriazad]

Again: DO you have Nvidia Experience installed? If yes, get rid of it for now and try it like that...the kernel crash issues related to that were really recent....so yea.

 

[xorbe]

That's a TDR. It is caused by many things, including: defective/oc gpu, overclocked cpu, system memory problems.

 

[BSim500]

Many people suggested to RMA the card. I did that yesterday ant today they told me that they tested everything and it works just perfectly for them with no issues. So they gave me the same card back.

I installed one recently for a friend on his HTPC that had a Seasonic G360 (360 watts) PSU and it ran flawlessly even under Furmark at barely 160w full system load, so I don't think a 600w PSU will be strained unless you've got a fault somewhere. (I installed just the driver and not the Nvidia Experience though if that makes any difference). Things to try:-

- Uninstall the driver then reinstall it without Experience (as Shehriazad said)
- Set a low Power Target limit of say 50% and see if it crashes then under Furmark?
- What are the temps like? Heatsink making proper contact?
- If all else fails, can you not take the card back and insist on swapping it for a different one of the same model? (It shouldn't crash at all even with 8x MSAA Furmark on.)

 

[KgD1001]

I installed one recently for a friend on his HTPC that had a Seasonic G360 (360 watts) PSU and it ran flawlessly even under Furmark at barely 160w full system load, so I don't think a 600w PSU will be strained unless you've got a fault somewhere. (I installed just the driver and not the Nvidia Experience though if that makes any difference). Things to try:-

- Uninstall the driver then reinstall it without Experience (as Shehriazad said)
- Set a low Power Target limit of say 50% and see if it crashes then under Furmark?
- What are the temps like? Heatsink making proper contact?
- If all else fails, can you not take the card back and insist on swapping it for a different one of the same model? (It shouldn't crash at all even with 8x MSAA Furmark on.)

- I tried installing with and without experience, no changes.
- Tried with 90% power limit - Crash. 75% power limit - crash. 50% - no crashes. Same with resolutions. If it's a low resolution - no crashes, even if GPU load is huge. If resolution is higher (mine is 1680x1050) - Crash. Vendor stress tested it on 1920x1080 and everything worked fine.
- Temps are perfect.
- I can only swap cards if this one is broken. They said this one is ok, so asking them for a swap is useless :/

 

[amenx]

I'll bet its an odd incompatibility issue with your motherboard and the 9xx cards. Even the 970 on a few older Asus boards did not work well but was fixed with a bios update. Try your card in another PC, but this time NOT one with the same board as yours.

 

[BSim500]

I'll bet its an odd incompatibility issue with your motherboard and the 9xx cards. Even the 970 on a few older Asus boards did not work well but was fixed with a bios update. Try your card in another PC, but this time NOT one with the same board as yours.

^ This may be the case : "I have the same mobo and the asus stryx gtx970 is not working in my PC. Tried almost everything."
http://www.tomshardware.co.uk/answers/id-2350262/gtx-970-comptabile-asus-m5a97.html#14483632

1. Make sure your BIOS is up to date
2. Try putting the 960 into the other slot on the motherboard
3. Similar problems on some boards in the past have been cured by resetting the BIOS / taking the CMOS battery out when unplugged from mains then putting it back in again, and then reentering your settings.

Other than that, I'm out of ideas (beyond testing it on a different motherboard).

 

[KgD1001]

Thank you for your replies! I will test my card with another psu and motherboard and see how it goes.

 

[KgD1001]

It appears that everything is fine with the PSU. I tried to run my 960 on an old Codegen 500W PSU. No changes at all. Then I went to my friend who has CX600M and everything worked just perfectly. No crashes at all.
So I suspect that the main problem is my motherboard indeed. My other friend with the same PSU (CX600) and motherboard (ASUS M5A97 LE R2.0) just like me has issues running my 960. It runs a little bit better but after some more intensive stress it crashes just like in my case, but insted of error code 3 it's error code 7. Not a big difference, really... So I'm thinking of getting a new motherboard (GA-970A-DS3P) but I'm still not sure if it's the main issue...