Asus GTX 960 crash issue

[Dr Prozac]

Hi,

It seems that I have the same problem as described in the SPCR review:

Typically I try to shoot for a standard GPU temperature of 85°C but unfortunately an odd bug that caused our test system to crash the moment the GPU temperature reached 77°C with the GTX 960. However, this seems to be an isolated problem as it easily handled temperatures of 90°C and above when tested in two different systems.

At default fan profile it is rock stable during a Furmark stress test (also combined with Prime95). Unfortunately when I set a quiet temperature/fan curve it crashes the system after two minutes or so. Around 75-80 C.

It's the same on last two driver versions. Any thoughts what can be done to fix this? My first idea was that maybe the poer section or memory section gets too hot, but if it's an issue that is present only on some systems, that it should not be the case.

My setups is as follows:
i5-4690k
Asus Z97 Pro
Asus GTX 960 Strix OC Edition
Team Vulcan 2x4GB 2400 MHz 11 CL
Be Quiet! Dark Power Pro 10 550 W
Crucial MX 100 SSD
WD Red 3TB
R5 case
Windows 7 Prof with all updates

[Dr Prozac]

Damn, normally it crashed after reaching 78 deg, so I tried to set a fan profile that would keep it at 75. But it also crashed after some time. I'm disappointed with this card.

[CA_Steve]

Any artifacting prior to the crash? Any crash error message? What kind of crash do you see? (BSOD, BSOD with error code, screen lock, ...)

Try underclocking the RAM or the GPU (separately) a bit and see if it still crashes at 78 or higher.

[quest_for_silence]

RMA it.

[Dr Prozac]

If it crashed at standard fan profile, I would RMA it straight away. But it survived ~30 minutes of Furmark at standard fan rpm. I'm not sure they'll accept it.

The screen goes blue/white or perhaps some other color (it's not a BSoD), than after a few moments the monitor loses signal. The system freezes (I can hear that it stops playing music in that characteristic manner). The GPU fans stop spinning, or at least it becomes quieter. Then I have to hit the reset button.
Nothing interesting in syslog (though I've disabled the pagefile, so it may have some problems writing anything in case of a crash).

It actually crashes around 75, but it lasts longer then. I tried to set up a Fander 92mm fan blowing at 800rpm from the side, around the VRM area. But it did not help. All is fine right before the crash (no artifacts, stable fps etc.).

Thanks for your suggestions, I'll make some more tests when I find time.

[CA_Steve]

VRM circuitry is designed to run hot. I doubt that's the issue. I think it's a marginal component (RAM or the GPU) that ends up in a timing race condition when temps rise past a certain point (CMOS gates slow down with temp increase). If you try the underclocking test and the device stays stable at 75C+, then you'll know. Try a 20% underclock and see if it's stable with Furmark. If it is, pop up to 5% underclock. Then go up or down as needed. If you can dial it in fairly close, then the next step would be to keep one clock at the given freq and run the other up and vice versa to see if it's the RAM or GPU.

[Dr Prozac]

At the moment I tried lowering the clock by 90 MHz (max that MSI Afterburner allows) and Memory by 200 MHz. It did not help. That's about 7 and 5% underclock.

Maybe the UnofficialOverclockingMode will allow more. I'll also try increasing voltage a bit.