Avoiding "Driver Display Has Stopped Responding" Crash Without Lowering OC...Can it be done?

ELB

Distinguished
Mar 15, 2006
80
0
18,640
Hello and thanks for reading. And if you don't like reading and prefer the short version, skip to the question in bold below.

I've recently had two GPUs in my machine, a long-time Gigabyte GTX 460 1GB OC edition and, more recently, a Sapphire AMD 7870 XT. The GTX 460 has been an ultra-reliable card for years; the 7870 XT I had to RMA twice and I just sold the third because I just wasn't happy with the card. (Side note: an MSI Nvidia GTX 760 OC Twin Frozr 2 GB is on its way to me now to replace the GTX 460.)

In the last few months I overclocked both the GTX 460 and the Sapphire 7870 XT. While I'm new to overclocking, I've read many articles and watched more than a few videos on it, so I think I get the gist. In both cases I was able to achieve at least a decent OC (around 15%-20% of the core clock and, in the case of the GTX 460, 10%-15% of the memory; the general consensus seems to be that OC'ing the 7870 XT's memory yields little result). The OCs were stable to the degree that they would run both Furmark and the Unique Heaven benchmarks all the way through without issue. Additionally, I could game with them for a time (at least a half hour, and often 1-3 hours) with no problems.

However, sooner or later the game would crash and I'd get the dreaded Driver Display Has Stopped Responding and Recovered error. While I didn't spend enough time in enough games/applications with the 7870 XT to truly establish its stability at stock clocks, I certainly have with the GTX 460. And there's no doubt that OC'ing the card — at least to the degree that I did (again, 10%+ core and memory) caused the issue. No OC, no issue.

For what it's worth, my Intel i5-2500k is running a healthy, though not especially aggressive OC. The only reason I bring this up is that I'm not OC-knowledgeable enough to know if this could be taxing my system to such a degree that it somehow compromises the GPU's overlclockability. Seems improbable, but, I'll repeat, I'm no expert. And the GPU clearly is the likely culprit, since when I return to stock clocks, the issue disappears.

So my question simply is, Is there any way — besides lowering the OC (possibly to stock) — to avoid the driver crash? Especially since the OCs I've used have allowed for stable gameplay/PC use for a good amount of time before the crash occurs:

•Would raising voltages help (I did raise the voltage modestly with both cards and never came close to having a heat issue — cards stayed below 70 C with fan at 75% max)?
•What about BIOS settings?
•Since the crashes happen after a good amount of use, is this more likely a memory OC issue, meaning I could OC the core?

Any input would be greatly appreciated. I'm okay running stock clocks, especially with the relatively far superior GTX 760 on the way, but I'm really curious, since so many folks seem to be able to attain great stable clocks and I have not. I understand that this could be luck of the draw, but with two cards now having the issue, I'm wondering if it's something else.

Thanks,

ELB

Potentially related system specs:

Mobo: ASUS P8Z68-V
PSU: Corsair 620HX
CPU: Intel i5-2500k
RAM: G.Skill Ripjaws X 8GB DDR3 1600
OS: Win 7 Home Premium
 
Solution
The reason you are getting driver crashes is (as you already know) the OC is unstable. There are 2 possible solutions to this.

1. increase the voltage and hope you can find a stable setting before damage is done
2. lower the OC.

since you don't want to lower the OC then the voltage increase is most likely your option.

Download OCCT and run the GPU test with error detection on. If the card errors then the OC is unstable.

First I would lower one of the clocks to stock. Example lower just the memory OC to stock and keep the core OCed then run OCCT. If no errors Switch the clocks to core at stock and Ram to OCed then test again. This will tell you witch or that both voltages will need to bumped up.

As each chip will OC to a...
The reason you are getting driver crashes is (as you already know) the OC is unstable. There are 2 possible solutions to this.

1. increase the voltage and hope you can find a stable setting before damage is done
2. lower the OC.

since you don't want to lower the OC then the voltage increase is most likely your option.

Download OCCT and run the GPU test with error detection on. If the card errors then the OC is unstable.

First I would lower one of the clocks to stock. Example lower just the memory OC to stock and keep the core OCed then run OCCT. If no errors Switch the clocks to core at stock and Ram to OCed then test again. This will tell you witch or that both voltages will need to bumped up.

As each chip will OC to a different point your yeild will differ from some out there and this is just how it is. I have given all the info you should need to finish your OC and find stability. If you have anymore questions I will do my best to answer them.

for shader settings in OCCT use 2-3 for Nvidia and 7-8 for AMD/ATI cards.

EDIT: It's been my experience that OCCT GPU test with its "Error Detection" can be more useful then running Furmark and Unique Heaven. This is not to say that these programs are not useful in finding stability but they need to used with OCCT in order to find the best stability possible. I use Furmark myself with OCCT to get the best OC's out of my GPU's. OCCT is also useful for stability testing the system as a whole by using the PSU test. The other tests I have found that other programs do a better job of stress testing than what OCCT does.
 
Solution

ELB

Distinguished
Mar 15, 2006
80
0
18,640


Fantastic answer. Obviously, exactly what I was looking for. Will try out OCCT following your instructions. Thanks so much, bgunner.