Just wanted to share my last night unpleasant experience with FurMark and a brand new watercooled GTX 690 and answer that question that I've read posted so many times on the web : is FurMark really dangerours for GPUs ?
Actually, it is, and not only because of overheating.
Last week, I received my second GTX690 in replacement of my previous one that died for an unknown reason after nearly one year (actually, less than 8 month) of valuable service. I then installed my EK waterblock and back plate, added it to my watercooled rig and started testing the loop. All went good, and the temperature of the card was even lower than with my previous 690 - I'm now way better in applying thermal paste and pad than one year ago.
But yesterday, after having changed a radiator on my system, I wanted to stress the GPU to ensure that all was fine with the temps. And for the first time, I decided to use FurMark.
And my GPU died.
After not even one minute of FurMark, at 1280x720 with no AA, the app suddenly freezed and I had to reboot the computer. By the time the app freezed, the temps were around 30°C on both GPUs with a load around 95-99%, so I was very confident that it was only a software issue : my 690 was not overheating.
The bad surprise came when Windows 8 finally booted : the desktop was displayed in 1024x768 instead of the monitor native 1920x1200 resolution. OK, not that bad, probably one of those Windows 8 mystery that was easily solved by switching the res back to 1920x1200.
And then I saw the EVGA Precision X popup telling me that no compatible hardware was found. Seriously ? A look at the devices manager of Windows confirmed that something was wrong : there was one yellow dot on each GPUs, and the card status was something ending by "Error 43" - the dreaded "Error 43".
I first thought that a reboot would fix the problem, but no. Then I removed/reinstalled the pilots but the error was still there. I then checked if my 6-pin power supplies were properly plugged (they were), and,in last resort, tested another power supply.
The Error 43 was still there. My GPU was officially dead.
Now that I think about it, the coil whine was very audible on this model, maybe this was a sign that one of the condensator or VRM was weak and FurMark, stressing everything on the GPU, only shortened the life of an already defective component. But that's very disturbing, nonetheless, to witness that a software can actually kill a GPUs in less than one minute, without even a single warning.
Now, the tough part begins : obtaining an RMA on a GPUs where the stock cooler has been replaced by a third party water cooling system.
Wish me luck !