Windows 7 draws random triangles and crashes with GTX 460

Question

I haven't used Windows for a really long time, but a few months ago I upgraded my Linux box with an NVIDIA GTX 460 graphics card and installed Windows 7 Ultimate (64 bit), just to be able to play StarCraft II.

I finished the single-player campaign, and haven't booted into Windows ever since.

A couple days ago, I booted Win 7 again to make sure it was ready for Crysis 2, but I was greeted with screen flicker and random triangles:

Now, everything was perfectly fine a few months ago, and everything still works perfectly under Linux, including 3D graphics. 100 iterations of MemtestCL up to 900 MB (the card has 1024 MB memory) reveal no errors either.

I've also managed to install all Windows updates and the latest drivers from Nvidia, but no luck there either.

To me, this definitely looks like a hardware problem. However, everything is fine under Linux, and I'm not a Windows expert, so I thought I'd give it one last chance before taking it for a repair/replacement and ask if anyone on SU has any idea what might be causing this.

UPDATE:

Following gordoco's answer, I checked the temperature and fan speeds, and the card is running at a cool 40-45°C, with the fan at 40%.

But things got stranger while I was running these tests. After a cold boot, I was unable to reproduce the problem. So a soft reboot from Linux to Windows reproduces the problem, but a cold boot into Windows doesn't.

One fact that might be related is that this box normally acts as a download box & server, so it runs headless and goes weeks (sometimes months) without a reboot.

UPDATE 2:

Between the cold boots, the new NVIDIA drivers and Win7 SP1, the problem seems to have resolved itself and I can't reproduce it anymore, even with a soft restart followed by hours of Crysis 2.

19 years, 2 CS degrees, and I still don't "get" Windows.

Anyway, thanks a lot to those who replied.

Answer 1

Is your GPU overheating? Perhaps the graphics driver settings in Windows are different than Linux, causing it to run hotter in Windows. The nVidia System Monitor can show you the temperature and fan speeds. Speedfan may be able to do that too. If it's running too hot, check that the GPU fan is running fast enough and it's not set for extreme overclocking.

Answer 2

After a cold boot, I was unable to reproduce the problem. So a soft reboot from Linux to Windows reproduces the problem, but a cold boot into Windows doesn't.

Definitely a hardware problem, but you've ruled out the likely culprit of temperature. It's a stretch, but if you can confirm this with more testing, it's conceivable that some incompatible state lingers in the video card's memory and survives the reboot. Stranger things have happened.

Another alternative, maybe equally likely, is that the card is defective.

A last question, perhaps the most apt culprit: what's your system's power supply? It'd be worth testing the system memory too (Windows 7 has a diagnostic you should be able to find in Control Panel or Administrative Tools). Anything weird or strange, especially hardware-related, almost always boils down to defects in one of these two, but from the behavior you're describing it points more towards PSU than RAM.