Graphics card on the way out?

[emyyhh]

Every time i play a game on my pc i am met with a directx error about 10 mins into gameplay. The screen turns black, flashes a bit, and then helpfully lets me know the game has frozen and tells me there was a directx error. Occasionally upon startup the pc BSOD's giving the error code 2057 - which i believe is graphics releated? Also, the gpu was artifacting heavily in COD wow the other day, although now it only crashes rather than artifacting. Other games such as anno1404 have caused it problems, although not to the same degree as COD.

My specs:

i7 920
Asus Rampage extreme
3Gb corsair 1333Mhz (memtest ok)
9800GX2 (not OC'd)
Some old HD (seagate)

The crashes occur whether my pc is overclocked or not. I have even reinstalled vista (64) incase i had some driver conflicts, but to no avail. I have tried a whole range of nVidias drivers, from old to new, with nothing fixing the problem. An 8600gt in place of my 9800gx2 solved the crashes, but turns games into slideshows.

I don't particularly want to replace my GPU as there are no 'exciting' upgrades that fit into a sensible budget (<£200) but if it has to be done, then so be it. What i want to know is; could it be my HD or RAM, and are there ways of knowing for sure?

 

[coozie7]

Random crashes/BSODs usually mean a power or temperature issue.
What is your powersupply?
Use either GPUZ or Rivatuner and check the graphics card temperatures.

 

[emyyhh]

Cheers coozie7,

The power supply is a Thermaltake toughpower 750W. Rivatuner tells me 'no supported drivers for this display adaptor', GPUz is telling me the GPU temp is 60 degrees, with the fan speed at 45%. Because rivatuner won't wokr i cannot increase the fan speed, although I do have a 12cm fan and a 9cm fan both providing cool air specifically to my GPU.

(using forceware 186.18 btw)

 

[coozie7]

I take it 60C is at idle? Not bad at all.
Also get Speed Fan to check the CPU temperatures then download and run Furmark to stress the card out and see what temperatures that causes.
One word of caution: Furmark is a vicious stress test-on steriods. Keep an eye on the system while it is running.

 

[emyyhh]

Right, speedfan says my cpu is at 30 degrees. GPU is again at 60. Furmark ran for about 20 seconds, when the temp of the GPU reached 75 degrees and the screen went black. The comp did not BSOD, but furmark appeared to hang. Only an alt-tab followed by a force-quit would cause the desktop to appear again. I treid leaving furmark open with the screen black, but nothing happened. Once the screen went black the GPU temp dropped slowly to ~60 again.

 

[coozie7]

The card should be able to reach much higher temperatures than that, but it is possible GPUZ is reading a little off.
To cover the bases:
Set a restore point then update Direct X, you will have to do this manually because this does not happen during routine Windows updates.
Make sure Windows is fully updated.
Strip and clean the card, using new thermal compound during reassembly, Arctic Silver 5 is popular but any premium product will be OK. Do not lose any thermal pads that contact the memory chips and you can apply thermal compound to them to ensure good contact during reassembly.

 

[emyyhh]

Right, that was a lot of effort. I had some Artic Sliver lying about, so i used that in place of the crappy nVidia stuff. However the problem still pursists. Am I right in thinking that after a clean vista install, directx updates (that i did not need) and trying numerous drivers (after using driver sweeper, and after a clean install) that the problem can be attributed to ONLY my GPU, or is it possible my HD is knackered, or perhaps my RAM?

Hmmmm, very frustrating.