I've been having a problem with my GPU and I think it's getting progressively worse. I'm pretty much at wit's end as to what is causing it, nor am I quite certain that it is my GPU. This is what happens:

After executing graphics-intensive code for a short time (~2 minutes) at what is about normal performance (I think it's slightly below normal these days), the current frame freezes completely for about 10 seconds, and then my framerate is fucked (~5 seconds per frame) until the process is aborted. When I had this problem before on Warcraft III, I would have a 10 second freeze and then it would resume a normal framerate for the remainder of the game and wouldn't freeze again that night. Now, however, I was playing Star Wars Battlefront II and I had to restart every time.

Best example: I tried running 3DMark

-It ran decently fine (30 - 10 fps depending on segment) for the first 1' 40" of Return to Proxycon. I think I used to get a couple more frames per second, but no big deal.
-Then at about 1' 40" it froze completely; after about 40 seconds 3DMark acted like Return to Proxycon was done. It then started Firefly Forest, which executed at about 0.2 fps, rather than the 15-22 I'm used to
-I pressed Esc during the fragged Firefly Forest and it terminated immediately and my system returned to normal

I have been having the freeze-and-resume-occasionally problem on Warcraft for months, but I ignored it. Now I don't get my framerate back after it. What is causing this/what do I have to do to fix it?

I'm running a eVGA 7900 GT on a socket-939 system and it's a little over two years old. I just noticed the serious frag problem, and WC3 has been freezing for about 6 months (it didn't before).

I would recommend monitoring the temperatures of the video card when this happens. Report back with temperature results.

This time it fragged a little way into Firefly Forest. The loss of graphics capability comes a little bit after the beginning of the jumpy temp section and the smooth decline is when I aborted the 3DMark test. I don't think 70C is that huge an issue.

Attachments:

It sort of sounds like your graphics card is crashing and then the driver is re-initializing it. I remember seeing it in action on an ATI graphics card, and I seem to recall hearing that nVidia had a similar feature, though it may function a bit differently (I've never actually seen it in action myself). At 70C, that could be just a bit too high if your card's chip is on the lower quality end of the scale (you never know whether it is or isn't, unless you've experimented with overclocking it before).

For someone who doesn't overclock and therefore doesn't have a lot of spare cooling apparatus lying around what is the bes (a) permanent and (b) temporary way to improve cooling? I'm thinking I could take the side of the case off and lay it open to try to keep it cool and see if it still crashes. My case is in a pretty poorly-ventilated area and it doesn't have the greatest fans.

Leaving the side panel off is definitely one way to help your video card get cool air. if you have a spare fan laying around, if you could temporarilly have it blow down onto the GPU that could help as well.

If this doesn't help, perhaps you could try to reinstall drivers. Although, I've heard a lot about 7900 series cards tanking after lengthy use.

Of course, there are some cases that leaving them open doesn't help cooling or can make it worse. It just depends on the case design, though.

clean the heatsinks. take those dusts off.

is not a very long time... i'd say you have a cooling problem. Since I had that same problem.

nVidia drivers include an overclocking feature (experimental) and a re-initialization 'thingy' that is supposed to change drivers. However, dust on the heatsinks could be a major problem and can even cause a fire. Clean them with a soft vacuum bristle head or even a tissue works. The chances of the card overheating are much grater than that of the drivers needing to be upgraded, in which can be done at the card's specific website. If, all else fails, try putting the card into another computer and see if SC will work. In most cases, you'll find the card drivers are out of date.

Oh, and with overclocking, do it very carefully, incorrect settings will stuff the card altogether, and, even can destroy critical settings.
Hope this helps, good luck.