I had this issue with a Radeon 9200SE once. It was faulty hardware (graphics card).

That's not encouraging, but I know it's always a matter of time. Especially in a Pentium 4 boiler of a system.
Having said that, while the exposed parts of the heatsink construction on this graphics card have felt very hot to the touch and that's concerned me to the point of putting an extra fan in near it, Speccy showed the temperature as ~40~50C or so, which isn't ridiculously hot.

But of course, simple age is just a factor too.

But naturally I hope it's not a hardware fault (unless perhaps it's just capacitors or something else relatively simple), so I hope there are some other possibilities of the problem. Seems like a notable card that could be hard to replace.

I would try Riva tuner or (something similar) and underclock the core and memory to see if that changes anything.

teh_Foxx0rz wrote on 2022-08-09, 22:08:

Having said that, while the exposed parts of the heatsink construction on this graphics card have felt very hot to the touch and that's concerned me to the point of putting an extra fan in near it, Speccy showed the temperature as ~40~50C or so, which isn't ridiculously hot.

Someone who used to work on hardware environmental testing told me that once something starts to fail due to the temperature fluctuations it has experienced over its lifetime - which he reduced the span of by subjecting things to frequent extreme fluctuations - while you might expect the failures to be more likely to occur when the dying equipment is at extreme temperatures, in his experience it more often failed when it was at moderate temperatures. This was in the context of me complaining about a router that I knew had been going bad for years, but strangely didn't crash more on hot days, and instead seemed to crash less often in summer.

I suppose that if you were to intentionally subject the card to higher temperatures, there is some chance it might help in the short term if you're lucky, but presumably it would make the card get to a completely unusable state much faster! So just to be clear, I wouldn't bother trying that myself 😀

So, update on this topic. I've tried a whole bunch of different late AGP, high-end Nvidia graphics cards (6600, 6800 and 7800) and they all seem to have this issue. Either these have all aged really poorly and no one is speaking about it, or the issue isn't actually the graphics cards themselves.

I've tried reinstalling the drivers but I get the same issues, however.

I wonder if anyone might have any other ideas.

Maybe the data the GPU is getting is already bad because the system RAM that it's coming from is faulty? Try MemTest86.

Alrighty, I tried with different RAM sticks but had no change in behaviour. However, just to explore everything after now being so certain that it's not a hardware issue, after having the same issue on so many cards, I looked into what drivers people were recommending for this series of cards on Windows XP and while perhaps not the only, I saw 93.71 mentioned, and tried installing that. And, well, it seems as though that's sorted it out.

I'm a bit bummed that I sold off that Golden Sample 7800GS+ now after getting the impression from here that it was most likely to be a hardware fault, but, it did run really hot and I wasn't too enthusiastic about its cooler, and I get the impression that a regular 7800GS isn't much different in performance and is pretty capable for Windows XP gaming anyway. Guess I'll see how that pans out.

This has obviously nothing to do with ram. That would make the system instantly to crash long before you notice glitches in video games.