For the last few months, my game machine has been intermittently stable. It'll run forever as long as I don't play any games, but as soon as anything 3D starts, I have a 50/50 shot of crashing within about an hour. It always sounded like a heat issue (which I've posted about before), but my case, PCU and mainboard temps have always been good. I've looked around inside and can see anything amiss.
Well, I finally the cause of my instability. When I came into my office this morning, I was greeted by a bad smell: burning electronics. You know the odor. Hot circuit boards. In a room full of computers, that's an alarming smell to say the least. So I got the flash light and started taking covers off PCs and sniffing. I found it on the first try: my game machine.
My video card (a Leadtek 64MB Geforce2 GTS Pro -- a nice card which I like a lot) was nearly smoking and the fan was making noises like a tiny rodent was being tortured inside it. I remembered seeing a utility that Leadtek bundles with their Windows drivers which had some sort of hardware monitor, so I found it and fired that up. The core temperature of my GPU was 91 degrees and the edge temp was 60-something degrees. As in degrees Celcius. The fan had crapped out, getting slower and slower over time.
I shut down and yanked my card out to see what was what. I'd never noticed the fan slowing because it's really hard to see. The AGP card is upside down, and with my tower case on the floor, you have to rest your head on the carpet and peer into the case to see that it was slow/stopped. The tiny fan barely turned when rotated manually.
I took the bad fan off and tried to find a replacement among all my junk. I gave up after about 5 minutes; there's too many boxes to go through. This is my lunch hour as well as my video card I'm burning, so a compromise had to be struck. I remembered where an old fan/heatsink from a K6-266 was, so I took the fan off that. The screw holes didn't even come close to matching up, so I attached the fan onto the card with a complex system of zip ties. It blows air down onto the heatsink (which is all the other fan did), and I figured that since it's a bigger fan, it'll cool better even if the marriage between heatsink and fan wasn't exactly a match made in heaven.
I reinstalled the card and booted Windows. After a couple minutes of non-3D activity, the core temp was 79 degrees C and the edge temp was 54 degrees C, which was not that great of an improvement in my opinion. So the only thing I could do after that was use the Leadtek overclocking utility to underclock my card. I dropped the "Graphic Clock" to 150MHz (from 200, I think) and the "Memory Clock" to 315MHz (from 350). After a couple minutes, the temps are 73 core and 49 edge. That's looks to be about as good as I'm gonna get.
I need to get a new fan. The old and busted one is a T&T brand, model MW-410H2S which is a little 40mm, 12v fan with a three-conductor cable and the tiniest Molex connector I've ever seen. I could only find one page for it so the new hotness might be A Sunon fan since other similar 40mm fans are easy to find. I might have to do a little cutting; the old fan has no "sides" and fits snugly into a socket-ish thing on the heatsink. I'd better buy two then.
I could also get a new video card but honestly I like the one I have as long as it doesn't get hot enough to boil water.