6800GT PCI-E running EXTREMELY hot

[jlambvo]

Ugh... I have a pair of XFX 6800GT PCI-E cards in my system. I purchased on off of eBay and hadn't had a single problem with the system until playing Oblivion recently, where my system would power off at random intervals. I noticed it was getting pretty hot, and when I enabled "Notify when GPU core temperature exceeds threshold," I was getting the popup so frequently I basically couldn't play the game.

Thing is, one of the cards has been idling in the 80C range. Right now, I have my case open and a standard room fan blowing onto it from a couple inches away, and it dropped to about 70C. The other card as always been about 15C cooler, and is currently at 59C. Under load, the main GPU goes over 100C pretty regularily. Given the default core slowdown threshold of 127 degrees, I didn't think it was much to worry about, but from descriptions online this sounds waaaaay not right.

Now I can't tell these two cards apart so I don't even know which one I bought off eBay... but I'm a little suspicious that the seller might have mounted a third party heatsink/fan on this card, and did a shoddy job of re-installing the factory one, so it isn't seated properly or the thermal glue is no good. There's an almost exactly 10C difference in temperature pretty consistently.... is that indicative of something else, or does this heatsink probably just need a re-seating?

Any tips or thoughts are appreciated!

 

[RussianSensation]

I'd probably isolate which card is overheating by testing each one separately in the same game. Then if the problem persists, I'd try remounting the heatsink. If the 2 cards run cooler alone, it just might be that in SLI the cards are too close together and one overheats the other. In that case you'd probably need an aftermarket heatsink for the one that's running hotter. Make sure there is no dust in the fans either.

 

[orangat]

The 6800gt is a moderately hot running card. My 6800gt summer temps run at 55C idle, 85C load give or take a few degrees. Your temps do sound way too high.

I'd definitely open up the problem card and check the heatsink. My guess is that the thermal paste was slopped on too thick.

 

[SynthDude2001]

That definitely sounds like you could benefit from reseating the heatsink. I bought my 6800GT new in 2004 and I still had that problem within the past few months. It had gotten to the point where my card was idling near 70C and load was as high as 105C. The 6800 series cards do run very hot, but they shouldn't really be that hot (unless you have a death wish for the card and you're running your PC outside or something ). I reseated my heatsink with AS5 (after removing the stock thermal stuff and cleaning all surfaces very well, both core and memory), and I got a 15C reduction in my load temps - big improvement, and no more crashes. I followed a guide here, and it was pretty easy. I'd go ahead and do that if I were you, it doesn't take all that long and the potential improvements are quite large.

 

[jlambvo]

Okay, just wanted to get some other opinions. Appreciated!

Thanks for the guide link, SynthDude, that will be a huge help.

 

[Nextman916]

sell them now on ebay in this condition and advertise them as brand new mint condition.

 

[jlambvo]

So, I FINALLY got around to attempting the process of reseating the HSF with new thermal grease, and think I successfully killed the card.

When I first tried putting it in after re-assembling, as soon as I powered up the system a broken-rhythm high pitched squeel/buzzing was being emitted from the 6800GT I had just re-assembled. After a few seconds it would become sustained and the system would power off. Attempting to power back on, it would almost immediately turn back off. Leaving it sit for a little while, it would stay on a little while longer and the squeeling more intermittent, but never actually POST.

I left the card out again for a while. I am afraid that I either tried putting it in too soon and the grease hadn't set properly or something (is that possible?) or that there was some residual isopropyl alcohol which hadn't evaporated and was shorting something out.

After a few hours, I gave it one more shot. This time there was no squeal/whining and it POSTed find, and even started booting into Windows. But just when it is about to draw the desktop, I get a BSOD noting a system checksum error or something (it only flashes briefly before the system resets).

It boots fine in safe mode. The same thing seems to happen whether the SLI bridge is on or off, and the selector card thingie on the MB is set to dual or single cards.

Is this thing beyond repair?