What could possibly cause this problem?

[RhoXS]

About 4 years ago I purchased an EVGA GeForce 7900GS video card. About a year ago, even though it was never overclocked, it started to intermittently display random characters, artifacts, and generally garbled images, to one degree or another. It even caused spontaneous reboots sometimes. Since it was out of warranty, I just replaced it with a new card.

I now would like to use it on a hand-me-down machine I use to check stuff out before I trust it on my main machine. The problem, once intermittent, is now full time with major artifacts appearing even during the post process before W7 even starts.

I initially thought this was a cooling problem so I removed the fan/heat sink and remounted it using AS5. Since the problem now occurs as soon as it comes alive during the post process, I do not think it is a cooling issue. In any case, I also tried cooling it with a big noisy high volume supplemental fan and also sprayed it top and bottom with "Coolit", all to no avail. Therefore, I do not think this is cooling related.

It troubles me to just put this old but elegant piece of technology in the trash (even though I just ordered an inexpensive card to use in the hand-me-down). Is this just a practically never to be fixed problem due to a bad GPU or memory chip and it is just time to send it to the recycle bin?

 

[Arkadrel]

Some cards last longer than others, it sounds like your 7900GS has lived a full life, time to throw in the bin.

 

[soccerballtux]

seems top of the line cards don't last as long as the lower end models, especially with nvidia ones...for me it was like mine just started producing more and more heat.

 

[RhoXS]

Well, I might have found the self imposed problem, fixed it, and then inadvertently destroyed what might have been a good card all along.

A few years ago, in an attempt to reduce noise, I replaced the EVGA cooler with a large fanless heat piped heat sink. It worked great but took up an enormous amount of space in the case. About a year ago I decided to get more control of the inside of the case and replaced the fanless device with the original EVGA cooler. The card worked briefly but then the problems began. I then simply replaced the card and did not bother with it again until just recently.

This is what I think happened.

The top surface of the memory chips are slightly lower than the top surface of the GPU. Therefore, if the EVGA cooler is resting on the GPU (as it should be), there is a very visible air gap between it and the memory chips. EVGA had each memory chip covered with a thick rubbery heat conductive tape. The thickness of the tape allowed the tape to come into good contact with both the lower surface of the cooler and the top of the memory chips.

When I replaced the fanless heat sink with the original cooler, I had no choice but to reuse the heat conductive tape as I had nothing else to bridge the air gap. Since the tape no longer stuck well, I put a small dab of AS5 on each side of it and I only used AS5 on the GPU. Since no tape was now used on the GPU, the bottom of the cooler was in very tight contact with it and squeezed all the AS5 out. The AS5 that got squeezed out managed to get onto some of the tiny conductive spots that ring the GPU. I also noticed some AS5 on the circuit points around some of the memory chips.

I cleaned all the AS5 off. To test the now cleaned card (that just a few minutes previously was headed to the trash), I tried it without a cooler as it was going to continue its journey to the trash if it still did not work. I assumed, very incorrectly, no harm would come from just quickly letting the machine boot and then promptly shutting it down. To my overwhelming surprise the card worked without an artifact through the post process. Almost immediately after the "Welcome" screen appeared artifacts started appearing. An instant later the screen went blank and the smell told the tale. The GPU was so hot I burned my finger just quickly touching it.

Obviously, there are some good lessons to be learned here.

 

[BD231]

lol rhoxs. I tried to make ramsinks out of pennies from 1981 or earlier and put them on a 9700pro. I did no sanding or lapping, I just put a layer of thermal tape between each penny about 5 pennies high then stuck one stack on every BGA chip. I couldnt even get the card to post after the application . Eventually it started working once I took everything off but I was almost certain I'd killed my $200+ dollar card with nothing but pennies.

Have you checked for any busted caps op? I'm pretty sure that's one of those cards you can throw into a conventional oven for a few minutes and fix, you might look into that.

 

[3DVagabond]

I'm pretty sure that's one of those cards you can throw into a conventional oven for a few minutes and fix, you might look into that.

Yea, try baking it. :thumbsup:

 

[RhoXS]

I'm pretty sure that's one of those cards you can throw into a conventional oven for a few minutes and fix

Are you serious? I think I already baked it real good when I lit it off without a heat sink on the GPU.

What temp and for how long? My wife liberally uses rosemary, salt and garlic when she uses the oven. Will that be useful here? I'd rather not use butter cause that might mess up the pristine interior of my new Lian Li mini tower.

BTW, I did check and all caps look intact.

 

[GodisanAtheist]

Yes they are serious. Putting a videocard in the oven and heating it causes the solder to soften and reflow, potentially bridging any cracked interconnects. Its almost common procedure on dead 8800's, although its worth mentioning that this often proves to be a temporary "tide me over" fix to a newer card.

I've had a pair of 7900GTs give up the ghost on me (worked fine, then random artifacting, then funny pixels, then whammo bios image corruption). The Ultrashadow engine on those cards NEVER worked right.

Seeing how the 7900's were Nvidia's 1st 90nm cards, they could have just had a higher than normal failure rate.

 

[zagood]

and also sprayed it top and bottom with "Coolit"

WTF?