Graphics card only working while overheating.

[Neuplop]

I have a GTX 680 2gb from gainward been working in this setup for a long time but died some time ago and yesterday I tried it again.

Trying it out yesterday was working, and today not, but still gives video output, and in the device manager says "Windows has stopped this device because it has reported problems. (Code 43)".

The fans are not spinning so it starts heating up (I have it on an open bench so I can touch it), after a while the pc restarts (the gpu burning hot), but it working again, and the fans start spinning, and when it cools down, around 55º C in msi afterburner it crashes and stops working again.
(I know it's pretty much a dead gpu but can't replace it rn, no money and if I had it I would wait for a 3060 or ti to be on stock)

Is there any chance to revive it? I don't want to just shove it in the oven because seems like the problem can be found and probably only heating the problem with a heatgun and not risking all the other components would be wiser. Is there any chance is not hardware related?

No dust, the cables are well connected, there are not more pcie ports to try it on, but I do have a worse graphics card that works with no problems on that pcie port so probably isn't that, no bios updates available, already uninstalled the drivers with ddu.

[emosun]

6 minutes ago, Neuplop said:

I know it's pretty much a dead gpu

correct. if you aren't going to replace it or even try reflowing it then theres not a whole lot you can do in the 30 seconds that it works as far as diagnostics.

I'd go to a pc shop and ask if they got any old free gpu's to just hold you over with a display output. least the machine will run

[Neuplop]

2 minutes ago, emosun said:

if you aren't going to replace it or even try reflowing it then theres not a whole lot you can do in the 30 seconds that it works as far as diagnostics.

What do I need to reflow it? And which diagnostics there are? I can turn the fans off with msi afterburner so it can last some minutes may be, even though probably isn't enough.

[emosun]

1 minute ago, Neuplop said:

What do I need to reflow it?

like you said a heat gun

2 minutes ago, Neuplop said:

And which diagnostics there are?

none , the gpu is damaged beyond running stable without any driver so from a hardware standpoint it's not functional. There are no software diagnostics to bother trying until the hardware is stable.

[Neuplop]

1 minute ago, emosun said:

like you said a heat gun

And any idea to which area would be the problem? I think is memory but I'm no expert. I don't want to burn areas that are completly fine.

[emosun]

Just now, Neuplop said:

And any idea to which area would be the problem? I think is memory but I'm no expert. I don't want to burn areas that are completly fine.

again , theres no such thing as determining the problem via "looking at it trying to run". without a proper diagnostic machine at the nvidia factory connected from several years ago you're playing a guessing game.

Seeing as the ram doesn't get very hot just reflow the gpu itself and see what that does first. But nobody ever said it'll actually work , just something that you can try and sometimes get lucky with.

[Mark Kaine]

it's indeed pretty much dead, you can try the reflow thing but even if it works it won't work for long (you need actually a reflow machine for it to having a "chance" to stay alive longer - those machines cost upwards $20k btw) 

 

 

[Neuplop]

1 hour ago, emosun said:

But nobody ever said it'll actually work , just something that you can try and sometimes get lucky with.

It's working now, it had a bit of stutter in some areas, and is at 75% power budget but did a pass on unigine, will play games later.