help GPU black screening

[Tirant]

My friend bought a prebuilt Alienware Aurora R11 that came with an EVGA manufactured 3080 that was having issues outside of the warranty window. I hooked it up and got into Windows, temps on the card looked fine, however after a few minutes the screen would flicker to black (sometimes come back for a sec) and then remain black but still outputting some kind of signal to my monitor. Windows looks to have lost the graphics device and could not reinitialize (and is completely unresponsive unless I hold the power button) gpu_crash.evtx. Once this happened, I could no longer successfully boot or even get into my BIOS (screen would just show a black signal). When I booted my PC using my integrated GPU and looked at the data for the 3080 using GPU-Z, it was detected as a 3080, but tons of info could not be read from the chip.

 

At this time, I decided to reflash the GPU using a compatible one online and double check the drivers and everything were up-to-date. I also use a 3080 normally (albeit a much nicer ASUS one), so I was confident my system could handle the card and any issues would be with the card itself.

 

 

So I took it apart to see if there was anything visually wrong with it using my eyes, and it looked fine. I don't have a microscope, so I could have missed something. I assumed at this time it might be a power/clock issue, like the clock is dropping too much, but I don't know enough about it and couldn't reliably boot to change settings and see if I could finagle it. I did take the pictures of the back of the 8-pin connectors, they are discolored but should be fine.

 

 

When I removed the heat spreader covering the RAM, the thermal pads got ruined and needed to be replaced. There isn't a lot of info online about this model, so I guessed 2mm which ended up being too thick. I know this because when I reassembled it, it actually booted and worked just fine for a while playing games. I was able to view it in GPU-Z and was otherwise a perfectly functional card.

 

 

 

The problem I had though was that the heatsink wasn't making good contact with the GPU based on the temps -- we're talking about huge fluctuations up to 70-80-90-100C, so I decided to go down to 1mm pads. When I removed the heatsink, the paste did not completely spread, so looks like it was a few hairs short of contact.

 

After doing all this with the new pads, the card wouldn't boot again at all (same black screen issue as before). I thought, huh, all that changed was the contact with the heatsink, so I started loosing the heatsink thinking it was a mounting pressure thing. Sure enough, after basically completely removing the heatsink, it worked again.

 

So, I'm assuming there is an issue with the mounting pressure on the GPU die. I think either the GPU die is toast, or there is an issue with some of the solder balls affixing it. I'm assuming this is because Dell cheaped out. I'm considering getting it reballed -- I'd try it myself since the card is broken anyways, but reballing a GPU die takes soldering expertise I lack. I could probably try and reflow it myself, but I wanted to look into options first since reflowing isn't as reliable as reballing. Is this a reasonable conclusion?

 

Sorry if this was long-winded, I'm just trying to explain everything in case someone has ideas. Thank you for any help!

 

tldr

GPU only shows a black screen if there is basically any mounting pressure.

[Bombastinator]

Intel or AMD cpu?

[Tirant]

1 hour ago, Bombastinator said:

Intel or AMD cpu?

Intel 12700k