[SOLVED] Burnt GTX 1080

[SuddenLoss]

I left my PC idling with New World in the background while showering and returned to my PC being shut off and releasing a burning smell. I dismantled every thing to find out where the smell came from. It was my GPU. Now, there is obviously a bit of damage but is there any hope in repairing the GPU? The warranty has already expired.
One thing to note is it has been approx. 4 months since cleaning out my case with air. And I have been keeping my two dogs (German Shepherd and Shiba Inu) in the computer room. Their fur plus other dust buildup must have had a big part in overheating.
NOTE: No visible damage to the backside. PC runs fine without GPU connected.

Specs:
All purchased in March/April 2017
PSU: EVGA Supernova 650 G2 80+ 650W
Motherboard: ASUS ROG Strix z270I Gaming
CPU: Intel I7 6700K
GPU: EVGA GTX 1080 FTW Gaming
RAM: VENGEANCE RGB PRO 32GB (2 x 16GB) DDR4 DRAM 3200MHz C16

 

[Ralston18]

I suspect that there is very little likelihood of repair.

Also:

What PSU is installed in the host computer: make, model, wattage, age, condition? May not be or no longer be able to meet the host computer's power demands.

Not sure about that thermal paste application - was that redone sometime before?

All in all a "perfect storm" of factors that destroyed the GPU.

 

[SuddenLoss]

I suspect that there is very little likelihood of repair.

Also:

What PSU is installed in the host computer: make, model, wattage, age, condition? May not be or no longer be able to meet the host computer's power demands.

Not sure about that thermal paste application - was that redone sometime before?

All in all a "perfect storm" of factors that destroyed the GPU.

The PSU is a EVGA Supernova 650 G2 80+ 650W. Nearing 5 years and still good condition. Thermal paste has never been reapplied and I bought the card brand new.
The PC runs fine without the GPU connected.

 

[USAFRet]

Thats a dead GPU.
There is no 'repair'.

 

[Remeca]

It's painful to see. Good luck with your search for a new GPU.

I would not use that PSU on any new one. Just in case.

 

[InvalidError]

Check Vcore-ground resistance using a meter with at least 0.01 ohm resolution. If you get more than 0.5 ohm after knocking off whatever may be shorted out on the Vcore VRM, then there is a chance you may be able to solder-in an aftermarket VRM to resurrect the GPU. If you cleaned up the charred VRM area and all seemingly shorted components yet still get less than 0.4 ohm on Vcore, the GPU die itself is likely shorted out and there is nothing that can be done about that.

 

[Remeca]

Check Vcore-ground resistance using a meter with at least 0.01 ohm resolution. If you get more than 0.5 ohm after knocking off whatever may be shorted out on the Vcore VRM, then there is a chance you may be able to solder-in an aftermarket VRM to resurrect the GPU. If you cleaned up the charred VRM area and all seemingly shorted components yet still get less than 0.4 ohm on Vcore, the GPU die itself is likely shorted out and there is nothing that can be done about that.

He gonna solder new VRAM on there?

 

[InvalidError]

He gonna solder new VRAM on there?

Why VRAM? The VRAM looks fine.

What he needs, assuming the GPU die itself isn't fried, is to bypass the burnt VRM with an aftermarket one. Not sure if anybody still makes those today but there was a time where some companies offered aftermarket VRMs for extreme overclocking. You rip the original VRM out, patch the VID and control pins through, solder jumper wires from the VRM board to the GPU's power planes and off you go.

 

[SuddenLoss]

Why VRAM? The VRAM looks fine.

What he needs, assuming the GPU die itself isn't fried, is to bypass the burnt VRM with an aftermarket one. Not sure if anybody still makes those today but there was a time where some companies offered aftermarket VRMs for extreme overclocking. You rip the original VRM out, patch the VID and control pins through, solder jumper wires from the VRM board to the GPU's power planes and off you go.

Any good resources you could point me towards to get a better understanding of how to go about attempting this?

 

[InvalidError]

Any good resources you could point me towards to get a better understanding of how to go about attempting this?

For the aftermarket VRM stuff, that was from a fairly long time ago. As I wrote earlier, I'm not sure this is still a thing, haven't seen mentions of this sort of OCing extreme in a long time. Back whenever that was, GPUs only had 6-8 phases VRMs. Ridiculously overkill aftermarket VRMs don't make anywhere near as much sense when you can buy a neatly integrated package at retail with 16-20 phases stock. Power supply rail tolerances have also become much tighter since then.

As far as the repair itself, you can look at Rossmann videos, he's repaired a few Macbooks with charred circuitry on them, even "repaired" at least one board by drilling a hole straight through an area that caught on fire during initial power-on testing as an hail-Mary data recovery attempt. Of course, nobody in their right mind should trust a computer or component with a charred PCB as a daily driver, that should only be done for data recovery purposes.

Since there is no irreplaceable data on a GPU and the damage is fairly extensive, I wouldn't bother with it. I'd never trust the repair any further than the nearest water bucket.