AMD R9 280x Toxic Problem.

[hypersomnia]

Good day everyone,

I figured i'd try to fix my old r9 280x toxic, just to see if something can be done.
Now before i explain what's the issue, be advised that the card probably came in contact with conductive paste and got fried because of it, still not 100 % sure about it.
So basically, one day i wanted to change the stock cooler because it wasn't working properly, i bought an arctic cooler and i thought it would have been a good idea to install vram heatsinks as well, that's where the thermal paste problably broke the gpu.
Every time i try to run it as of now, there are permanent glitches on the screen, these are not driver related, they are literally colored stripes on the screen.
On top of that if you try to install any sort of driver the gpu will just crash, it's only usable when it's not using an amd driver.
The gpu started experiencing problems right after it started lagging while playing a game, after that it got corrupted in some way.
Of course any other gpu works fine on the system, the problem is the r9 280x itself.

What do you guys think? Is it dead?
I will upload images of the glitches soon, thank you all for the help in advance.

 

[jay32267]

I think I would already have gotten rid of it. Yes, I think it's dead.
All the symptoms you mention go with "dead video card".

 

[hypersomnia]

I see, thank you for the info, let's see what others think about it.
I really want to investigate more.

 

[Eximo]

So I am guessing you used thermal adhesive on the VRAM, that usually isn't a problem unless you got it everywhere, I don't think the stuff Arctic provides is conductive. Now the heatsinks themselves are conductive, so if you managed to short something out with them that might explain some of the issues. Maybe some filter caps are bridged and causing memory instability.

Those heatsinks can be removed just twist the heatsinks gently back and forth with a pair of pliers until the adhesive gives way. Don't just wrench it in one direction harshly, that might pull the memory off the card.

The Arctic cooler itself might also be bridging something, you can try the original cooler back on there if you still have it. Excessive installation pressure might also be bowing the PCB, a loose connection in the PCIe slot can cause the symptoms you describe. So it might just be a matter of loosening it and re-applying the cooler more gently.

Last resort is stripping the card completely of everything and attempting a re-flow or simple re-heating in a toaster oven. This can correct small connection failures on occasion. Linus Tech Tips has a decent video on how to do it (though they did it to an SSD as I recall) The key is not to get it warm enough to melt everything, the components will just fall off. (You should not use a household oven as there is a potential for food contamination as the card may outgas chemicals) I recommend finding a toaster oven at a second hand store.

 

[hypersomnia]

Thank you for the long message, however i've already tried everything that you stated save the oven method, that probably isn't going to fix anything.
The gpu clearly slowed down and crashed when it supposedly came in contact with the paste, after that there were permanent graphical glitches.
I didn't use thermal pads, and also yes, the thermal paste provided by arctic is conductive and thus dangerous for the pcb.
But hey, you could be right, i would love to think the artifacts come from an instability of some sort, even though i highly doubt it.

 

[Eximo]

Do you mean to say you used electrically conductive thermal compound around the GPU? Old Arctic Silver, yes, is conductive. Something like Arctic MX-4 would not be conductive. (Those are different companies by the way) You didn't use the thermal adhesive on the GPU did you?

If so, what you should do is take the cooler off, take a toothbrush soaked in isopropyl alcohol and go to town on the surface mount components surrounding the GPU. Should be filter caps and pull up/down resistors only there. Maybe some opto isolator packages. One of those being shorted could easily cause instabilities, but if they were, and were unshorted, they should still be fine. Current just wasn't passing through them.

You might as well experiment with it, I suppose. Can't hurt at this point. If all fails, maybe put the GPU coolers up for sale, someone can probably make use of them.

 

[hypersomnia]

Hmm, that could be a good thing to try Eximo.
The paste i used was the one provided with the arctic extreme III cooler, it's not the same type as the mx-4.
And yes i applied said thermal compound on the vrams, then placed the heatsinks on top of the paste.
Although, it would be nice to know if the gpu itself can be repaired with the use of specific tools, and i'm talking about professionals, in case everything else fails. I have no knowledge about the boundaries of gpus damage and how much of it can be repaired.
If there is some kind of hope, i would like to make it run again.
As for the cooler, it's apparently useless, it doesn't dissipate heat properly anymore, reason is unknown. Last test i did: 100 % fans usage, max temp 100 degrees celcius, of course resulting in thermal throttling.

 

[bignastyid]

Please don't bump posts.

 

[Eximo]

I've had occasion to buy one of Arctic's coolers, not the Extreme III though. Pretty sure a small tube of MX-4 and an aluminum tube of thermal adhesive is what it came with. The current listing on Newegg states "MX-4 pre-applied" Only the adhesive was meant for the VRAM and loose heatsinks. MX-4 is a fairly mid-range thermal compound, but quite versatile.

Even if the heatpipes were damaged it is still a decent chunk of metal and should offer better cooling than that. It is possible the GPU is drawing way too much current and overheating. That could also explain artifacts.

Go ahead and give it a thorough cleaning around the GPU and see what happens.