Kahenraz wrote on 2022-03-30, 16:51:

The heatsink is cemented on and is making good contact. That's not the problem.

I didn't think it was, I thought I'd share my own experience of a short lived used graphics card that I purchased.

>3o ohms across one of the 22uF tantalums

30 ohms is huge, not a short https://www.calculator.net/ohms-law-calculato … =watt&x=61&y=20

I seem to remember Mobile Radeons 9xxx dying in thinkpads, and reballing as successful fix, as in the defect wasnt directly in the die bonds like in Nvidia GPU plague, but early RoHS balls cracking due to thermal stress

I've had lots of things arrive dead, but it's harder to think of things that legitimately worked for at least a few seconds and then quickly died.

One that comes to mind was an MSI 440BX motherboard. I remember booting it up on the first try, and thinking it was interesting that it reported onboard voltages on the POST screen. I don't remember if it froze or kept running, but after that one time, I could never get that board to POST ever again.

I also had an Asus nForce(1) motherboard that booted up and within a few seconds a capacitor exploded. I replaced it, and come to think of it, I've still never retested the board after that.

I had a laptop RAM module die after about an hour of memtest86, it had been working perfectly until the moment it fell over. I'm still unsure of the wisdom of even running memtest on laptops because of that experience - heat is an issue with those things.

Kahenraz wrote on 2022-03-30, 01:12:

It's a FireGL 8700 (Radeon 8500). I removed the tantalum and it tests fine. There is still a low resistance between the pads but it's not a dead short and this may be normal for the circuit. I don't have another card to compare.

Unless I can find a bad capacitor, there is nothing visibly wrong with it. No missing surface mount components or broken traces.

Very odd.

Can you provide us with further info on what the card does / doesnt do? (Do you even get a garbled output? Any sort of "Im missing a graphics adapter" Beeps? Anything? Looking at the generic picture in TechPowerup maybe it would be a good idea to actually replace all the capacitors. I know its a bit before the "Great Capacitor Catastrophe" but alas, its a 20 year old card. 8500 are quite durable. Used to have a couple. One was the vanilla 8500, and the other was an All in Wonder. It would sometimes, very rarely behave stupid, but would recover from it. And I ended up killing it by accident.
Anywho, more info please? Thnx.

I've since removed all of the capacitors from this card and they all test fine. They're all polymer capacitors, so I would have been surprised if this was the issue. I also lifted one side of each tantalum capacitor and verified that none of them are shorted.

Since I'm confident that it's already dead, I'm slowly removing each component from this to see if I can find out where the fault is. I'm wondering if it might be the crystal? Of course, it could also be the GPU itself.

I tried to remove the heatsink and found it to be VERY well attached. I've never tried the freezer method of removal, and after leaving it in overnight, I was only able to get one side of the epoxy to "crack" a bit before the thing had warmed back up. I've put it back into the freezer and will try again later. I'm using a nylon spudger which doesn't provide as much leverage as a screwdriver, but has less risk of damaging the board.

Freeze spray works the best, a good few seconds blasting the heatsink, and it will pop off with minimum effort. I've used the technique many times now, and it's the best for removing stuck on heatsinks, I reckon.

By "freeze spray", do you mean an upside-down can of compressed air or something else?

The kind you would find from your hardware/electronics stores, meant for rapid cooling. The can I have will take things down to -55°c apparently. They typically use different chemicals to achieve this, so more effective than compressed air upside down, though if that's what you have, it could be worth a shot.

In a situation like this my first guess would be that the seller sold me a baked GPU in hopes that it will last long enough so you can't reasonably open a case.

RockstarRunner wrote on 2022-05-05, 13:19:

Freeze spray works the best, a good few seconds blasting the heatsink, and it will pop off with minimum effort. I've used the technique many times now, and it's the best for removing stuck on heatsinks, I reckon.

I agree, it always worked for me as well, although I don't use a freeze spray, I just put the card inside the freezer at -21 degrees C, and let it sit there for a few hours.
After that, twisting the cooler will usually work, although those with superhuman glue might also need a twisted screwdriver between the PCB and the cooler (with an old credit card in between). This requires a bit of experience, though, because you're just a slip away from kissing that card goodbye 😁 When the epoxy/glue is frozen, it's usually enough to just do a gentle twist and it will pop right off.

Freezer works pretty well for epoxied heatsinks ~

The argument to be made for Freeze spray, over putting in the freezer would be Freeze spray takes only a few seconds, and can be localized to the heatsink, instead of freezing the whole card for hours in the freezer.