This looks like dead ram, but it can be a somewhat dead die, too.
I would touch up the connections that you can access. c91 is crooked I guess.
I am very sure that it is not the power delivery, it should not glitch but just crash in that case.
I would recommend trying to contact someone like TAGG XOC (on youtube) before holding the heat gun to the die.
He knows how to be gentle with this vintage stuff.

Robertkopp wrote on 2022-04-05, 00:34:

I would recommend trying to contact someone like TAGG XOC (on youtube) before holding the heat gun to the die.

Oh, nice little channel. Will definitely check it out.

I think I'll order some RAM in the meantime, while searching for ideas and asking around.

Still the problem with the glued heatsink. There must be some kind of nifty chemical that can eat through the glue without rendering the rest of the card into a puddle.

If you have any industrial strength solvent, wear gloves and wet a dab on a cloth. Used it on cars very carefully while detailing. Soak and rub the glue for a bit, dry it off and let it settle being careful not to get any in contact with any metal ect. Repeat. Wear gloves.

If you got, try electronics contact cleaner or brake cleaner. I have no idea what's in there, but old baked thermal compound just vanishes.
Otherwise you can dunk the whole card in 99% IPA for a day without issues, try harder solvents afterwards.
Maybe first lighter fluid and if that does not work you need to get the acetone out I guess.
With those I would however fill up a jar to the top, preferably with a opening that just fits the heatsink and then try to dunk the card flat
and upside down with the heatsink first into the solvent, to avoid the pcb or other components from from touching it for a longer period.
You never know how good the pcb coating is. The last thing you want is to dissolve the pcb into glass fiber pulp.

https://youtu.be/2bjHLg9zkf4?t=10

That heatsink, just, wrong.

Can confirm that Geforce 256 chips don't mind acetone.
Dissolves many things, but will it reach enough of the stuff to safely remove the heatsink?

Maybe heating up the thing close to the chip's destruction temp might melt the glue, but I don't think so.

As a last resort, you could mechanically grind down the aluminum, last stage before the glue carefully with sand paper over a smooth block.

Or you could carefully submerge it head first in sulfuric acid or sodium hydroxide - but don't try that at home. 😁

Thank you for your care, good luck!

What about soaking a thread in solvent and sawing through with that? Direct application to the area needed.

How come nobody suggests chilling it? That's my first thought. Stick it in a ziplock bag and into the freezer for 3 hours, remove and apply heavy vice grips to a fin, it might tear off the fin so maybe use needlenose and length-wise to grip more area. Once clamped, spray the heatsink with an upside-down can of air, liquid CO2 should chill it even more. Start twisting lightly in a controlled fashion, pump it, and pray to the compiler gods for success. It might even be a good idea to somehow add a counter-force to the graphics chip to counter some of twisting force, even if you just pinch the chip between your fingers, any force would help. And for the best chance of success, sacrifice a virtual chicken to the compiler gods before any of this, just with an imaginary chicken and do the motions with your hands, it usually works wonders for me.

maestro wrote on 2022-04-12, 01:04:

What about soaking a thread in solvent and sawing through with that? Direct application to the area needed.

Would take a pretty thin and tough thread. Maybe dental floss, if that won't just dissolve itself.
Or real silk yarn.

Thank you all for the suggestions 😀

ODwilly wrote on 2022-04-06, 07:29:

wear gloves and wet a dab on a cloth

No way to apply anything directly with a piece of cloth. But the glvoes are a must of course!

Robertkopp wrote on 2022-04-11, 21:16:

I have no idea what's in there, but old baked thermal compound just vanishes.

Well, this much I can tell: This stuff is no normal thermal compound.
I would even be surprised if it actually is anything *thermal*.

Cuttoon wrote on 2022-04-11, 23:01:

Dissolves many things, but will it reach enough of the stuff to safely remove the heatsink?

The gap between heatsink and chip is pretty big. I think I can use a small syringe and a lot of patience.
Heating up wouldn't work I think. Grinding everything down, would be absolutely worst case. I believe that would put way too much stress on the solderjoints.
Sulfuric acid would probably mean that I will just have the glue left, but nothing else 😉

maestro wrote on 2022-04-12, 01:04:

What about soaking a thread in solvent and sawing through with that? Direct application to the area needed.

That is an awesome idea. A good alternative to just dripping something on there with a syringe. This way I can apply some force too, without overdoing it.
Freezer only if I really run out of ideas. That never worked for me in the past and I'm already cringing at the thought of twisting the heatsink.
But I will take the chicken advice. Maybe even a goat! Usually reserved for compiling shaders 😉

===

Current status:
Still looking for a RAM supplier. Shouldn't be that big of a problem, I was just lazy 😁
But I've already got my hotair station out of storage last weekend.

My goal is to first get the heatsink off. If that won't work, I don't even need to bother with the RAM.
The poor card would just go up in flames after a short period. Don't trust that glue AT ALL.

If the card behaves the same after (eventually) removing it ... the adventure can continue.

Digitalzombie wrote on 2022-04-12, 08:02:

The gap between heatsink and chip is pretty big. I think I can use a small syringe and a lot of patience.

Definitely try acetone then. Among readily available, relatively harmless chemistry, that stuff is pretty potent.
Local home improvement stores here sell it by the liter, under their own brand - probably intended as paint thinner or to clean stuff for painting.
Consider gloves, specs and provide some ventilation, not to sniff too much of it. Fire hazard, obviously.

! definitely make sure it won't attack the chip's plastic and don't be sure about the writing on it !

Digitalzombie wrote on 2022-04-12, 08:02:

Grinding everything down, would be absolutely worst case. I believe that would put way too much stress on the solderjoints.

No expert on solder joints, but since the force would be mostly "lateral" or "in the same plane as the PCB" I don't think it matters, as they would be stressed all at the same time. So, not like a crowbar to an edge.
And, if anything, I was thinking more if "hobby precision belt saw" or Dremel than of "230 mm angle grinder".
The main catch is rather, apart from the hazard of fucking up, that the last mm will take ages if you dont want to damage the upper crust of the chip's hull.

Digitalzombie wrote on 2022-04-12, 08:02:

Sulfuric acid would probably mean that I will just have the glue left, but nothing else 😉

That wasn't meant too seriously, since sulfuric acid is no toy and depending on your location, buying it in any quantity might land you on a no-fly-list 😜
But, while SA will really go to town on your cotton clothes at any concentration, it does get sold in plastic containers. So, up to a test, but the chip may be fine.
Thing is, aluminium isn't a "noble" metal at all. It doesn't rust, but acid might make short work of it. Same with sodium hydroxide - aka "caustic soda", not that hard to come by. Used to develop photo-chemical PCBs.
So, that will have to be tried some time, I shall make some photos. Just not on a precious card like that.