Is repasting killing my cards?

[da na]

Odd question but I have had now two rare old graphics cards die within days of being repasted with Arctic MX-4. 

I am confident that my repasting method isn't causing any issues - cleaning the die with 99% isopropyl, scraping off any hardened old paste with a plastic blade, and applying a moderate amount of paste to the die. Of course there's the occasional runoff of alcohol or crusty paste onto the PCB but I've had that happen with all sorts of hardware and it hasn't caused problems.

But is the general idea of repasting somehow bad? It seems remarkably odd that a GeForce 7950 Go GTX and ATi Radeon X1800M would both perish from normal maintenance after years of running in the high 80s-90s under load because of dry paste and clogged fans. Does running at reasonable temperatures under load once again somehow damage the BGA solder? 

I understand how stupid of a question this is, but I'd love to not have any more hard-to-replace cards die on me, if I'm doing something wrong please let me know.

[Somerandomtechyboi]

did you cover the entire die of the gpu or did you just put a small dot like you would on a cpu and called it a day? cause you have to cover the entire gpu since theres no ihs, i dont really like spreading thermal paste so i usually just use a ton of thermal paste to really make sure the die gets covered, did that on my 580 2048sp and no issues though there was alot of thermal paste since i wanted to make sure the corners didnt dry out overtime

 

if the die got chipped thatd probably mean instant death for the gpu so i dont think thats whats happening here

[SImoHayha]

they were probably already on the verge of dying and you repasting them didn't really do anything. honestly you running those cards at 80/90s probably accelerated the damage of them. 

[da na]

2 minutes ago, Somerandomtechyboi said:

did you cover the entire die of the gpu or did you just put a small dot like you would on a cpu and called it a day? cause you have to cover the entire gpu since theres no ihs, i dont really like spreading thermal paste so i usually just use a ton of thermal paste to really make sure the die gets covered, did that on my 580 2048sp and no issues though there was alot of thermal paste since i wanted to make sure the corners didnt dry out overtime

 

if the die got chipped thatd probably mean instant death for the gpu so i dont think thats whats happening here

I properly spread the paste across the whole die, I tend to use a little more thermal paste than what's needed on GPUs anyway

[da na]

2 minutes ago, SImoHayha said:

they were probably already on the verge of dying and you repasting them didn't really do anything. honestly you running those cards at 80/90s probably accelerated the damage of them. 

Yeah unfortunately I don't know the history of these laptops I buy used, and I also don't have them turned on for very long before repasting the GPUs. I just know neither card didn't show issues before being serviced besides running very very hot

[da na]

The reliability is so inconsistent too - system that, 15 minutes ago, was doing this:

Spoiler

is noww POSTing with no artifacts in sight. Very very odd

Spoiler

So they could've had problems for years that I would've never known about. Seems to only happen after I hit the cards with a high load, this ATI died after 15 seconds of a benchmark

[8tg]

I think it’s more likely that it’s related to more recent use. Some hardware will be fine for function after being entirely unused for decades, but start running power through some components and they start to actually show their wear. 
This has been the case with my micron transport xpe laptop from 1997, where under normal circumstances, it just would turn on fine.

But now after having it for a while and running it a few times, it’s showing signs of capacitor failure where it’ll fail to keep power. So booting once or twice it didn’t have any issues, but it was really accelerating the wear on the power delivery doing so.

 

It’s like if you shattered a laminated window. It will still remain a window indefinitely, it’s precarious but stable, and you can probably open and close it a few times. But every time you do, it gives into the stress a little bit more. Until eventually opening the window causes the pane to bow in, and closing it causes it to fall out of the frame.

[da na]

12 minutes ago, 8tg said:

I think it’s more likely that it’s related to more recent use. Some hardware will be fine for function after being entirely unused for decades, but start running power through some components and they start to actually show their wear. 
This has been the case with my micron transport xpe laptop from 1997, where under normal circumstances, it just would turn on fine.

But now after having it for a while and running it a few times, it’s showing signs of capacitor failure where it’ll fail to keep power. So booting once or twice it didn’t have any issues, but it was really accelerating the wear on the power delivery doing so.

 

It’s like if you shattered a laminated window. It will still remain a window indefinitely, it’s precarious but stable, and you can probably open and close it a few times. But every time you do, it gives into the stress a little bit more. Until eventually opening the window causes the pane to bow in, and closing it causes it to fall out of the frame.

Thank you.

That checks. I'd never quite thought about it that way and now that you've made the analogy, it makes perfect sense.

To replace the fried 7950 GTX in my XPS M1710, I bought a new-old-stock Quadro FX 1500m still in its Dell factory packaging. That card has been working totally fine for months, handling whatever I throw at it. Never quite thought about how hammering a card then leaving it unused for a decade could cause damage.

Guess I never considered it because of all of the old cards that do work fine - it's odd, I've almost only had GPU death in Dell XPS laptops, save for one HP Pavilion. But a lot of my working cards are in laptops that were clearly loved and used for a very long time versus being left in storage for most of their life. Hell, I've got a Compaq workstation with a Quadro FX 1500m I bought of a guy who was still daily driving it until last year. That card runs like a dream...

 

Can't believe I was too stoopid to realize this, but it's very obvious now.

Death is caused by the fact that I only put a heavy load on the cards after repasting; of course it wouldn't be the repasting itself.

[PDifolco]

47 minutes ago, da na said:

Odd question but I have had now two rare old graphics cards die within days of being repasted with Arctic MX-4. 

I am confident that my repasting method isn't causing any issues - cleaning the die with 99% isopropyl, scraping off any hardened old paste with a plastic blade, and applying a moderate amount of paste to the die. Of course there's the occasional runoff of alcohol or crusty paste onto the PCB but I've had that happen with all sorts of hardware and it hasn't caused problems.

But is the general idea of repasting somehow bad? It seems remarkably odd that a GeForce 7950 Go GTX and ATi Radeon X1800M would both perish from normal maintenance after years of running in the high 80s-90s under load because of dry paste and clogged fans. Does running at reasonable temperatures under load once again somehow damage the BGA solder? 

I understand how stupid of a question this is, but I'd love to not have any more hard-to-replace cards die on me, if I'm doing something wrong please let me know.

A possibilty is that you applied too much pressure on the IHS-less dies when remounting the card

 

[da na]

11 minutes ago, PDifolco said:

A possibilty is that you applied too much pressure on the IHS-less dies when remounting the card

 

Possible but I really doubt it. I work with IHS-less mobile CPUs, chipsets, and soldered graphics frequently, and I've only seen these symptoms on a very particular era of high performance Dell socketed mobile GPUs. Would be a bit odd if I somehow damaged the die on just these.

The heatsink holes are numbered and I follow that order for removal and re-assembly. 

[Mark Kaine]

3 hours ago, da na said:

Odd question but I have had now two rare old graphics cards die within days of being repasted with Arctic MX-4. 

I am confident that my repasting method isn't causing any issues - cleaning the die with 99% isopropyl, scraping off any hardened old paste with a plastic blade, and applying a moderate amount of paste to the die. Of course there's the occasional runoff of alcohol or crusty paste onto the PCB but I've had that happen with all sorts of hardware and it hasn't caused problems.

But is the general idea of repasting somehow bad? It seems remarkably odd that a GeForce 7950 Go GTX and ATi Radeon X1800M would both perish from normal maintenance after years of running in the high 80s-90s under load because of dry paste and clogged fans. Does running at reasonable temperatures under load once again somehow damage the BGA solder? 

I understand how stupid of a question this is, but I'd love to not have any more hard-to-replace cards die on me, if I'm doing something wrong please let me know.

i think the main issue is probably mounting pressure  - how much is too little, how much is too much? how'd you even know?  you got the torque numbers from manufacturer? 

 

 

i doubt it's the paste,  it literally shouldn't matter as long its not conductive,  mx4 is surely better than some old dried up stuff lol

[da na]

1 minute ago, Mark Kaine said:

i think the main issue is mounting pressure  - how much is too little, how much is too much? how'd you even know?  you got the torque numbers from manufacturer? 

 

 

i doubt it's the paste,  it literally shouldn't matter as long its not conductive,  mx4 is surely better than some old dried up stuff lol

That's an interesting point, perhaps loose mounting pressure might be causing the die to separate from the board....

I'd been hand-tightening the screws, not wanting to crack the die but in so doing I very well may have applied inadequate pressure. 

[Godlygamer23]

11 minutes ago, da na said:

That's an interesting point, perhaps loose mounting pressure might be causing the die to separate from the board....

I'd been hand-tightening the screws, not wanting to crack the die but in so doing I very well may have applied inadequate pressure. 

This seems like only an issue if you're not allowing the screws to turn all the way until they stop - once they get to the end of their range of motion, it's really about the final bit of turning reasonably left that creates some kind of clamping force. 

[Godlygamer23]

14 minutes ago, Mark Kaine said:

i think the main issue is probably mounting pressure  - how much is too little, how much is too much? how'd you even know?  you got the torque numbers from manufacturer? 

Even having a set torque value would still open up questions because the method that you use to tighten the hardware matters a lot here.

 

Since you're typically compressing a spring, that creates pushback against the threads, which causes the torque driver(no matter what kind you use) to experience more resistance(and thus more torque required to turn the hardware), so the tool might read the right torque value, even if it hasn't been reached.

 

This is also my problem with using torque values - they can be misused by individuals. Torque values simply correspond to the approximate point at which the hardware has reached optimal tightness, but that doesn't mean it actually has, since other factors influence the required torque, including at which point the tool was actually applied.