Artifacts after GPU thermal paste and pads replacement

[Siberia35]

Hi everyone,

 

I changed my thermal paste and pads on my Zotac RTX 3070.

It isn't my first GPU thermal paste replacement, but the first time I messed up.

I've done the exact same procedure as Chapter 1 on 3 others cards in the past.

 

I did overclock the card a little bit, but not that much, was perfectly stable, didn't do any big stress test or didn't touch the voltage tho.

GPU temps never did go over 80°C.

 

Chapter 1: The origin

 

I changed the GPU paste to the Thermal Grizzly Kryonaut one (12,5 W/MK)

I removed the old one with microfiber towel and a very little isopropanol alcohol.

I carefully remove all pads with metallic tweezers without touching any other metallic piece.

Then I cleaned the VRAM cells again with microfiber towel and isoprop, and dried it.

Then I added brand news one Adwits 6W/MK (which aren’t that good after benchmark, but cheap).

I measured the exact height of each original with a caliper and then cutted the exact dimension for the new pad.

 

Result: Then the card didn't even seem to boot, and was not even recognized in GPU-Z.

 

My first mistake: using microfiber towel may have caused static. I facepalmed afterwards when thinking about it (even it's so damn effective compared to paper towel) 

Second mistake: I've put a little bit too much paste, not like the Vergy McCumshot, but a little too much, which of course overwhelmed on the decoupling capacitors nearby.

Third mistake: I screwed the sockets screw a little bit too hard, not at maximum but maybe a little too hard, I guess.

 

Chapter 2: The revenge of the screws

 

Dissembled the whole thing again, cleaned all the decoupling capacitors who had thermal paste on them, with towel and isoprop.

I repasted very carefully the GPU with the exact amount of paste I needed too.

Next, I only lightly screwed the socket screws.

Double checked everything, found out that they were also pads on the other side as well, between the card and the backplate, cleaned and changed them too.

 

Result: Worked, but heavy blur and artifacts.

I don't have any artifacts in the UEFI or BIOS tho.

The card is now recognized in GPU-Z, only the memory doesn’t appear.

                 

 

 

Chapter 3: A new hope

 

Connectors seems good. I can't plug it through anything else than HDMI (I've 3x DP port + 1 HDMI), otherwise black screen.

Did a DDU in safe mode, reinstalled the drivers manually, resettled the resolution at 1920x1080, blurness away but still artifacts.

 

Result: Worked, but still artifacts.

 

 

 

Chapter 4: Endgame

 

Maybe flashing the BIOS will do?

 

Aaaaaaaaaaand there isn't any corresponding BIOS files available for my GPU, so I didn't flash it.

 

I can't even save the actual BIOS in GPUZ, when GPUZ stops the driver, black screen and it doesn't come back.

 

     

 

Conclusion:

I still have those damn artifacts. Memory not detected in GPUZ, Afterburner doesn't even detect the card.

 

I didn't have any anti static equipment at the time, but now I do, and I now only use isoprop with a strong paper towel.

I still was careful the whole time, unscrewing, dissembling, cleaning and didn't scratched the card or dropped it. I knew the risks.

 

Things I've learned so far:

-Overwhelming paste can cause trouble on nearby capacitor, even if I thought thermal paste was totally dielectric since its silicone based, turn out to not be totally true for some paste, which can cause troubles when overwhelming.

-Take static more seriously.

-Artifacts are mostly a VRAM problem, which can be caused by over screwing, lack of pads or good contacts between the pads and the radiator, corrupted BIOS and few others reasons.

 

So, what is the next step? What did I miss ?

 

Crying alone in a corner on my miserable life or repair shop ? Is it even repairable ? Are spare pieces even available at the current time ? Should I pay respect and move on ?

 

Thanks in advance.

 

PS: I've considered a Chapter 5: The Oven strikes back at 280°C during 12 mins to solve any soldering issues but I somehow doubt this will do anything since the card did perform well before the pad change. I didn't know if this trick it still relevant in 2021 with RTX cards.

[Falkentyne]

What exact cloth did you use?  Take a picture?  Were you working on carpet?  Did you have shoes and socks on ? (I always go barefoot).

 

I usually use charmin toilet paper, but I also use those cloths that come with Hotline games mouse feet sometimes (Usually directly on the GPU though, for cleaning after initial wipe), and sometimes lint free eye liners./lip gloss applicators (VERY USEFUL).  https://www.amazon.com/gp/product/B085XXXPMM/

 

[SamucaFO]

I think you screwed up the VRam chips when you screwed the screws. This a typical VRAM problem. See if it has damage in the vram chips and the memory temp, it can be temp problem because of the new pads.

[Falkentyne]

8 hours ago, Siberia35 said:

Hi everyone,

 

I changed my thermal paste and pads on my Zotac RTX 3070.

It isn't my first GPU thermal paste replacement, but the first time I messed up.

I've done the exact same procedure as Chapter 1 on 3 others cards in the past.

 

I did overclock the card a little bit, but not that much, was perfectly stable, didn't do any big stress test or didn't touch the voltage tho.

GPU temps never did go over 80°C.

 

Chapter 1: The origin

 

I changed the GPU paste to the Thermal Grizzly Kryonaut one (12,5 W/MK)

I removed the old one with microfiber towel and a very little isopropanol alcohol.

I carefully remove all pads with metallic tweezers without touching any other metallic piece.

Then I cleaned the VRAM cells again with microfiber towel and isoprop, and dried it.

Then I added brand news one Adwits 6W/MK (which aren’t that good after benchmark, but cheap).

I measured the exact height of each original with a caliper and then cutted the exact dimension for the new pad.

 

Result: Then the card didn't even seem to boot, and was not even recognized in GPU-Z.

 

My first mistake: using microfiber towel may have caused static. I facepalmed afterwards when thinking about it (even it's so damn effective compared to paper towel) 

Second mistake: I've put a little bit too much paste, not like the Vergy McCumshot, but a little too much, which of course overwhelmed on the decoupling capacitors nearby.

Third mistake: I screwed the sockets screw a little bit too hard, not at maximum but maybe a little too hard, I guess.

 

Chapter 2: The revenge of the screws

 

Dissembled the whole thing again, cleaned all the decoupling capacitors who had thermal paste on them, with towel and isoprop.

I repasted very carefully the GPU with the exact amount of paste I needed too.

Next, I only lightly screwed the socket screws.

Double checked everything, found out that they were also pads on the other side as well, between the card and the backplate, cleaned and changed them too.

 

Result: Worked, but heavy blur and artifacts.

I don't have any artifacts in the UEFI or BIOS tho.

The card is now recognized in GPU-Z, only the memory doesn’t appear.

                 

 

 

Chapter 3: A new hope

 

Connectors seems good. I can't plug it through anything else than HDMI (I've 3x DP port + 1 HDMI), otherwise black screen.

Did a DDU in safe mode, reinstalled the drivers manually, resettled the resolution at 1920x1080, blurness away but still artifacts.

 

Result: Worked, but still artifacts.

 

 

 

Chapter 4: Endgame

 

Maybe flashing the BIOS will do?

 

Aaaaaaaaaaand there isn't any corresponding BIOS files available for my GPU, so I didn't flash it.

 

I can't even save the actual BIOS in GPUZ, when GPUZ stops the driver, black screen and it doesn't come back.

 

     

 

Conclusion:

I still have those damn artifacts. Memory not detected in GPUZ, Afterburner doesn't even detect the card.

 

I didn't have any anti static equipment at the time, but now I do, and I now only use isoprop with a strong paper towel.

I still was careful the whole time, unscrewing, dissembling, cleaning and didn't scratched the card or dropped it. I knew the risks.

 

Things I've learned so far:

-Overwhelming paste can cause trouble on nearby capacitor, even if I thought thermal paste was totally dielectric since its silicone based, turn out to not be totally true for some paste, which can cause troubles when overwhelming.

-Take static more seriously.

-Artifacts are mostly a VRAM problem, which can be caused by over screwing, lack of pads or good contacts between the pads and the radiator, corrupted BIOS and few others reasons.

 

So, what is the next step? What did I miss ?

 

Crying alone in a corner on my miserable life or repair shop ? Is it even repairable ? Are spare pieces even available at the current time ? Should I pay respect and move on ?

 

Thanks in advance.

 

PS: I've considered a Chapter 5: The Oven strikes back at 280°C during 12 mins to solve any soldering issues but I somehow doubt this will do anything since the card did perform well before the pad change. I didn't know if this trick it still relevant in 2021 with RTX cards.

 

 

Card died for one of three reasons:

1) You knocked something off the PCB and didn't notice that you did.

2) BGA solder balls became loose (Core most likely, GDDR6 possibly) from overtightening and the PCB flexing and something getting severed.  Failures like this are very often solder issues.

3) Contaminants got between the VRAM contact chips or under the chips.  Give the card a complete full bath in 99% isopropyl alcohol--douse the entire card in it for several minutes, then wipe it with some lint free lip gloss applicators (I use these for cleaning VRAM and chips: https://www.amazon.com/gp/product/B085XXXPMM ) or coffee filters along with 99% isopropyl alcohol.

 

Only 3 has a chance of recovery.  1 and 2 seem more likely (still not sure how you managed to do this).

[Siberia35]

20 hours ago, Falkentyne said:

What exact cloth did you use?  Take a picture?  Were you working on carpet?  Did you have shoes and socks on ? (I always go barefoot).

 

I usually use charmin toilet paper, but I also use those cloths that come with Hotline games mouse feet sometimes (Usually directly on the GPU though, for cleaning after initial wipe), and sometimes lint free eye liners./lip gloss applicators (VERY USEFUL).  https://www.amazon.com/gp/product/B085XXXPMM/

 

Well normal cotton cloth. Nothing special, barefoot with flipflops, no carpets, was working on a plastic cutting board.

[Siberia35]

19 hours ago, SamucaFO said:

I think you screwed up the VRam chips when you screwed the screws. This a typical VRAM problem. See if it has damage in the vram chips and the memory temp, it can be temp problem because of the new pads.

I'm gonna retry with news Thermalright pad.

 

Somehow, i can't see the memory temp in GPU-Z

[Siberia35]

14 hours ago, Falkentyne said:

 

 

Card died for one of three reasons:

1) You knocked something off the PCB and didn't notice that you did.

2) BGA solder balls became loose (Core most likely, GDDR6 possibly) from overtightening and the PCB flexing and something getting severed.  Failures like this are very often solder issues.

3) Contaminants got between the VRAM contact chips or under the chips.  Give the card a complete full bath in 99% isopropyl alcohol--douse the entire card in it for several minutes, then wipe it with some lint free lip gloss applicators (I use these for cleaning VRAM and chips: https://www.amazon.com/gp/product/B085XXXPMM ) or coffee filters along with 99% isopropyl alcohol.

 

Only 3 has a chance of recovery.  1 and 2 seem more likely (still not sure how you managed to do this).

Is the oven an option for n°2 ?

 

Okay, i'm gonna try the isoprop bath. Are cotton swabs all right or does it really need to be lint free like some cotton makeup remover ?

 

Even if number 3 doesn't work, can you repair it with like, buying new VRAM chips and using a hot air gun for soldering ? (Not planning to do it myself, but is it somehow fixable by a pro ?)

[Falkentyne]

1 hour ago, Siberia35 said:

Is the oven an option for n°2 ?

 

Okay, i'm gonna try the isoprop bath. Are cotton swabs all right or does it really need to be lint free like some cotton makeup remover ?

 

Even if number 3 doesn't work, can you repair it with like, buying new VRAM chips and using a hot air gun for soldering ? (Not planning to do it myself, but is it somehow fixable by a pro ?)

No, cotton swaps are definitely NOT all right.  And no to your other answer.  Once the magic smoke comes out, it's done (How would you even know what failed and where?)

 

This is what I use.  I used this on my Out of Stock 3090 FE and other video cards (and you know I'm not going to risk that card).  Also used these to clean up LM spills on motherboards (along with iso).

Never had a problem.

 

https://www.amazon.com/gp/product/B085XXXPMM/

[darustla]

I think you killed the VRAM with static most likely, artifacting is definately a VRAM problem and I doubt your replacement pads are causing it from not showing in GPU-Z either.

 

Big F in these current times