[SOLVED] Replaced thermal pads on Gigabyte RTX 3080 Gaming OC and bricked it

Rangs641

Distinguished
May 7, 2013
27
0
18,530
Hi Experts

My Gigabyte RTX 3080 Gaming OC GPU's memory temp was crossing 105C consistently so I decided to replace stock thermal pads with GELID ULTIMATE 2mm. While replacing the pads I damaged the board near the bottom screw. That's the only screw that is directly on the board. Attached pics.

I replaced the pads, connected the fan cable (?!?) and put the GPU back in my system. Now when I turn on the PC, the GPU's LED and fans turn on, but there is no signal to the monitor or my LG TV. I tried display port and HDMI, and neither seem to generate a signal.

I tried this card in a different PC and same behavior!

But when I turn on the system via mobo display port, and check GPUZ and HWiINFO, it seems to detect the GPU and show all stats. Idle memory temps have improved by at least 10C - earlier it used be 60C, now 50C. The card itself seems to be working, but is just not outputting signal!

Does anyone have any suggestions? Do you guys think the damage to the board caused this problem? I'm pretty sure I have voided the warranty by damaging the board. So sending the card to gigabyte is my last option.
 
Last edited:
Solution
OK. So gigabyte won't be able to fix this even if I pay for the repair?
They have the obligation to flat out refuse repair due to damage they suspect was caused by the user. How are you going to fib your way around what happened in that first image?
They're going to see that, point the finger at you, "You damaged the gpu. We will not fix this. You get it back as is.", and you will be stuck with it... or I guess you could try to sell it for parts..?


Are you suggesting that the card cannot be repaired at all?
Quite likely, but we don't know the degree of damage.
1)The first image suggests a possible fire hazard. If the screw makes contact with exposed wiring, it could cause a fire.
2)The pads you used compared to the...

Rangs641

Distinguished
May 7, 2013
27
0
18,530
Also, I used Cooler Master MasterGel Pro thermal paste on the chip.

Initially I had applied a lot of paste. I cleaned it and applied a thin layer the second time. Made sure all the thermal pads are in place. No signal whatsoever.
 
Last edited:

Rangs641

Distinguished
May 7, 2013
27
0
18,530
How did you end up with that? There seems to be some exposed contacts on the PCB, where the screw is threading is...I think you're shorting the board.

I pulled the board apart without removing that screw and it broke. :-( I tried running the GPU without that screw, but made no difference.
 

Phaaze88

Titan
Ambassador
If there are exposed contacts in the PCB like Lutfij says, even if you somehow manage to get the gpu working again, it is now at risk of becoming a house fire like the NZXT H1 riser disaster.

If you try sending to Gigabyte, they're going to see that damage, and will do nothing other than send it back to you.

Electrical traces may have been broken too, depending on how hard those new pads were. That'll definitely kill a card.

The card is a paperweight now.
 

Rangs641

Distinguished
May 7, 2013
27
0
18,530
If there are exposed contacts in the PCB like Lutfij says, even if you somehow manage to get the gpu working again, it is now at risk of becoming a house fire like the NZXT H1 riser disaster.

If you try sending to Gigabyte, they're going to see that damage, and will do nothing other than send it back to you.

Electrical traces may have been broken too, depending on how hard those new pads were. That'll definitely kill a card.

The card is a paperweight now.
OK. So gigabyte won't be able to fix this even if I pay for the repair? Are you suggesting that the card cannot be repaired at all?
 

Phaaze88

Titan
Ambassador
OK. So gigabyte won't be able to fix this even if I pay for the repair?
They have the obligation to flat out refuse repair due to damage they suspect was caused by the user. How are you going to fib your way around what happened in that first image?
They're going to see that, point the finger at you, "You damaged the gpu. We will not fix this. You get it back as is.", and you will be stuck with it... or I guess you could try to sell it for parts..?


Are you suggesting that the card cannot be repaired at all?
Quite likely, but we don't know the degree of damage.
1)The first image suggests a possible fire hazard. If the screw makes contact with exposed wiring, it could cause a fire.
2)The pads you used compared to the original ones. Changing pads should be a last resort, but I can wave that off due to the temperatures you reported.
Replacing pads is a crapshoot, with high risks. Knowing thickness isn't enough, these things come in different degrees of hardness too, which can lead to 2 unfavorable scenarios:
A)Too soft. Pads don't make proper contact with the cooler's heatsink, and thus the memory runs hot.
B)Too firm. Pads don't have enough give and when securing the PCB to the cooler, the PCB will either warp, causing uneven mating and high gpu core thermals, or crack the PCB.

The w/mk 🐂 💩. Higher is not better here. In fact, higher is harder. I looked up those Gelid pads, and they claim 15w/mk.
If that is true, and you matched the depth of the original pads...
Then upon securing the cooler to the PCB, you cracked it.
 
Solution

Rangs641

Distinguished
May 7, 2013
27
0
18,530
They have the obligation to flat out refuse repair due to damage they suspect was caused by the user. How are you going to fib your way around what happened in that first image?
They're going to see that, point the finger at you, "You damaged the gpu. We will not fix this. You get it back as is.", and you will be stuck with it... or I guess you could try to sell it for parts..?



Quite likely, but we don't know the degree of damage.
1)The first image suggests a possible fire hazard. If the screw makes contact with exposed wiring, it could cause a fire.
2)The pads you used compared to the original ones. Changing pads should be a last resort, but I can wave that off due to the temperatures you reported.
Replacing pads is a crapshoot, with high risks. Knowing thickness isn't enough, these things come in different degrees of hardness too, which can lead to 2 unfavorable scenarios:
A)Too soft. Pads don't make proper contact with the cooler's heatsink, and thus the memory runs hot.
B)Too firm. Pads don't have enough give and when securing the PCB to the cooler, the PCB will either warp, causing uneven mating and high gpu core thermals, or crack the PCB.

The w/mk 🐂 💩. Higher is not better here. In fact, higher is harder. I looked up those Gelid pads, and they claim 15w/mk.
If that is true, and you matched the depth of the original pads...
Then upon securing the cooler to the PCB, you cracked it.
When I screwed back the 4 screws behind the gpu chip, I did hear a cracking sound. Maybe it was the pcb breaking.

This was a bad idea from the beginning and I knew it too. Sigh! Anyway, I'll go to the gigabyte service centre tomorrow and tell them what happened. If they can't fix it, then no games until Rtx 4080 is out!
 

Rangs641

Distinguished
May 7, 2013
27
0
18,530
Handed over the card to gigabyte service centre today. They saw the pcb damage near the screw and accepted it to repair under CID status. Customer damaged or something like that. They said they'll do their best to repair the card, but if they can't, they'll send it back. Fingers crossed.
 
Last edited:

Rangs641

Distinguished
May 7, 2013
27
0
18,530
To close this thread, gigabyte said they can't repair the gpu because the pcb had cracked in the middle because of incorrect thermal pad thickness. So gg! The wait begins for RTX 4000 series......
 

Phaaze88

Titan
Ambassador
Sorry to hear that.
Gigabyte should correct themselves - in a sense, you did use the correct thickness, but they were too hard. This can be offset by using thinner pads, but one can only go so thin with these things.


So yeah, next time the idea comes to mind, remember that knowing pad thickness isn't enough.