Fried? 3080 Founders black screen after replacing thermal pads

[teskone]

Hello there. I have a 3080FE, 12 months old. I bought the 2mm Gelid Extreme pads for the front side and I managed to reduce mem temps by 20c, the gpu die got hotter so I reopened it and tried to squish the pads a bit more and repaste the gpu. After several attempts I could not manage to reduce the die and hotspot temps, eventually it stopped working and now there is no video signal and no leds, I just see the fan spinning but they are not at 100%. I tried to repaste the die again, cut the excess of thermal pads due to me maybe squishing them too much,I used a different mobo to make sure I did not damaged the pci slot, I used a different psu and even a different 12 pin cable adapter, still no signal and the gpu is cold. I inspected the pcb, I cannot see damaged or missing components, the 12pin connector is stiff.

I might have unscrewed the leaf bracket screws unevenly, causing the gpu die to bend or to allow micro air bubbles to get in? Would be the next thing to do to just send it to a specialised reballing laboratory and see what I get? The pcb is not 100% clean after all those attempts so I wonder if the kryonaut paste would be conductive enough to get a shortage. I will try to clean it a bit, try the 1,5 Gelid extreme pads instead of the 2mm that also started to crack after all these attempts, but I doubt it will solve the issue.

I also cheched a couple of resistances with the multimeter and probed the first and fourth last pins of the pcie but I get similar values to other people.

I thought I would never fail a simple task like this, but her I am, asking for your help after a rejected RMA by Nvidia and an integrated videocard in my pc in 2022.

 

[nguyen]

Hello there. I have a 3080FE, 12 months old. I bought the 2mm Gelid Extreme pads for the front side and I managed to reduce mem temps by 20c, the gpu die got hotter so I reopened it and tried to squish the pads a bit more and repaste the gpu. After several attempts I could not manage to reduce the die and hotspot temps, eventually it stopped working and now there is no video signal and no leds, I just see the fan spinning but they are not at 100%. I tried to repaste the die again, cut the excess of thermal pads due to me maybe squishing them too much,I used a different mobo to make sure I did not damaged the pci slot, I used a different psu and even a different 12 pin cable adapter, still no signal and the gpu is cold. I inspected the pcb, I cannot see damaged or missing components, the 12pin connector is stiff.

I might have unscrewed the leaf bracket screws unevenly, causing the gpu die to bend or to allow micro air bubbles to get in? Would be the next thing to do to just send it to a specialised reballing laboratory and see what I get? The pcb is not 100% clean after all those attempts so I wonder if the kryonaut paste would be conductive enough to get a shortage. I will try to clean it a bit, try the 1,5 Gelid extreme pads instead of the 2mm that also started to crack after all these attempts, but I doubt it will solve the issue.

I also cheched a couple of resistances with the multimeter and probed the first and fourth last pins of the pcie but I get similar values to other people.

I thought I would never fail a simple task like this, but her I am, asking for your help after a rejected RMA by Nvidia and an integrated videocard in my pc in 2022.

Have you tried baking it? no further harm than it's already in.

I think you crushed the BGA on the memory modules because the thermal pad would expand when it heat up under load.

Try baking your video card and replace the thermal pad with 1.5mm ones.

 

[teskone]

Have you tried baking it? no further harm than it's already in.

I think you crushed the BGA on the memory modules because the thermal pad would expand when it heat up under load.

Try baking your video card and replace the thermal pad with 1.5mm ones.

I will try to go to a specialized lab to reball the mem modules and the core, maybe the vrm too.
I got schematics for the 3080FE, I can provide those to them I guess, if the reballing wont work

 

[Count von Schwalbe]

It's pretty simple to try a reflow if you think the solder is the issue. Just warm the card up, apply flux until well coated under the dice, then pop the card in the oven until cooked through

 

[teskone]

It's pretty simple to try a reflow if you think the solder is the issue. Just warm the card up, apply flux until well coated under the dice, then pop the card in the oven until cooked through

I'm actually hoping that if I send it to a lab the could spot any other possible issue for me

 

[RJARRRPCGP]

Are the video card prices in your region still anything like '21?

 

[delshay]

I'm actually hoping that if I send it to a lab the could spot any other possible issue for me

I would check the card again, it must get warm to the touch which indicate the card is not dead. If it's still cold then there must be an issue with the power cable or the socket Do not bake the card, HDMI, display port or the power socket may melt, even thou they can withstand more heat than the fan header.

I had an issue here with my Vega card thinking it was dead as it was cold to the touch. The issue is the power socket on the card will go HR (high resistance). It's not the first time my card has tried to trick me in thinking it was dead. ...Unplug & re-plug the power cable a few times at the GPU & the PSU end if you have too.

 

[ARF]

Looks nasty

 

[xtreemchaos]

you have the right idea with taking it to a tech, you may get lucky and the tech will find something simple like a mosfet or choke cracked away from the PCB . reflowing yourself is a very hit and miss and if it works its not very often it lasts.

 

[FreedomEclipse]

Keep us updated. Would be interesting to know what an experienced certified tech finds

 

[teskone]

I would check the card again, it must get warm to the touch which indicate the card is not dead. If it's still cold then there must be an issue with the power cable or the socket Do not bake the card, HDMI, display port or the power socket may melt, even thou they can withstand more heat than the fan header.

I had an issue here with my Vega card thinking it was dead as it was cold to the touch. The issue is the power socket on the card will go HR (high resistance). It's not the first time my card has tried to trick me in thinking it was dead. ...Unplug & re-plug the power cable a few times at the GPU & the PSU end if you have too.

I will then. but I changed psu and 12pin adapter power cable.
how did you solve the HR problem?

Looks nasty

View attachment 263557

not enough pressure on the cooler I guess

Keep us updated. Would be interesting to know what an experienced certified tech finds

of course, if I can only find one.
I don't even what to look for apart from a reballing lab

 

[ARF]

not enough pressure on the cooler I guess

Oor too much pressure which resulted in those earthquake-like cracks on the dried, solid thermal paste...

 

[teskone]

Oor too much pressure which resulted in those earthquake-like cracks on the dried, solid thermal paste...

I applied fresh paste several times during those attempts, maybe the last one wasn't well spread, but I definitely got the same result with the previous repasting

 

[registertotypetostrangers]

You probably bent the PCB enough that some of the pads got ripped on the memory or perhaps other components. Thermal pads are not as easy to replace as it seems since you really need to accurately measure the thickness and compressing them to fit will just cause more issues than pads that are too thin. Its good you are going to take it to a professional repair shop. Re-flow jobs just ruin the card even more and most GPU repair shops will refuse to work on cards that have been "Baked".

 

[delshay]

I will then. but I changed psu and 12pin adapter power cable.
how did you solve the HR problem?

I just kept plugging in the power cable into the GPU socket. It took around 9 - 12 times for it to work. It must be the "sense pin" in the socket or the plug that has tarnished. You can see this, there is a difference in color, grey in color where some of the pins are lighter in color. where I have been plugging in the cable many times. The socket itself is supposed to be gold plated, not so sure on my card maybe it's I worn away. I started a thread on this where I was going to change the socket but stopped left it at that as there are no sockets available until end of year 2023. You can tell if the card is faulty because my card was cold to the touch, but I unplugged the fan & notice it was getting warm. If you think your card is cold & not getting warm stop the fans, then check to see if it's getting luke warm. .It should be getting at lease luke warm to the touch with fans unplugged.

Note if the GPU has poor contact with the heatsink, it will shut down. I tested my card with no heatsink but kept my finger on the naked GPU die, with my other hand on the main power switch ready to turn it off as soon as it heats up. It takes less than 6 seconds for the GPU to heat up, so I was quick to cut power as soon as I got a display output or before the GPU overheats, whichever came first. This is how I established my card was not faulty & not dead. This is how I found out it was the power socket at the GPU end that was playing tricks on me.

EDIT: Sorry I should have added I also had a HDMI cable issue. My Vega card had issues with certain brand of HDMI cable, not sure why this is as other GFX cards worked fine with the old cables. The second issue here is I switched to a slightly shorter higher quality HDMI cable.

 

[teskone]

Update:

I just tried the gpu without the cooler. the gpu started heating so it should still be alive, I did not wait enough to get video signal though since I was too scared after the first 4 secs.

should I now try to get a video signal or mount everything back and see what happens?

might this little dent on the pin be the problem?

 

[delshay]

Update:

I just tried the gpu without the cooler. the gpu started heating so it should still be alive, I did not wait enough to get video signal though since I was too scared after the first 4 secs.

should I now try to get a video signal or mount everything back and see what happens?

might this little dent on the pin be the problem?

You do not need to wait for the boot screen menu to show-up as some GPU die heats up very quickly. You just need to see some activity on the monitor. My TV shows 1080p signal detected well before the boot screen shows-up. You can leave your finger(s) on the die until you can no longer keep it there, then cut the power with your other hand at the mains or back of PSU. If you suffer burns to your finger(s) during this test I will not be held responsible, you do this at your own risk.

If you did not get a signal during this time, then I would check the HDMI/Display port cable has not gone faulty. ..Cables do break & can break at any time with any movement. My issue was I was getting intermittent black screen. Some user(s) will blame this on poor drivers when in-truth it can sometimes be the cable. The cable works fine on other GFX cards zero issues but my Vega card is is a little picky about cable quality especially during gaming, ie black screen for a few seconds.. ..Anyway your problem is you need to confirm your cable has not gone faulty.

EDIT: Not sure what to make out of the photo. ...checking

 

[teskone]

I tried again, this time I kept it on for 6-10 seconds but it did not seem to get hot enough to feel any pain.
I tried 2 different dp cable on 2 ports, an hdmi too. If I could get at least 1 second video signal I would hope it could be a faulty port.
Anyway, since I didn’t get it on different ports, even if one of them is defected would not it be enough to basically flatline a whole gpu, right?
So I am back thinking I knocked a capacitor even if a checked for missing smd, or it would be the damn core and I am screwed then

 

[Shrek]

Looks nasty

View attachment 263557

I agree, clean this up first and see if it works.

Is that thermal paste under the memory chip?

 

[teskone]

I agree, clean this up first and see if it works.

Is that thermal paste under the memory chip?

I cleaned everything up, but apart from me using my dirty hands it did not spread outside the core.
Thats the squished thermal pad residue

 

[delshay]

I tried again, this time I kept it on for 6-10 seconds but it did not seem to get hot enough to feel any pain.
I tried 2 different dp cable on 2 ports, an hdmi too. If I could get at least 1 second video signal I would hope it could be a faulty port.
Anyway, since I didn’t get it on different ports, even if one of them is defected would not it be enough to basically flatline a whole gpu, right?
So I am back thinking I knocked a capacitor even if a checked for missing smd, or it would be the damn core and I am screwed then

I'm not familiar with your card, but I would suspect it would get very hot within 3 - 6 seconds where you can no longer keep your finger on the die. If it's not getting very hot, then this is your main problem not the display cable.

You either created a hardware fault or there's a problem with the power cable that plugs into the GPU. I know Nvidia like to fit SMD fuses to they cards so this is the next thing I would check along with voltages.

At this stage put everything back together & take it to a repair shop to prevent any further damage.


As a side note you can fit thicker thermal pads. My card was fitter with slightly too thicker thermal pads where the PCB was slightly warped & core temperature did go up, but I already knew over time with heat cycles this will disappear & core temperature will come down. This is all down to me understanding the mounting pressure what type of thermal pads I'm dealing with. I always suggested users should read the PDF Documentation on the type of pads they are dealing with. All thermal pads are different, so it's a good idea to understand them.

 

[neatfeatguy]

I'm not familiar with your card, but I would suspect it would get very hot within 3 - 6 seconds where you can no longer keep your finger on the die. If it's not getting very hot, then this is your main problem not the display cable.

You either created a hardware fault or there's a problem with the power cable that plugs into the GPU. I know Nvidia like to fit SMD fuses to they cards so this is the next thing I would check along with voltages.

At this stage put everything back together & take it to a repair shop to prevent any further damage.


As a side note you can fit thicker thermal pads. My card was fitter with slightly too thicker thermal pads where the PCB was slightly warped & core temperature did go up, but I already knew over time with heat cycles this will disappear & core temperature will come down. This is all down to me understanding the mounting pressure what type of thermal pads I'm dealing with. I always suggested users should read the PDF Documentation on the type of pads they are dealing with. All thermal pads are different, so it's a good idea to understand them.

Speaking of thermal pads, I did have a read the other day about how some FE use different thickness of pads.

I believe it went something along the lines that the first FE cards used 2mm on the front and 3mm on the back. Then the newer versions that started at some point coming out in 2021, started using 1.5mm on the front and sometimes 2.5mm or 3mm on the back.

When you replaced the pads, did you verify the thickness?

 

[DrCR]

If you get it to a professional shop, I’ll be curious as to what company you use.

 

[teskone]

I'm not familiar with your card, but I would suspect it would get very hot within 3 - 6 seconds where you can no longer keep your finger on the die. If it's not getting very hot, then this is your main problem not the display cable.

You either created a hardware fault or there's a problem with the power cable that plugs into the GPU. I know Nvidia like to fit SMD fuses to they cards so this is the next thing I would check along with voltages.

At this stage put everything back together & take it to a repair shop to prevent any further damage.


As a side note you can fit thicker thermal pads. My card was fitter with slightly too thicker thermal pads where the PCB was slightly warped & core temperature did go up, but I already knew over time with heat cycles this will disappear & core temperature will come down. This is all down to me understanding the mounting pressure what type of thermal pads I'm dealing with. I always suggested users should read the PDF Documentation on the type of pads they are dealing with. All thermal pads are different, so it's a good idea to understand them.

I read about the possible different pads following the 2nd version of the 3080, I also think they changed the pad thickness at this point, after gpu core went up I ordered 1.5 mm pads but since it stopped working is not relevant anymore. I went with the most used brand pads and thickness, thinking I wouldn’t waste time on attempts.

If you get it to a professional shop, I’ll be curious as to what company you use.

What do you mean? I live in Italy, so I cannot rely on well known youtubers like northridgefix to fix the card. Here we have electronics repair shop or IT shops, otherwise you have to try home-based professionals.
Electronics repair shop have zero knowledge about dedicated gpu pcb, I got the schematics for the 3080 but I don’t even trust them opening the card.
IT shops rely on little server maintenance, small networks and basic repair, they will send the card to a lab they work with, would be better to go with the lab myself.
These labs are specialized in pcb trace repairs and reballing, that would be a plus compared to the general electronic repair shop so I thought it would be the best choice. i just looked for reballing labs in my city as I will be able to tell them better what I did and what to look for.

 

[delshay]

Speaking of thermal pads, I did have a read the other day about how some FE use different thickness of pads.

I believe it went something along the lines that the first FE cards used 2mm on the front and 3mm on the back. Then the newer versions that started at some point coming out in 2021, started using 1.5mm on the front and sometimes 2.5mm or 3mm on the back.

When you replaced the pads, did you verify the thickness?

No

2mm pads is all I had at that time & given I had them for over a year, I think it was best to use them before they self-harden. My pads are different to most users as once it is set it can never be disturbed in the life-time of the card. Both my Nano cards R9 & Vega should be using 1.5mm but using 2 mm pads does not matter I just cut the pads a little smaller than normal for the "inductors only". This will increase the mounting pressure on the pads, so they spread out more to the side.