My GTX 1080 has died

Trace6x

n00b
Joined
May 11, 2021
Messages
10
To cut a long story short, I think my 1080 has died, it lights up, but I don't get any output to a monitor nor do the fans spin.
I've confirmed it doesn't work in a different machine either, and confirmed that the rest of the components are working properly. I was experiencing crashes under load before it failed outright.
It's out of warranty and I'm out of ideas! Considering the current market it's a bit of a bummer for it die now.

Someone over on LTT forum suggested I should post here and at RazorWind directly, if you or anyone else had any ideas/suggestions I'd be grateful to hear them! I own a multi meter and a soldering iron so I could follow some instructions to check various compoents but I'm not sure what I'd be looking for.
Thanks!
 
What model of 1080 Ti is it? Can you post some photos of it, particularly the back of the card?

Edit: Oh, you said it's a non-Ti.
 
Last edited:
Does the 2nd machine you tested in have a powerful enough PSU?
Is there a QR code (on the mobo) or beeps (needs an internal speaker) when you power up?
Does the original machine function ok with onboard video or another gfx card. If so which card?
 
bunch of smokers here! i took mine apart when it died, i thought, and it was paste that had dried out for brick making. cleaned up the gpu and reapplied paste. it is now fully functional and works like i just got it out of the box.
 
What model of 1080 Ti is it? Can you post some photos of it, particularly the back of the card?

Edit: Oh, you said it's a non-Ti.
I've attached some pictures! It's an msi gtx 1080 gaming x 8GB, non ti


Does the 2nd machine you tested in have a powerful enough PSU?
Is there a QR code (on the mobo) or beeps (needs an internal speaker) when you power up?
Does the original machine function ok with onboard video or another gfx card. If so which card?
Yeah the second machine has a 750w PSU, no beeps or QR code, the machine actually boots up I can access it fine remotely, the original machine functions fine with the onboard graphics and my gtx 970.
 

Attachments

  • PXL_20210511_214923592.jpg
    PXL_20210511_214923592.jpg
    306.2 KB · Views: 0
  • PXL_20210511_215039481.jpg
    PXL_20210511_215039481.jpg
    670.5 KB · Views: 0
Yeah the second machine has a 750w PSU, no beeps or QR code, the machine actually boots up I can access it fine remotely, the original machine functions fine with the onboard graphics and my gtx 970.
My next suggestion is as DrDoU said, repaste it and cross your fingers thats all it needs.

Before applying the paste, after the card is disassembled, check to see if there are any darkened/burned areas, especially near the power in socket and the PCI-e slot.
Check both sides of the card.
(The photos appear ok, but the eye can see more)
 
Last edited:
Shhh, we only tell our favourites the best technique!
In case you haven't figured it out, OP, this is a meme. We had a new user come on here a couple of weeks ago, asking for help after he had apparently put a graphics card in a microwave. It appeared to have gotten so hot that the copper traces delaminated from the PCB. Obviously, you should not put a graphics card in a microwave.

Now, let's take a few resistance measurements. Put your meter in ohms mode, and take a measurement between each of the points I've marked and a ground, such as the housing of the DVI connector, or one of the through-plated screw holes.
front_resistance.jpgback_resistance.jpg

Here's what you should expect to see:
A: > 10K
B: > 10K
C: Not sure - maybe 5K?
D: > 10K
E: Not sure, maybe 10K?
F: ~350
G: 50-150
H: 0.5

If your meter isn't a fancy one, you may have to adjust the range to get a good reading. Don't worry right now if your numbers don't match mine - just take the measurements and report back to us.

Also, take measurements at each of the terminals of that little four terminal thing above F.
 
In case you haven't figured it out, OP, this is a meme. We had a new user come on here a couple of weeks ago, asking for help after he had apparently put a graphics card in a microwave. It appeared to have gotten so hot that the copper traces delaminated from the PCB. Obviously, you should not put a graphics card in a microwave.

Now, let's take a few resistance measurements. Put your meter in ohms mode, and take a measurement between each of the points I've marked and a ground, such as the housing of the DVI connector, or one of the through-plated screw holes.
View attachment 355552View attachment 355551

Here's what you should expect to see:
A: > 10K
B: > 10K
C: Not sure - maybe 5K?
D: > 10K
E: Not sure, maybe 10K?
F: ~350
G: 50-150
H: 0.5

If your meter isn't a fancy one, you may have to adjust the range to get a good reading. Don't worry right now if your numbers don't match mine - just take the measurements and report back to us.

Also, take measurements at each of the terminals of that little four terminal thing above F.
Party Pooper :(
 
Have you replaced all the thermal pads yet? Repasting (with tons of thermal paste) usually revives cards for me that are old and just have caked paste, but a lot of them also have old, crappy thermal pads that need replacement as well.
 
My next suggestion is as DrDoU said, repaste it and cross your fingers thats all it needs.

Before applying the paste, after the card is disassembled, check to see if there are any darkened/burned areas, especially near the power in socket and the PCI-e slot.
Check both sides of the card.
(The photos appear ok, but the eye can see more)
I'll try re-pasting, but surely the card should boot up without a heatsink on anyway?


In case you haven't figured it out, OP, this is a meme. We had a new user come on here a couple of weeks ago, asking for help after he had apparently put a graphics card in a microwave. It appeared to have gotten so hot that the copper traces delaminated from the PCB. Obviously, you should not put a graphics card in a microwave.

Now, let's take a few resistance measurements. Put your meter in ohms mode, and take a measurement between each of the points I've marked and a ground, such as the housing of the DVI connector, or one of the through-plated screw holes.
View attachment 355552View attachment 355551

Here's what you should expect to see:
A: > 10K
B: > 10K
C: Not sure - maybe 5K?
D: > 10K
E: Not sure, maybe 10K?
F: ~350
G: 50-150
H: 0.5

If your meter isn't a fancy one, you may have to adjust the range to get a good reading. Don't worry right now if your numbers don't match mine - just take the measurements and report back to us.

Also, take measurements at each of the terminals of that little four terminal thing above F.
Yeah I guessed! Poor sod!
Sorry I forgot to take the front plate off (let me know if you need better photos, google is gimping the quality)
Also my multimeter is super old, hope I'm doing this right, red to ground black to the contact?

Probably also worth noting I put electrical tape around the die to protect from liquid metal thermal paste leaking a few years back, the liquid metal was crap so I switched to regular paste but didn't remove the tape, it definetly was burnt on since then hence the dark spots around the die.

A: 50k (was rising slowly)
B: 7.5k also rising
C: virtually no reading at all on this one
D: 49k
E: 64.5 (without the K)
F: 10.16k
G: 0.4, tough to get a good reading on this one
H: 0.5

Wasn't sure which one was the 4 terminal thing so I did both:
1:
prong 1: 332
prong 2: 456
prong 3: 11.55k
prong 4 (right): 455

2:
Prong 1: 18.04k
prong 2: 18.01k
prong 3: 18.03k
prong 4: No reading at all

Thanks for the help all I really appriciate it! (including the microwave advice ;)
 

Attachments

  • PXL_20210512_094139696.jpg
    PXL_20210512_094139696.jpg
    325.9 KB · Views: 0
  • PXL_20210512_095410183.jpg
    PXL_20210512_095410183.jpg
    396.8 KB · Views: 0
  • 4 terminal.jpg
    4 terminal.jpg
    403.5 KB · Views: 0
Last edited:
That would be instant magic smoke.
These are not like CPUs where you have to manually fit the cooler.
There is no protection.
Ah good job I didn't try that! I'll give it a shot with some new paste.
Edit: no luck, applied new paste and no change, gpu lights up but no spin on the fans and no display :(
 
Last edited:
Ah good job I didn't try that! I'll give it a shot with some new paste.
Edit: no luck, applied new paste and no change, gpu lights up but no spin on the fans and no display :(
Best see what Razorwind comes up with.
 
That would be instant magic smoke.
These are not like CPUs where you have to manually fit the cooler.
There is no protection.
Actually no. You wouldn't want it to start and run for very long, but you'd be amazed at how tough the actual GPU itself is, and that's the only component on the card that absolutely needs a heatsink.

I'll try re-pasting, but surely the card should boot up without a heatsink on anyway?



Yeah I guessed! Poor sod!
Sorry I forgot to take the front plate off (let me know if you need better photos, google is gimping the quality)
Also my multimeter is super old, hope I'm doing this right, red to ground black to the contact?

Probably also worth noting I put electrical tape around the die to protect from liquid metal thermal paste leaking a few years back, the liquid metal was crap so I switched to regular paste but didn't remove the tape, it definetly was burnt on since then hence the dark spots around the die.

A: 50k (was rising slowly)
B: 7.5k also rising
C: virtually no reading at all on this one
D: 49k
E: 64.5 (without the K)
F: 10.16k
G: 0.4, tough to get a good reading on this one
H: 0.5

Wasn't sure which one was the 4 terminal thing so I did both:
1:
prong 1: 332
prong 2: 456
prong 3: 11.55k
prong 4 (right): 455

2:
Prong 1: 18.04k
prong 2: 18.01k
prong 3: 18.03k
prong 4: No reading at all

Thanks for the help all I really appriciate it! (including the microwave advice ;)
When you say "no reading at all" do you mean the meter said something like "OL" or do you mean zero? When taking resistance measurements, it doesn't really matter which way the probes go, but for most other measurements, it does. The red goes to the positive side of the circuit, and the black goes to the negative, or ground.

Your readings all look OK, except for G, which looks like it may be in short. I don't have a 1080 to compare to handy, but you should have somewhere in the tens of ohms on that circuit. If you really have less than 1, that's no good. The good news is that, if we can find and clear the short, there's a good chance the card would work. The bad news is that finding the short is likely to be a pain in the ass, and you'll probably need hot air to remove components without destroying the board.

Do you have a bench power supply, or access to hot air?
 
Actually no. You wouldn't want it to start and run for very long, but you'd be amazed at how tough the actual GPU itself is, and that's the only component on the card that absolutely needs a heatsink.


When you say "no reading at all" do you mean the meter said something like "OL" or do you mean zero? When taking resistance measurements, it doesn't really matter which way the probes go, but for most other measurements, it does. The red goes to the positive side of the circuit, and the black goes to the negative, or ground.

Your readings all look OK, except for G, which looks like it may be in short. I don't have a 1080 to compare to handy, but you should have somewhere in the tens of ohms on that circuit. If you really have less than 1, that's no good. The good news is that, if we can find and clear the short, there's a good chance the card would work. The bad news is that finding the short is likely to be a pain in the ass, and you'll probably need hot air to remove components without destroying the board.

Do you have a bench power supply, or access to hot air?
Sorry I got a reading of zero, good job the probe orientation didn't matter!
Ah bummer, did another reading and it's definetly 0.04. But hey finding it is good news!

Not to hand regarding the power supply/hot air but I can either get access to both at work, or purchase a hot air gun off amazon if that helps.
 
Sorry I got a reading of zero, good job the probe orientation didn't matter!
Ah bummer, did another reading and it's definetly 0.04. But hey finding it is good news!

Not to hand regarding the power supply/hot air but I can either get access to both at work, or purchase a hot air gun off amazon if that helps.
Ok, next step is to check the measurements on C and G one more time, so we can be sure that what you reported is correct. You shouldn't have zero at any of the locations I marked, so if you do, that means you have a failure of some kind in that circuit. C and G aren't really related, so that suggests that you may actually have two shorts, which is kind of unusual.

Normally, next step would be to use current injection into the shorted circuits to identify which components are bad. Unfortunately, the gddr5x memory on this card can apparently be damaged by doing that. I think my next step instead would be to check the gate pins on the relevant mosfets to see if any of them are also shorted to ground. If you find any that are, you change out that mosfet package. Maybe also check for any visibly cracked ceramic capacitors.

Can you post some closeup photos of the areas I've marked in red, here?

front_suspect_areas.jpg

The memory on GDDR5X cards has a secondary 1.8V rail. It's possible that's what the buck converter on the far right is, and if that's the case, and both are shorted, that strongly suggests a failure on the GPU die.

In which case you should sell this card on ebay as "parts only." Dead ones sell for so much these days, it almost doesn't make sense trying to repair them.
 
Ok, next step is to check the measurements on C and G one more time, so we can be sure that what you reported is correct. You shouldn't have zero at any of the locations I marked, so if you do, that means you have a failure of some kind in that circuit. C and G aren't really related, so that suggests that you may actually have two shorts, which is kind of unusual.

Normally, next step would be to use current injection into the shorted circuits to identify which components are bad. Unfortunately, the gddr5x memory on this card can apparently be damaged by doing that. I think my next step instead would be to check the gate pins on the relevant mosfets to see if any of them are also shorted to ground. If you find any that are, you change out that mosfet package. Maybe also check for any visibly cracked ceramic capacitors.

Can you post some closeup photos of the areas I've marked in red, here?

View attachment 356133

The memory on GDDR5X cards has a secondary 1.8V rail. It's possible that's what the buck converter on the far right is, and if that's the case, and both are shorted, that strongly suggests a failure on the GPU die.

In which case you should sell this card on ebay as "parts only." Dead ones sell for so much these days, it almost doesn't make sense trying to repair them.
Couldn't agree more, the "for parts" prices of some of these GPUs right now are insane.
 
Couldn't agree more, the "for parts" prices of some of these GPUs right now are insane.
I always wonder who's buying them. Given the time and effort it takes to fix them, I can't imagine it making financial sense to buy dead ones and repair them for the purpose of mining if you also have to pay that much for them up front.

I guess maybe if you're unscrupulous, you could buy them, fix them and flip them as working for huge money right now?
 
I always wonder who's buying them. Given the time and effort it takes to fix them, I can't imagine it making financial sense to buy dead ones and repair them for the purpose of mining if you also have to pay that much for them up front.

I guess maybe if you're unscrupulous, you could buy them, fix them and flip them as working for huge money right now?
Who knows, the kind of nonsense that goes on in ebay everyday is sad. I mean they run the same scam almost every day on there (ridiculously low priced GPUs/CPUs) and people seem to keep buying them.
 
Is a paid repair from MSI even an option I wonder? Even a 1080 non-ti is still probably worth 3 or 4 hundred in this market. A $100-200 repair bill still puts it in the plus column.
 
Is a paid repair from MSI even an option I wonder? Even a 1080 non-ti is still probably worth 3 or 4 hundred in this market. A $100-200 repair bill still puts it in the plus column.
Not really. Unless they try to warranty it I don't think they will repair cards. And no proof of purchase will make warranty repair difficult.
 
Is a paid repair from MSI even an option I wonder? Even a 1080 non-ti is still probably worth 3 or 4 hundred in this market. A $100-200 repair bill still puts it in the plus column.
Generally speaking they do not repair cards themselves. There are many third party services that do these types of repair. I'd do a quick search and see what you come up with. You may be surprised. Some people offer services relatively cheap if the repair is not super time consuming.
 
All I can say is, good luck OP - I feel for you. This would suck any normal year but it's even shittier because of the bullshit scalpers/miners. I hope it works out for you - I've got a 1080 TI and I might just repaste it just as a precaution, it's certainly been long enough
 
Ok, next step is to check the measurements on C and G one more time, so we can be sure that what you reported is correct. You shouldn't have zero at any of the locations I marked, so if you do, that means you have a failure of some kind in that circuit. C and G aren't really related, so that suggests that you may actually have two shorts, which is kind of unusual.

Normally, next step would be to use current injection into the shorted circuits to identify which components are bad. Unfortunately, the gddr5x memory on this card can apparently be damaged by doing that. I think my next step instead would be to check the gate pins on the relevant mosfets to see if any of them are also shorted to ground. If you find any that are, you change out that mosfet package. Maybe also check for any visibly cracked ceramic capacitors.

Can you post some closeup photos of the areas I've marked in red, here?

View attachment 356133

The memory on GDDR5X cards has a secondary 1.8V rail. It's possible that's what the buck converter on the far right is, and if that's the case, and both are shorted, that strongly suggests a failure on the GPU die.

In which case you should sell this card on ebay as "parts only." Dead ones sell for so much these days, it almost doesn't make sense trying to repair them.
Upon recheck of C and D:
C: 0.06
G: 0.05

Regarding the gate pins, you'll have to walk me through checking those, not sure what they are I'm afraid! I couldn't see any cracked or damaged capacitors but I'm not 100% sure.
I've attached some close ups.
Good to know, I'm pretty certain I was having issues with the memory before the card failed outright, I was getting artifacting when switching refresh rates. I'll sell it on if there's nothing else I can do!
Is a paid repair from MSI even an option I wonder? Even a 1080 non-ti is still probably worth 3 or 4 hundred in this market. A $100-200 repair bill still puts it in the plus column.
I did contact MSI for that reason, they told me to contact Amazon. I don't think it'll go anywhere as it's out of warranty and I bought it second hand anyway. Here's a fun bonus, the day I decided to contact MSI and find out if the card was under warranty (because it was artifacting when switching refresh rates) was the day after the warranty expired.
 

Attachments

  • PXL_20210514_073128025.jpg
    PXL_20210514_073128025.jpg
    680.1 KB · Views: 0
  • PXL_20210514_073205878.jpg
    PXL_20210514_073205878.jpg
    620.5 KB · Views: 0
  • PXL_20210514_073218015.jpg
    PXL_20210514_073218015.jpg
    666.6 KB · Views: 0
  • PXL_20210514_073231774.jpg
    PXL_20210514_073231774.jpg
    676.4 KB · Views: 0
card looks kinda dirty.....Maybe a run thru the dish washer? :D And then bake it dry
 
Back
Top