"Fixed" a shady GTX 1080 today. It can't be that easy, right?

[lexluthermiester]

Hi to all, hope your day is going great.

I bought a GTX 1080 EVGA today, for 5 euro. The card had a reported problem of "crashing with a black screen and artifacts" under load.

I got the card in my hands and instantly tried undervolting and underclocking it, to no avail. It still crashed when I entered Furmark. I flashed the latest VBIOS for the card, that still didn't do the job. And then as a final try, I decided to look at the board.

This is what greeted me.

View attachment 400519


A stinky dusty card with what looks like cheap thermal paste and flux around the die. I cleaned up the card, the power phases, VRMs ; etc. Changed the thermal paste and tightened the back screws as needed. (They were loose before.)

And, OH WOW! IT WORKED!

View attachment 400522

It can't be that easy, right? Tell me what was the culprit? Was the memory overheating due to not making sufficient contact with the thermal pads? Was it just luck on my part?

What was it, I need suggestions.

Thanks for reading

My guess, the contact with the dies and the junk all over the place causing capacitance issues or intermittent shorts. This is another example of why it's important to keep things clean.

I ran Furmark for 20 minutes

Stop using Furmark for a decent start. No game is ever going to push the card as hard as that..

My advice is to get some IPA, ideally 90% or better but 70% will still work, and a soft brush toothbrush, soak the board in the IPA for a few minute and lightly brush off all the gunk and stuff. Then rinse it off with the same IPA and let it dry. You can help that out with a can of air or an air blower. Then put it all back together with careful use of a good thermal paste. It should be fine. The fact that the card is working and then goes wonky very likely means is needs some TLC and it's otherwise fine.

 

[Shrek]

I agree, a good scrub with isopropyl alcohol might help.

 

[rusty caterpillar]

Too* stop talking rubbish though, not like he's baking it at 400c and leaving it in the oven for 2 hours, stop moving your goalposts, obviously you have "I know better complex"

You throwing stones blindly

Is very likely that soldering on 1080 contains lead(between 40-60 %) which it should, without lead soldering is brittle, but in order to melt soldering with lead you need high temps, way over 300c, and if PCB has copper layers and ceramic layer those will spread the heat around therefore you have to increase the backing time.

In one instance I had to raise over 380 C for the heat gun to re-solder 2 x knock off SMD from a MBO. It worked but, soldering was far from good, not to say I expose those SMD to way too much heat and I had to covered some SP caps and E capacitors with steel plate for shielding but, needless to say I targeted a certain area.

My advice is to get some IPA, ideally 90% or better but 70% will still work, and a soft brush toothbrush, soak the board in the IPA for a few minute and lightly brush off all the gunk and stuff. Then rinse it off with the same IPA and let it dry. You can help that out with a can of air or an air blower. Then put it all back together with careful use of a good thermal paste. It should be fine. The fact that the card is working and then goes wonky very likely means is needs some TLC and it's otherwise fine.

I can confirm this small finger toothbrush is very good for PCB cleaning, no cotton fiber or bristles left behind

 

[lexluthermiester]

I can confirm this small finger toothbrush is very good for PCB cleaning, no cotton fiber or bristles left behind

I've seen those, never used one. They seem like they would work well.

 

[Shrek]

My advice is to get some IPA, ideally 90% or better but 70% will still work

Walmart prices (for the 32 fl oz)

• 70% is $2.98 which works out to 13.3 ¢/fl oz for the IPA content
• 91% is $3.98 which works out to 13.7 ¢/fl oz for the IPA content

so, there is no real economy to getting the 70% over the 91%.

 

[lexluthermiester]

so, there is no real economy to getting the 70% over the 91%.

Economy is not the concern, it's the water content. 90%+ has less water and thus dries faster and easier. The difference is small but can save time and effort.

 

[CheteriagalenTriagalnik]

Economy is not the concern, it's the water content. 90%+ has less water and thus dries faster and easier. The difference is small but can save time and effort.

Yes. In medicine 70% is preferred not so much because of the economy, but because having less water means a slower drying time, ergo the disinfectant effect lasts for longer. The water also helps in destroying the cell membranes of various pathogens. This has been proven to kill more germs and is the reason 70% is the standard. Also helps that it emits less fumes and is slightly cheaper.

In this case, the issue is cleaning debris/buildup and not disinfection, so 90% would be better. Even better than that? 99%. This is best used on items that are sensitive to water, such as GPUs. If you can find 99%, I'd argue this is the use case for it.

 

[lexluthermiester]

In this case, the issue is cleaning debris/buildup and not disinfection, so 90% would be better. Even better than that? 99%.

That is the point I was making. 99% is a bit on the expensive side, but for something like the situation subject of the OP, it might be worth it.

 

[Shrek]

For me it is about cost

“An engineer is someone who can do for a dime what any fool can do for a dollar.” — Arthur M. Wellington

If the 70% were cheaper, I'd use it.

Even better than that? 99%. This is best used on items that are sensitive to water, such as GPUs.

In what way are GPUs sensitive to water?

 

[koralski]

I am gonna try as a last ditch effort both of your tips.

I am going to buy a toothbrush (more accessible here in EU) and 90% IPA. Gonna scrub the card and try. If that doesn't work, I am gonna ask my dad about his hot air gun and just blast the card.

What flux should I buy? It seems there are 10000 different types. I am lost.

I got another bargain this month too. The direct competitor to this card. The Vega 64. I got a Liquid Cooled edition for about 23 bucks or so. It works perfectly without any issues. I was planning to give my girlfriend the 1080 if I managed to "fix it" fix it.

EDIT: Can someone give me the TL;DR on the Nvidia Internal Memory testing Linux ISO? How do I use it, where do i acquire it?

However, before reflowing, I'd test the VRAM

One chip looks suspect

View attachment 401000

This was gunk stuck on top of it. I know it looks like a burn mark, but it isn't one.

 

[Shrek]

I just use a rosin base liquid flux and clean it off after with isopropyl alcohol; avoid plumber's flux.

 

[CheteriagalenTriagalnik]

In what way are GPUs sensitive to water?

Well, they are electronics. If the water has not fully evaporated by the time it is plugged in, the GPU could easily be fried. Going with a more concentrated IPA solution reduces the risk of this happening by shortening the drying times. Do keep in mind some small deposit of water could still be present around some component even if the card itself looks and feels dry.

I am gonna try as a last ditch effort both of your tips.

I am going to buy a toothbrush (more accessible here in EU) and 90% IPA. Gonna scrub the card and try. If that doesn't work, I am gonna ask my dad about his hot air gun and just blast the card.

What flux should I buy? It seems there are 10000 different types. I am lost.

I got another bargain this month too. The direct competitor to this card. The Vega 64. I got a Liquid Cooled edition for about 23 bucks or so. It works perfectly without any issues. I was planning to give my girlfriend the 1080 if I managed to "fix it" fix it.

EDIT: Can someone give me the TL;DR on the Nvidia Internal Memory testing Linux ISO? How do I use it, where do i acquire it?


This was gunk stuck on top of it. I know it looks like a burn mark, but it isn't one.

Good order of operations. You have a plan, stick to it and good luck.

 

[eidairaman1]

Walmart prices (for the 32 fl oz)

• 70% is $2.98 which works out to 13.3 ¢/fl oz for the IPA content
• 91% is $3.98 which works out to 13.7 ¢/fl oz for the IPA content

so, there is no real economy to getting the 70% over the 91%.

Take caution with 90 proof, i actually damaged a case's plastic with it.

 

[lexluthermiester]

For me it is about cost

“An engineer is someone who can do for a dime what any fool can do for a dollar.” — Arthur M. Wellington

If the 70% were cheaper, I'd use it.




In what way are GPUs sensitive to water?

Fair points, but a key aspect is effectiveness of application. Lower water content means the IPA potency will be higher and the cleaning effect will be greater. 90ish% is the sweet-spot of cost and usefulness. 70% will still get the job done, as would 50%. One just has to use more of it.

Take caution with 90 proof, i actually damaged a case's plastic with it.

True. Care should be taken to avoid using 90%+ on certain plastics.

 

[DirtyDingusMcgee]

However, before reflowing, I'd test the VRAM

One chip looks suspect

View attachment 401000

yeah, that doesn't look good

 

[rusty caterpillar]

I've seen those, never used one. They seem like they would work well.

Got exactly this:


Cheaper, but I really recommend, worked better than anything I used before.

Fair points, but a key aspect is effectiveness of application. Lower water content means the IPA potency will be higher and the cleaning effect will be greater. 90ish% is the sweet-spot of cost and usefulness. 70% will still get the job done, as would 50%. One just has to use more of it.


True. Care should be taken to avoid using 90%+ on certain plastics.

Using 90 and over, never used less concentration for PCBs. If you clean PC Case plastics than I would go with 50 %. Take the 90 and diluted with distilled water.

 

[Shrek]

Take caution with 90 proof, i actually damaged a case's plastic with it.

Also avoid any solvent with acetone in it.

 

[koralski]

Bought IPA from a local vendor, cleaned the card and just plugged it in. On the first boot there were artifacts, and my mobo's VGA light was on. After the restart it seems it was okay, but I was tightening the bolt that secures the GPU. It could be from that. Don't know. Gonna run DDU and install drivers and test it out!

I am gonna run Unigine Heaven and OCCT for memory. Gonna try Furmark too. Let's see what happens.

 

[Marcus L]

Bought IPA from a local vendor, cleaned the card and just plugged it in. On the first boot there were artifacts, and my mobo's VGA light was on. After the restart it seems it was okay, but I was tightening the bolt that secures the GPU. It could be from that. Don't know. Gonna run DDU and install drivers and test it out!

I am gonna run Unigine Heaven and OCCT for memory. Gonna try Furmark too. Let's see what happens.

Don't run furmark it's a power virus for older cards of that gen

 

[koralski]

Don't run furmark it's a power virus for older cards of that gen

I've seen several people run it. What stress test should I use then?

I ran Furmark, it instantly crashed.

Restarted the system, tightened the back screws even more. Downclocked and ran Unigine Heaven without issue. What seems to be the problem here?

BTW, after cleaning it the artifacts changed, they were little colored rectangles. Not a corrupt image.

My guess is that the die has degradated beyond repair.

And now its fixed again??? Im about to go apeshit.

 

[Marcus L]

I've seen several people run it. What stress test should I use then?

I ran Furmark, it instantly crashed.

Restarted the system, tightened the back screws even more. Downclocked and ran Unigine Heaven without issue. What seems to be the problem here?

BTW, after cleaning it the artifacts changed, they were little colored rectangles. Not a corrupt image.

My guess is that the die has degradated beyond repair.

Artifacts usually indicate a vRAM issue not the GPU die, furmark will stress a GPU unlike any other benchmark or games will ever come close to, in fact back in the GTX 480 days people literally killed their cards running furmark as it used to put such a stress on them, AMD and NV had to change their drivers so that they would detect when furmark was being run so that it wouldn't allow it to use as much power and stress the GPU as much which would not happen under real world scenarios. You shouldn't force the screws to be over tightened either or you risk damaging the card further, specifically the GPU die in this case, with the spring loaded screws you should be able to feel when they are tightened enough as there will be no tension left to screw them further unless you use exessive force.

You say you ran unigine without issue then just after that say there is artifacting, which is it? if you always experience artifacting under any kind of load and/or just 2D load then it is likely a faulty vRAM chip/s and there is in all likelyhood nothing you can do in that regard short of aquiring spare identical vRAM chips and finding the faulty ones, desoldering them and replacing them, which is 99.99% something considering the GPU and the cost to you you wouldn't be doing.

Did the vRAM chips have thermal pads? were they intact? did you replace them?

You have done what most people would have done in trying to get it working, I still stand behind baking/using a heatgun as a final attempt at breathing some life into it as everything else has failed and you will only be $5 out of pocket at the end of the day!

Edit: Seeing your edit, forget benches for now, run some games and see how it performs/if it artifacts, not a fan of either furmark or kombuster as both stress the GPU to levels that games typically don't, sure they can be useful though as your GPU seems to be "on the edge" just play some games on it and see what happens, good luck

 

[koralski]

Artifacts usually indicate a vRAM issue not the GPU die, furmark will stress a GPU unlike any other benchmark or games will ever come close to, in fact back in the GTX 480 days people literally killed their cards running furmark as it used to put such a stress on them, AMD and NV had to change their drivers so that they would detect when furmark was being run so that it wouldn't allow it to use as much power and stress the GPU as much which would not happen under real world scenarios. You shouldn't force the screws to be over tightened either or you risk damaging the card further, specifically the GPU die in this case, with the spring loaded screws you should be able to feel when they are tightened enough as there will be no tension left to screw them further unless you use exessive force.

You say you ran unigine without issue then just after that say there is artifacting, which is it? if you always experience artifacting under any kind of load and/or just 2D load then it is likely a faulty vRAM chip/s and there is in all likelyhood nothing you can do in that regard short of aquiring spare identical vRAM chips and finding the faulty ones, desoldering them and replacing them, which is 99.99% something considering the GPU and the cost to you you wouldn't be doing.

Did the vRAM chips have thermal pads? were they intact? did you replace them?

You have done what most people would have done in trying to get it working, I still stand behind baking/using a heatgun as a final attempt at breathing some life into it as everything else has failed and you will only be $5 out of pocket at the end of the day!

First of all I ran Furmark. With artifacting.

Then restarted, ran Unigine, without any problems. I am currently running Kombustor, no problems. Its been 20 minutes.

The big issue is that I don't own a heatgun, and I am not comfortable baking a card in the oven I eat. I do not care if it is "safe" and such. I just can't do it.

I am gonna try and acquire a heatgun somehow. What wattage should I purchase so it would get hot enough to do the job?

I got my multimeter, if someone can tell me what to measure, and how; I'll do it.

 

[Marcus L]

First of all I ran Furmark. With artifacting.

Then restarted, ran Unigine, without any problems. I am currently running Kombustor, no problems. Its been 20 minutes.

The big issue is that I don't own a heatgun, and I am not comfortable baking a card in the oven I eat. I do not care if it is "safe" and such. I just can't do it.

I am gonna try and acquire a heatgun somehow. What wattage should I purchase so it would get hot enough to do the job?

No idea, I have baked a couple in the oven, 190-200c for 8-10 mins, likely less "toxic" than using an oven cleaner, but you do you! GL

I got my multimeter, if someone can tell me what to measure, and how; I'll do it.

I don't think your issue is a power one, you should really try and focus on one thing at a time

 

[koralski]

No idea, I have baked a couple in the oven, 190-200c for 8-10 mins, likely less "toxic" than using an oven cleaner, but you do you! GL


I don't think your issue is a power one, you should really try and focus on one thing at a time

Okay, any suggestions for a heat gun then?

You're right. I am gonna focus on one thing at a time.

 

[Marcus L]

Okay, any suggestions for a heat gun then?

You're right. I am gonna focus on one thing at a time.

No, never used one, someone further up in the thread mentioned it, maybe drop them a PM