Question Intentional gpu damage?!

[Joseph_87]

Bear with me, I'm getting truly annoyed... So last year I bought 3 Gtx1080ti Poseidon cards for some deep learning project concept nothing crazy. Then I sold one of them as used last September while leaving the olders for myself as a gaming card in SLi. Last month the person contacted me that the card is faulty and he wants his money back. We made an agreement that I'll give him my other card of the same type instead and send in the faulty one for rma to the retailer. He agreed, sent me the card that was totally brick and I sent him the newer leaving me with the oldest. Then I sent the faulty GPU to the retailer. Two weeks later he contacted me AGAIN to let me know that the other card is a brick and that he bought from another guy a new one and that's fine so I have to pay him back the original price.
Now, one card? I understand tho I used all of them, but the other too?! I told him that I would not repay him until I the retailer has a clear indication that the cards has not been purposely damaged but even then they'll replace the cards.
Now comes the fun part. The cards does not have any indication of being purposely damaged but none of them work... like at all. The retailer highly doubts that both cards end up dying on the same way and so do I, but we can't point out how on earth this is possible. Anyway the distributor denied the 2nd rma as quote no way that's not intentional.
I have a couple theories how he might have achieved it like 4.5v battery connected to pcie pins or anything crucial but not a capacitor as it would not leave any obvious sign but should be enough of damaging certain components that expect 3.x volts.
We're just playing in the dark and try to prove an intentional damage. Note that the 3rd card (while have a slight issue when not being elevated) is completely fine.
I cannot send the cards to the manufacturer because the warranty is only valid through the retailer in my country.
The other possibility I can think of is manually stopping the fan while using afterburner for example and let it still run for a couple hours, but shouldn't that leave an obvious sign? Getting out of ideas...

 

[Metal Messiah.]

I'm not quite sure what could be the reason, but have you asked the BUYER, which Power supply model he was using with the 1080 Ti GPU ? There might be some fault on his system as well, if those cards previously worked fine on your RIG, as you claim.

Else, he may be lying to you, though this is just my vague guess....But even if the cards aren't showing any indication of being damaged, there might be some internal damage, VRAM corruption or a capacitor issue...I'm not sure though.

Have you asked the buyer, what does he really mean by "faulty" ? Was he able to boot his PC, and play games ? Was he getting monitor signal ? Did he plug in all the PCI-e connectors properly, 6/8-PIN as applicable, assuming he is having a decent Power supply unit. ? When did the card die ? Was he able to install the GPU drivers ?

Or, the card didn't get past the POST BIOS screen ?

 

[13thmonkey]

sold in sept, faulty in feb, sorry but that's tough luck in the secondhand market, he could RMA, you should have walked.

 

[USAFRet]

He had it for 6 months, and now wants a refund? Twice?
No way.
Second hand, as it, where is, no warranty.

 

[cryoburner]

One could probably just short some pins on the board with a screwdriver or something to break the card. Or it might just be that his system has some serious power supply issue. It does sound a bit suspicious if he supposedly bought a new card in between the two failures that is working fine though.

As for paying back the original price, it sounds like he had the card for almost half a year and was likely using it for months prior to the initial failure, so it apparently wasn't DOA. If the card was sold as used, unless you had some agreement at the time of purchase to guarantee it against issues for an extended period, then you are likely not under any obligation to provide a refund.

Was this sold on an online marketplace, or locally?

 

[Joseph_87]

Actually sold one in Sept then it was faulty at the end of Feb. I gave him a completely fine card that I have been using then mid March bamm faulty again.
While here I am using the third one without any similar symptoms. Just the standard poseidon symptoms that is the card sags it overheats...
I don't think it's his power supply as that should have some traces on the pcb like popped capacitors or vrms or something. Overheating again, something on the pcb...
Shorting the board while it's on... Kinda dangerous for himself too and needs some luck for it not to leave traces...
The retailer asked the distributor to send the card to the manufacturer for proper diagnosing... On my expense obviously. Will see what's gonna come up, but idk if they could say anything else....

Plus info, I tried out both cards in my systems and none of them work. Fans spin up then stop as it should but that's about it. No signal. By using another gpu the card is not detected I device manager. Like nothing is plugged in...

 

[4745454b]

My guess is he just pushed them too hard. Either with a weak PSU or just flat out OCed them too far. It can die leaving no visable evidence. If you could handle the RMA for him and let hi use a second card while that happens that's great. But when he comes back a month later saying his second card died, too bad. That's the problem with second hand stuff. Get the first card back from RMA and keep it It's his problem with HIS card. He needs to find out what's wrong and how to stop it from happening again.

 

[ElectrO_90]

Stop being a good samaritan. He is now trying to take advantage.
6 months is more than enough for a second hand GPU.
Just send him the receipt of the one he has and tell him to claim RMA himself.

 

[cryoburner]

If one were performing a scam, it could also be that they bought another known defective card of the same model for cheap, and then swapped it with the one you sold them, hoping to get a refund, while still keeping the good card. You could note the serial number before sending the card to them, and seeing whether it matches upon return, though if that's a on a sticker, or on the cooler, then it could potentially get swapped around. Putting a tiny mark on the PCB with a permanent marker before sending it to them might help you verify that it's the same card if you ever were to send them a working card again, such as if the RMA for the second card managed to go through.

 

[Joseph_87]

Nah the serial number is the correct one that I am sure about. I did failed ocs in the past but never killed any components with it. Even when I got black screen I could remove the overclock from the card using builtin gppu for signal. But here... both cards I gave him are flat out dead. One heading to Asus for diagnostics and another rma denied by the distributor...

 

[4745454b]

A failed OC is one thing. Putting a experimental bios in there and running it for months with a higher than should be used Vcore could kill it. It could have happened with the first one. He didn't run the second one for months, but he could have tried something similar. When selling used I usually allow a 7-10day window for people to make sure it will work for them. But if someone came back after 6mo and said it died I'd ask what they wanted me to do about it. That's way to long for me to cover under a "warranty". I could see helping them with an RMA if it's in my name or I had the reciept. But I would have never sent another card while we waited for it to come back

 

[Joseph_87]

Interesting idea. I guess the manufacturer can diagnose that? That the BIOS has been tampered with? We'll see. The retailer had another idea of static electricity but we agreed that shorting the gpu while on sounds dangerous and could damage the mobo too. Well we'll see what comes back from Asus, at least we have another card to play with.

 

[Joseph_87]

Well, here's the update. Nobody knows and the manufacturer replaced the GPU with an equal one so at least I have no loss. Funny thing is he took back the replacement card with the EXACT same problem as the previous three to the retailer and demanded his money back. I think after this, I'm safe to say that either his PSU bricks the GPUs or he's doing it in some smart way that even the manufacturer couldn't detect.

 

[iMatty]

I would've walked away since beginning, killing two cards that were working perfectly fine then he is doing it intentionally.

 

[Joseph_87]

Yeah but that was the problem. Not a single person could prove it. Not even ASUS themselves. A faulty bios flash should have been diagnosed according to asus. Frying the GPU with a faulty PSU also should have left some traces but they didn't state what exactly and tbh I'm not even interested. I have a tip but I'm not breaking my last gtx1080ti just to try, maybe if I have a scrap around I'll play along with it. I'm pretty sure that a 9V battery like THIS ONE and two wires randomly pushed under the backplate, or heck even on the pcie contacts could brick a gpu. What I'm shocked tho that a person could do this to 2x POSEIDON 1080Ti's.... geez sure the cards have their own faults but comon... these are still awesome... and not only he did but he did it and not even once left a single trace. Not a single popped capacitor.

 

[iMatty]

If you get that card RMA'ed then just keep it for your self and don't pay him anything, he came after long time saying a card is faulty is not your fault.

 

[Joseph_87]

the retailer already exchanged the killed gtx1080ti for an rtx2080 plus paid the difference in cash. So he got a brand new card + nearly half the price of another brand new rtx2070 which silenced him. I think anyone would be silenced if he got a brand new card + nearly the same cash that you bought the used one for... He got the win in this case, but admit he is smart. Breaking a GPU without anyone finding out how... Literally without signs, and not even asus can clearly state whether the gpu is trully faulty or it has been tampered with. And by us, we can't remove the warranty stickers without actually loosing the warranty, that was also intact. Interesting topic, I think I'm going to purchase a GT710 and mess around with it to get any similar results or having power (both power LEDs on the GPU are white fans are spinning, GPU gets warm but no signal on anything). I'll definitely give this 9V battery a shot to see if it does anything.

 

[Joseph_87]

So, I bought two cheap used GT710 without warranty for just fooling around with it. Originally, it was working, I could kill it. Yeah, you can blow up capacitors with a single 9V battery pretty easily, but that leaves a nice clean evidence of the damage, and we had no such thing on the 1080ti -s. On the other one 710 I was more lucky. I think I hit the bios IC, small smoke no visible evidence of burnt up component but windows won't recognize the card anymore. I checked under 10x zoom, and still couldn't say that there's anything wrong with that IC. I wonder if the 1080ti-s had the same fate, but then I wonder; how come ASUS themselves couldn't find that? I mean they certainly could replace the bios chips on the 1080ti and give it back to him so that he can fool around more and get caught. Anyway, as both of the 710s are dead now, I think I'll donate them to one of my friends who might be able to use components off those.
P.S.: My original assumption of randomly attaching the 9V battery to the pcie pinouts made nothing at all. Like nothing happened, tho I expected that it would do the most damage as that's the data transfer between the RAM <-> GPU - VRAM but apparently that well protected as it seems.

 

[4745454b]

Somehow I doubt he was actually trying to destroy it. It makes for more sense that he just OC'd it and drove it too hard. TBH I think you are worrying about this way to much.

 

[Joseph_87]

Not that I worry, but became interested. OC'ing I have some doubts. Asus can read the bios of the card so if it was pushed too hard there would have been a trace. if it was indeed oc'd he should have needed to flash the bios which would have been obvious again and I guess the would be gone.

Yeah now this is just for my curiosity. Let's face it. I sold a completely fine 1080ti poseidon in September. He clearly demanded money back in February. Instead of giving him money, I replaced the card and sent the faulty for rma myself. Then just 2 weeks later again give back money. Not that this card is faulty or something. No. Money money moneyy as that song says. And I've been using that card up till that point. Rock solid!
Both cards same symptoms. Fans spin up, heat is there, windows won't recognize them. But according to Asus, not the BIOS. And no trace of tampering. This made me curious thus I bought the 2 cheap 710s.

In the end the retailer replaced his faulty card with an RTX 2080 but since that's a lot cheaper he also got half the price of an RTX 2070 in cash. Win win scenario at least for him.

 

[13thmonkey]

Not that I worry, but became interested. OC'ing I have some doubts. Asus can read the bios of the card so if it was pushed too hard there would have been a trace. if it was indeed oc'd he should have needed to flash the bios which would have been obvious again and I guess the would be gone.

Yeah now this is just for my curiosity. Let's face it. I sold a completely fine 1080ti poseidon in September. He clearly demanded money back in February. Instead of giving him money, I replaced the card and sent the faulty for rma myself. Then just 2 weeks later again give back money. Not that this card is faulty or something. No. Money money moneyy as that song says. And I've been using that card up till that point. Rock solid!
Both cards same symptoms. Fans spin up, heat is there, windows won't recognize them. But according to Asus, not the BIOS. And no trace of tampering. This made me curious thus I bought the 2 cheap 710s.

In the end the retailer replaced his faulty card with an RTX 2080 but since that's a lot cheaper he also got half the price of an RTX 2070 in cash. Win win scenario at least for him.

I don't think ocing leaves a trace in bios, and you don't have to flash the bios to oc.