Zotac - GTX 460 SLI, up in smoke!

[VirtualLarry]

This is a Folder's worst fear. Card meltdown. One reason that I stopped overclocking the cards, and stopped CPU folding, because the temps just get too high. High enough, that the PSU heats up from all of the hot air generated by the GTX460 and the Q9300, that I can barely hold my hand to the top of my case.

I'm working on getting a hold of a pair of Antec 100 cases, hopefully the increased airflow will help my situation.

 

[Black96ws6]

I was under the impression that the computer was sitting idle at home and and he came home and the damage was done. I have never in my life heard of a card melting at idle unless mabe it got wet. If the card got hot it would shut down auto matically, even if he was folding, which he never said he was.

I did see that, however the thing that sticks out as weird to me (and maybe it's just me), is, he said:

"I came home from work on Tuesday, 4/19, to find my system shutdown and a strong burning smell in the air."

So...you left your system on all day while you were at work.....why???

Something doesn't add up...but again, could be reading too much into it...

 

[dakU7]

I did see that, however the thing that sticks out as weird to me (and maybe it's just me), is, he said:

"I came home from work on Tuesday, 4/19, to find my system shutdown and a strong burning smell in the air."

So...you left your system on all day while you were at work.....why???

Something doesn't add up...but again, could be reading too much into it...

Why? why not? my system is on 24/7 and i shut it down usually once a week. Thats not the part where things doesn't add up, leaving a computer on is very common.

 

[thilanliyan]

So...you left your system on all day while you were at work.....why???

There are times when I leave my comp on overnight and several days too. Not uncommon.

 

[Idontcare]

There are times when I leave my comp on overnight and several days too. Not uncommon.

I've got computers (5) in my basement that have been running 24x7 for nearly 3 yrs. Plus my desktop computer.

The thing that does not add up here is that you've got multiple fail-over points in the electrical circuit.

The PSU should have failed, either by fuse blow or by internal circuit death, long before the mobo (PCIe socket) or GPU got enough amps through it to sustain a flame long enough to ignite the HSF shroud.

The shroud plastic materials themselves are spec'ed to be flame-retardant. (not that they can't be set on fire, but the fire should extinguish itself unless a fuel-source is adding to the equation)

Barring the PSU killing the power to the short-circuit, the circuit breaker for the house should have tripped and done the same.

I've shorted out lots of computer parts, not proud to admit it but it is true, and not once has my circuit breaker failed me. I killed my raid-0 I-RAM drives because the ram dimm on one touched and completed an electrical short to a PCB pin on the backside of the other (with my face about 4 inches from it at the time) and whiz-pop went both the I-ram cards and my circuit breaker.

I don't doubt this guy's cards and/or mobo killed themselves, but warranty fraud is a fact of life (see Zap's thread) and it would not be the first time that someone got a brilliant idea on how to RMA their OC'ed to death video cards...so we all have the right to be a little guarded and a little suspicious, technically we'd be naive if we weren't a little bit suspicious.

 

[Arkadrel]

The owner would have to prove that the cards died because of something that went wrong with them (specifically) and not because of something that went wrong with something around them (PSU/mobo/user-error in assembly).

He would have to prove that the root-cause was not his mobo, not his PSU, not his power cables, or the possibility that he hooked it up incorrectly (did not fully insert the GPU into the PCIe slot, etc).

He can easily prove he has non-functional GPU's, not so easy to prove that the root-cause of them no longer functioning is solely due to manufacturing defects or flaws.

Doesnt it work the otherway around?

Isnt it up to the warrenty people, to exsame stuff and figour out if faul play was at hand, and not the casuality of a hardware malfunktion?

If hes sure enough, I guess he could always take it to court.

 

[notty22]

This comment is based on the Op's psu, I noted its a Corsair 850.
This is a topic, I referred to before. Antec instructs about the dangers of single rail psu's. There is a technical reason why, a 'smart' ocp protection is not implemented on big single rail pc's.
Antec telling consumers to stay away from single rail PSUs?
A Antec rep comments in this thread.


Myth 3: A Single Rail Power Supply Is As Safe As Multi Rail Power Supply!

edit: Its a controversial topic, single rail psu's are so popular. I'm just wondering if this issue contributed to what looks like a extended burn in the ops, computer.

 

[tigersty1e]

Can you post a bare shot of the 460 culprit pcb?

Usually, these cards vrm will fail and cause burning, but from the backside shot, it looks like the mobo pcie slot caught fire. These cards draw 150 watts from the slot. If OC'd crazy high, that might be too much for the board.

Check the 24 pin connector on mobo and psu.

 

[nyker96]

from OP's pictures it looks like the motherboard chipset cooler melted. How I don't know maybe the PSU blew, I once had a PSU blew, took the MB and cpu with it. I'd say the culprit is PSU blew due to heat and when it blew it burnt out both the motherboard and the 2xgcards at the same time. What I call 'total death'. On one hand you could say the cards might be responsible probably because it drew so much power, but Zotec could also claim their cards fine just OP's case isn't well ventilated or the PSU is too weak to begin with.

 

[Elixer]

Dang, looking at those cards you are really lucky to not have had a house-fire from it.

You can say that again.
If his house burned down because of those cards, you can bet that Zotac would be paying damages.

I am not sure how you can tell who is at fault here, both the mobo and the cards suffered lots of damage...


All I can say is, good luck, OP!

 

[Texashiker]

You posted the thread on the zotac forum on April 19, 2011. Have you any response from them?

There are no replies to you post over there. Seems kinda strange for an official support forum.

Barring the PSU killing the power to the short-circuit, the circuit breaker for the house should have tripped and done the same.

Who said the damaged happened all at one time? Could it have been spread out over a period of time?

 

[OILFIELDTRASH]

If gpu makers want to not honor warranties based on overclocking they need to use a bios switch like the 6990's. If you flip the switch they aren't liable. The card makers brag about how cool they run, overclocking ability, etc.. But then if something goes wrong they don't want to honor it. I have no doubt he had these cards overclocked. What enthusiast with a more than capable power supply and sli 460's wouldn't? I overclock the piss out of my 580 (within reason) and I fully expect it to be covered under warranty should something go wrong. I don't see anything this guy did wrong. He bought enthusiast hardware and pushed em some. He wasn't volt modding his Dell tower.

 

[apoppin]

Do we yet know if it was the video cards that caused it, or not?

My own Gigabyte x58 MB malfunctioned and ran enough current through the 12V cable to melt two of the PSU's 24-pin ATX plastic connectors using GTX 560 TI SLI at stock settings on a 775W Thermaltake PSU. Not the PSU's fault nor the video cards; it was caused by the MB's malfunction. That said, i got great service from Gigabyte RMA and there are no further issues with GTX 590, GTS 480 or GTX 580 SLI and everything works great with HD 6990+6790 Tri-Fire-X3.

i now use my PSU with the damaged ATX connector for powering multiple video cards as a secondary unit.

If gpu makers want to not honor warranties based on overclocking they need to use a bios switch like the 6990's. If you flip the switch they aren't liable. The card makers brag about how cool they run, overclocking ability, etc.. But then if something goes wrong they don't want to honor it. I have no doubt he had these cards overclocked. What enthusiast with a more than capable power supply and sli 460's wouldn't? I overclock the piss out of my 580 (within reason) and I fully expect it to be covered under warranty should something go wrong. I don't see anything this guy did wrong. He bought enthusiast hardware and pushed em some. He wasn't volt modding his Dell tower.

 

[Seero]

Something isn't right about the pictures. First, burn mark doesn't start from the PCB. You have 2 cards, the lower cover of the top card is melted, but clearly it was melted from outside to inside, indicating the source is from outside. Then you have the lower card, those are plastic burns from the top card, again not from inside to outside. The location of the GPU, VRM and RAM looks perfectly fine, indicating the abnormal temp from these components can be ruled out. But if the fire didn't start from the PCB on the TOP and BOTTOM, then what causes the burn?

You said you SLI, why do you place the card back to back where you really should populated socket 1 and 3? Edit: I really mean PCIe 1 and 4, as I mistakenly think that the second PCIe was a PCI.

Suppose the origination is indeed on the video card, then it means there exists some components that were so hot that melted the cover of the top card from the back side of the second card, but there are actually nothing on the PCB located at the lower part of the card.
Picture from hardwareheaven.com

Here is the interesting part, look at the light plastic mark on the back of the second card where the GPU is, it seems to be from the plastic at the bottom and got spun to the top. Indicating the fan of the top card works. Look at the blade of the top card, the tips are also darkened, supporting my theory. That the fire didn't start off from the card, but there were sitting on top of something that melted the cover of the top card. Now look at the mobo pic

The clip of socket 3Edit, again I meant the 4th PCIe slot, not 3rd. was narrowed, indicating OP couldn't have a populated that slot during the incident. Look at the HS on the mobo, it actually matches the melt part of the top card, indicating it was extremely hot during the incident. From the picture I say the mobo temp is way over 100c during the incident, but why didn't heat got vented?

Because the back side of the 2nd card is blocked the intake of the first card. It was spinning, but not moving air!

There were 2 things that are hot, the top card and the HS on top of NB. In conjunction with the fact that intake of the top card is blocked, temperature at the surface of the NB HS gets so high that it melted the bottom of the top video card's HS's cover.

First, these video cards are not meant to be placed back to back as those ain't centrifugal fan, it require open space for the intake to work properly. Even with a blocked fan, it couldn't cause this, since the GPU will cease to function when it is too hot, but the GPU wasn't too hot at the time, I am saying this out loud but the GPU as well as the HS reached around 100c which is not enough to trigger the temperature production of the video card, but there is another source of heat, the HS off the mobo, making the bottom part of the card exceed plastic melting point. Here is something interesting i found from an old review about that particular mobo,
GA-X58A-UD5, from overclock.net.

Whenever a heatsink gets hot you know it’s doing its job, and well the gigabyte heatsinks get very hot, yet we never see Northbridge temperatures above 40C at stock and 50C overclocked, that is without supporting fans cooling everything down, but the heatsink temperature is another story, measured at stock load the heatsinks get up to about 70C and at overclock about 80-90C if you average them all out, the most heat ends up at the Northbridge cooler as its where all the heat pipes meet, and the cooler is perfect for a small 40mm or 60mm fan, yet gigabyte does not provide one, because it isn’t really necessary with the north bridge temperatures being that low. It would have been nice though. As you can see the heat moves from the block to the fins, on the UD7 on top of the fins is a waterblock, not the most effective way to design a cooling apparatus, but easy to implement because you can switch out the waterblock for the “silent hybrid” heat pipe apparatus the UD5 does not come with and the UD7 and UD9 do get.

So, it isn't the video card, PSU or the mobo, but the combination of 3 factors, 1) intake of the top card is blocked, 2) extremely hot NB HS (lack of NB fan), and 3) poor case ventilation. Look again at the third PCIe slot, there are no burning plastic, but are de-shaped, indicating that temperature was so high that it actually melted it.

Here is my guess, OP's goal is a ultra quiet PC, thus use H50 without OC. There may be 1 outtake near the mobo which is set very low because of the noise and it really doesn't move hot air. The airflow at the lower part of the case is next to completely dead except for the second card. The temp of the first card is high, near 90c, but OP didn't pay attention about it. Unfortunately, there are no thermometer at the point of ignition as therefore OP didn't even know there exists a vent problem.

 

[OCGuy]

If gpu makers want to not honor warranties based on overclocking they need to use a bios switch like the 6990's. If you flip the switch they aren't liable. The card makers brag about how cool they run, overclocking ability, etc.. But then if something goes wrong they don't want to honor it. I have no doubt he had these cards overclocked. What enthusiast with a more than capable power supply and sli 460's wouldn't? I overclock the piss out of my 580 (within reason) and I fully expect it to be covered under warranty should something go wrong. I don't see anything this guy did wrong. He bought enthusiast hardware and pushed em some. He wasn't volt modding his Dell tower.

Huh? Unless your AIB partner specifically covers overclocking in the warranty, you can't expect to be covered.

 

[OILFIELDTRASH]

Huh? Unless your AIB partner specifically covers overclocking in the warranty, you can't expect to be covered.

Sure I can. Aib's like to sit on the fence and encourage overclocking then if it something goes wrong they want to deny coverage. I paid the premium for the "voltage tweak, cooler running, 50% faster" and I plan to get my money worth out of it.

 

[yacoub]

I think Seero's on to something.

It looks more like some level of installer error with rig configuration may be involved (i.e., putting the two cards in slots right next to each other which robs the top one of proper ventilation) combined with it sitting on top of a motherboard northbridge that runs too hot and may have even initiated the failure.

 

[OILFIELDTRASH]

They are dual slot cards. It is not possible to put them in slot 1&2.

 

[DaveSimmons]

Edit: nvm, Seero raised some excellent points.

 

[3DVagabond]

If gpu makers want to not honor warranties based on overclocking they need to use a bios switch like the 6990's. If you flip the switch they aren't liable. The card makers brag about how cool they run, overclocking ability, etc.. But then if something goes wrong they don't want to honor it. I have no doubt he had these cards overclocked. What enthusiast with a more than capable power supply and sli 460's wouldn't? I overclock the piss out of my 580 (within reason) and I fully expect it to be covered under warranty should something go wrong. I don't see anything this guy did wrong. He bought enthusiast hardware and pushed em some. He wasn't volt modding his Dell tower.

The bios switch does not stop you from O/C'ing your card w/o flipping it. It can still O/C just like any other card without it can.

 

[Seero]

They are dual slot cards. It is not possible to put them in slot 1&2.

My mistake, there are 4 PCIe slot in total. Slot 1&2 are blocked by the top card, the second card is at slot 3, but it really should be at slot 4. I will make that correction.

 

[Seero]

I think Seero's on to something.

It looks more like some level of installer error with rig configuration may be involved (i.e., putting the two cards in slots right next to each other which robs the top one of proper ventilation) combined with it sitting on top of a motherboard northbridge that runs too hot and may have even initiated the failure.

Edit: nvm, Seero raised some excellent points.

Thank you and thank you.()

 

[d4a2n0k]

Sure I can. Aib's like to sit on the fence and encourage overclocking then if it something goes wrong they want to deny coverage. I paid the premium for the "voltage tweak, cooler running, 50% faster" and I plan to get my money worth out of it.

Not to derail this thread but damn straight, couldnt agree more.

I recently had a big problem with my 2600k/ASUS Maximus Extreme and I'm still not exactly sure what happened. http://forums.anandtech.com/showthread.php?t=2152116&highlight=d4a2n0k All I know is it happened whilst overclocking. In the past, if I fried something, I pay for it.

This time though, Intel decides to make a K processor version that is from the box cover "Unlocked and Unleashed" and my motherboard has a friggin LN2 switch on it so it can boot at 200 degrees below 0, I didnt lose any sleep returning the Mobo and RMA'ing the CPU.

Gotta say though, Intel, ASUS and Newegg covered me and sent brand new replacements no questions asked.

 

[Pontius Dilate]

I just want to make some observations here.

Regarding the matter of placing the cards in the 1st & 3rd PCIe x16 slots, those are the only full x16 electrical slots on the board. The 2nd & 4th PCIe x16 slots are actually x8 electrical. For full x16 links to both cards in SLI you have to use the 1st & 3rd slots. This is true for many other motherboards. Also for 3-way SLI the cards have to be sandwiched. nVidia-supplied 3-way SLI bridges require that the cards are sandwiched, and some 2-way SLI bridges require that the cards are sandwiched. Therefore while we could say that this is not an ideal configuration (especially for the top card) it is certainly a fully approved and expected configuration by the hardware designers and manufacturers.

I don't see how the southbridge heatsink could possibly get hot enough to help start a fire on the cards or the board. That just doesn't sound possible to me. The board would have shut down due to electrical problems before that happened I would think.

I'm no forensic scientist, but it looks to me like flame shot out from the lower x16 PCIe slot and burned the back of PCB of the lower card and the heatsink shroud of the upper card. A sudden massive short in the lower card's PCIe slot which caused an intense electrical fire before the computer shut down? It's not at all clear that the card itself is specifically to blame. The OP could just as easily be requesting a new board from Gigabyte.

 

[KaosFaction]

I just want to make some observations here.

Regarding the matter of placing the cards in the 1st & 3rd PCIe x16 slots, those are the only full x16 electrical slots on the board. The 2nd & 4th PCIe x16 slots are actually x8 electrical. For full x16 links to both cards in SLI you have to use the 1st & 3rd slots. This is true for many other motherboards. Also for 3-way SLI the cards have to be sandwiched. nVidia-supplied 3-way SLI bridges require that the cards are sandwiched, and some 2-way SLI bridges require that the cards are sandwiched. Therefore while we could say that this is not an ideal configuration (especially for the top card) it is certainly a fully approved and expected configuration by the hardware designers and manufacturers.

"fully approved and expected configuration" would be based on reference designs, which are centrifugal fans. These are not.

Seero's analysis seems fairly indicative of probably the actual occurrence. I initially thought it was interesting to run a 920 @ stock but use an H50 at the same time. Either the OP was stretching the truth to favor himself (what I initially believed, give the image that everything was running status quo, no overclocks or anything) or that he was running a "silent" rig (what I now believe as the evidence points a bit more in that direction). I actually see more possibility of this type of anomaly coming from cutting the corners for silence then juicing an overclock, basically because there are more fail safes in the overclock scenario.

Definitely unfortunate. However threatening to give bad PR is pretty insulting to me. If you are truly not in the wrong in any way, you shouldn't have to threaten a company to do the correct thing. Saying things like "...strong burning smell in the air. After a quick inspection..." don't meld well to your threats either. I think anyone on this board or any other would not just pop the case and unscrew the PSU if there was a strong burning smell, even if you quickly decided the source the of the smell was the PSU. I would certainly be checking all my components to ensure that the whole system was safe.

Do I think you should get a refund/RMA/replacement. Of course, but your approach is appalling.