Advanced search

Message boards : Graphics cards (GPUs) : "Invisible" Graphics Card (780ti)

Author Message
valterc
Send message
Joined: 21 Jun 10
Posts: 21
Credit: 6,161,484,672
RAC: 4,196,238
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 41500 - Posted: 7 Jul 2015 | 14:04:39 UTC
Last modified: 7 Jul 2015 | 14:06:57 UTC

Some time ago I bought four 780ti graphic cards, same brand, not reference. All of them died exactly the same way, I was able to get a replacement for the first three (another brand) but not for the last one (out of warranty). I just would like to tell you 'how' they died, in a very strange way, at least for me:

At a certain time, after months of continuous BOINC running 24/7 without any kind of problem, no temperature or dirt issues, the cards suddenly went into 'invisible' mode.

The host computer simply were not able to detect it, booting (without issues) like no graphics card were available. No noise, fan spinning or lights on the card. I tried to reseat it, moved it to another PCI slot, moved it to another computer, without success, another card inserted in the same slot works perfectly. The host were four different computers, both Linux and Windows. I removed the cooler from the last one looking for strange things on the pcb like burns or fat capacitors or strange smells, nothing at all.

Any hints or comments about this?
Thanks in advance

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2343
Credit: 16,201,255,749
RAC: 7,520
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 41501 - Posted: 7 Jul 2015 | 17:35:30 UTC - in response to Message 41500.
Last modified: 7 Jul 2015 | 17:51:51 UTC

My GTX690 died the same way. Later I've found the two 12V pins on the 24-pinned MB power connector was burnt. These 2 pins power the 4 PCIe slots and the memory and the chipset. The latter two are negligible when there are 4 high-end GPUs installed. The total power consumption of 4 high-end cards from the PCIe slots could be 4*75W at worst case. There is only that two pins on the 24-pin power connector which have to supply that much power. That is 300W+ combined, or 25A@12V. 25A on two pins is simply too much, so it is quite advisable to choose a MB which has extra PCIe power connector(s) to power 4 high-end GPUs in the long term.
I suggest you to check the 24-pin power connector on your motherboard & PSU for burn marks on the 12V (yellow cable) pins. If the PSU is a modular one then check both ends of that cable.

eXaPower
Send message
Joined: 25 Sep 13
Posts: 293
Credit: 1,897,601,978
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwat
Message 41502 - Posted: 7 Jul 2015 | 21:10:25 UTC
Last modified: 7 Jul 2015 | 21:28:41 UTC

The total power consumption of 4 high-end cards from the PCIe slots could be 4*75W at worst case. There is only that two pins on the 24-pin power connector which have to supply that much power. That is 300W+ combined, or 25A@12V. 25A on two pins is simply too much, so it is quite advisable to choose a MB which has extra PCIe power connector(s) to power 4 high-end GPUs in the long term.

Here's an example of extra 6 pin and (4pin) connector MB. (located in between the pcix1 slot and left side VRM cooling element. 4 pin next to the 8pin.) Some boards have an extra 8pin instead of a 6 pin near PCIe. MB also have both an extra 8 and 6 pin.





valterc
Send message
Joined: 21 Jun 10
Posts: 21
Credit: 6,161,484,672
RAC: 4,196,238
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 41506 - Posted: 8 Jul 2015 | 13:43:38 UTC - in response to Message 41501.
Last modified: 8 Jul 2015 | 13:44:08 UTC

My GTX690 died the same way. Later I've found the two 12V pins on the 24-pinned MB power connector was burnt. These 2 pins power the 4 PCIe slots and the memory and the chipset. The latter two are negligible when there are 4 high-end GPUs installed. The total power consumption of 4 high-end cards from the PCIe slots could be 4*75W at worst case. There is only that two pins on the 24-pin power connector which have to supply that much power. That is 300W+ combined, or 25A@12V. 25A on two pins is simply too much, so it is quite advisable to choose a MB which has extra PCIe power connector(s) to power 4 high-end GPUs in the long term.
I suggest you to check the 24-pin power connector on your motherboard & PSU for burn marks on the 12V (yellow cable) pins. If the PSU is a modular one then check both ends of that cable.

The four cards were never together (in the same chassis). They died while inside different computers, only the first one died while together with another one in the same computer. The last one died while alone for a long time. I was able to get 980s as replacement for the first three, I just inserted them in the same slot without issues (I'm quite confident that this not a MB or PSU issue).

I will double checked the 24pin power connector but I do not think this was the problem.

mikey
Send message
Joined: 2 Jan 09
Posts: 290
Credit: 2,036,666,115
RAC: 10,309,983
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 41512 - Posted: 9 Jul 2015 | 11:36:26 UTC - in response to Message 41506.

My GTX690 died the same way. Later I've found the two 12V pins on the 24-pinned MB power connector was burnt. These 2 pins power the 4 PCIe slots and the memory and the chipset. The latter two are negligible when there are 4 high-end GPUs installed. The total power consumption of 4 high-end cards from the PCIe slots could be 4*75W at worst case. There is only that two pins on the 24-pin power connector which have to supply that much power. That is 300W+ combined, or 25A@12V. 25A on two pins is simply too much, so it is quite advisable to choose a MB which has extra PCIe power connector(s) to power 4 high-end GPUs in the long term.
I suggest you to check the 24-pin power connector on your motherboard & PSU for burn marks on the 12V (yellow cable) pins. If the PSU is a modular one then check both ends of that cable.

The four cards were never together (in the same chassis). They died while inside different computers, only the first one died while together with another one in the same computer. The last one died while alone for a long time. I was able to get 980s as replacement for the first three, I just inserted them in the same slot without issues (I'm quite confident that this not a MB or PSU issue).

I will double checked the 24pin power connector but I do not think this was the problem.


A friend is checking a 760 that 'just died' on me, he said he found a short between the power block and the fan housing where they were touching, I had replaced 2 fans on it, and he thinks a plastic spacer or aftermarket fans may bring it back to life. He said that normally the clearance is pretty small so maybe I squeezed it too hard at some point. Mine did exactly the same thing, my fans didn't quit they were just running slowly, resulting in very minor overheating, ie 72c temps, and blowing them out didn't help so I bought 2 replacement fans. After the fan replacement the thing was never recognized by Windows or Linux again, I figured some capacitor had popped but he said nope not that.

Post to thread

Message boards : Graphics cards (GPUs) : "Invisible" Graphics Card (780ti)

//