9800 GX2 reliability?

[Azurael]

Both of my 9800 GX2s now appear to have died... In slightly different ways.

The EVGA (newer of the two by a month) died a couple of weeks ago, it was in as the second card, so the first thing I knew about this is that its fan didn't spin down once the drivers loaded - all the GPUs were still showing up in device manager at this point but the two on the EVGA card were showing up with the warning exclamation mark... A few boots later, only one of the two GPUs on the second card is showing up, but it's working fine for CUDA processing (obviously SLI doesn't work with three GPUs in this configuration ) The fan control is on the lower of the two cards (as mounted), so the fan remains at full speed and that's also the primary display device, so I can't use the EVGA as the only card in the system any more.

Pulled that one out so I at least got SLI back, then a couple of days ago, one of the GPUs also disappeared on my XFX GX2 without trace... Evidently, it's the secondary (top) PCB that's gone on this one, as the device still works as the only display device, and the fan control still works. Which is especially odd, because the nF200 is on that PCB, so you'd the the whole card would die if it went.

Neither of the 'absent' 'cards' show up in NVflash...

Everest has also gone completely mental:

PCI-E 2.0 x16 port #0 In Use @ x16 (nVIDIA nForce 200 (BR04) 32-Lane PCI Express 2.0 Switch --> nVIDIA nForce 200 (BR04) 32-Lane PCI Express 2.0 Switch, nVIDIA nForce 200 (BR04) 32-Lane PCI Express 2.0 Switch --> nVIDIA nForce 200 (BR04) PCI Express 2.0 Switch --> nVIDIA nForce 200 (BR)

It also claims the link is running at PCIe 1.0 rates:

Bus Type PCI Express 1.0 x16 @ x16

Hmm, could just be symptomatic of the dying card though. I guess I should try out my GX2s on another mobo.

Anybody else had any experience of this? The cards have not been stressed at all. They probably only have 4-5 hours gaming time each as I've not had the time to play since I built the rig, and I really intended them (though never got round to it) to run 4 x F@H GPU clients.

p.s. since I posted this @ OCW originally, I've noticed the the XFX was getting really hot when powered down with HybridPower (wheras in the past it would be stone cold when the card was 'disabled'.) so I pulled it out and now I'm running with just the stunning Gf8200 IGP in nForce 780a....

I can't even put my GTS back in while I'm waiting for warranty replacement of the XFX 'cause I've lost my DVI cable and hybridpower isn't supported so the display would have to be connected to the DVI port on the card, not the HDMI port on the mobo

[Azurael]

Heh, to bump an old thread, I ended up building a frankenGX2 with the two remaining working PCBs, then flashing them with a matching pair of BIOSes because I couldn't be bothered to find all the paperwork and RMA the cards. It's been working fine since, although I reassembled it without the shell, which substantially lowered temperatures and may have been a contributing factor. I don't think I'll ever buy a dual GPU card again (let alone two) - I think TDP this high in a 2xPCI form factor without watercooling is inevitably bad news, but I resent how expensive GPU waterblocks are and like my CPU running uber-cool anyway.