can't get gtx 970/980 to fold.

[Tobben]

i recently received a msi gtx 970, i have it installed alongside a sapphire 7950 gpu. I followed the instructions here:
/viewtopic.php?f=67&t=19989&p=199379#p199379

if i edited the current slots the fah client kept resetting the slot to -1, so i removed both gpu slots and added 2 new slots, one with gpu index 0 as my 7950 gpu appeared first in cmd, and cuda/opencl index to 0, as my 7950 appears first in gpu-z, then on the second slot i set the gpu index to 1 and open cl/cuda index to 1 for the same reasons, but i keep getting this error on the gtx 970, slot:
12:33:25:WARNING:WU03:FS02:FahCore returned: GPU_MEMTEST_ERROR (124 = 0x7c)
any idea what's causing it or how to fix it? i get the error pretty much instantly.

Mod Edit: added 980 to title

[7im]

If you remove the 7950, do you get the same error on the 970?

[Tobben]

If you remove the 7950, do you get the same error on the 970?

interesting that you should say that, i hadn't really done any proper testing, if i remove the 7950 and install the gtx 970 in the top slot, my pc doesn't seem to want to boot properly, first it displays B2 then a single underscore _ that is blinking on an otherwise black screen. Currently typing on my secondary screen from the integrated intel graphics..

if i look very closely at the pcb of the 970 there is a slight color discrepancy, one part is slightly darker than the other, almost looks like moist or a sticky substance, but it's not, although the difference really isn't that clear. Not sure if i got a doa or something else is going on.

seems the only reason i was able to install drivers in the first place is because i was using a 7950 as primary gpu.

[bruce]

The correct value for gpu-index does not depend on which one is in the top slot. Windows uses a proprietary method of enumerating GPUs; Linux does not.

When the value is set to -1, it is configured by the client. In your case, the first one identified by the client, the -1 becomes the value of 0.

[Breach]

Try:

1. Reseating it carefully - make sure it's fully inserted in the PCI-E slot - if you have another one to try in do that. Make sure that the power connectors are properly and full connected.
2. Clear your CMOS (might want to do a backup first).
3. Make sure your MB has the latest BIOS - if not update. DO NOT use the MSI Windows Live Update utility - it's been known to brick boards.

If still no go I'd say it's DOA. Good luck.

[Tobben]

Try:

1. Reseating it carefully - make sure it's fully inserted in the PCI-E slot - if you have another one to try in do that. Make sure that the power connectors are properly and full connected.
2. Clear your CMOS (might want to do a backup first).
3. Make sure your MB has the latest BIOS - if not update. DO NOT use the MSI Windows Live Update utility - it's been known to brick boards.

If still no go I'd say it's DOA. Good luck.

thanks, i was hoping to avoid all this :p, but a bios update might be worth it? wouldn't have thought that would be necessary though. Never liked doing bios updates..

The correct value for gpu-index does not depend on which one is in the top slot. Windows uses a proprietary method of enumerating GPUs; Linux does not.

When the value is set to -1, it is configured by the client. In your case, the first one identified by the client, the -1 becomes the value of 0.

i didn't do it based on what card was in the top slot, i followed the "guide" or what you want to call it that i found on this forum, how would i go about finding out what cuda index and open cl index a card has though? be it 0 or 1

[Breach]

thanks, i was hoping to avoid all this :p, but a bios update might be worth it? wouldn't have thought that would be necessary though. Never liked doing bios updates..

Honestly? I think the card is dead. But if you want to try before you RMA give it a try - a BIOS update is a long shot, but you never know.

[bruce]

It's not 100% clear to me if you're using the latest drivers, but i think it goes without saying that upgrading them from the manufacturer is highly recommended. I can't comment about the other recommendations ... except that personally, I would have gone for the RMA before this point.

[Breach]

The point is, on the off chance that this is due to a MB incompatibility, that even if he RMAs, his replacement card's behaviour won't be any different. But indeed, the fact that there's apparent damage to the PCB, the BIOS error code and the failed GPU mem self-test points rather to a dead card.

[Tobben]

The point is, on the off chance that this is due to a MB incompatibility, that even if he RMAs, his replacement card's behaviour won't be any different. But indeed, the fact that there's apparent damage to the PCB, the BIOS error code and the failed GPU mem self-test points rather to a dead card.

It's not 100% clear to me if you're using the latest drivers, but i think it goes without saying that upgrading them from the manufacturer is highly recommended. I can't comment about the other recommendations ... except that personally, I would have gone for the RMA before this point.

If you remove the 7950, do you get the same error on the 970?

ok guys, thanks for all the help. According to msi the the new gtx 900 series doesn't come with a regular legacy bios, so i did actually need a bios update to support the UEFI VBIOS, and the sticky substance is residue from cleaning in the manufacturing process, which is not uncommon apparently. Both cards are folding just fine now. But the bios update made it so that my fans are running at more or less 100% ... so with 2 gpus and one bigadv machine on my shelf this room has become laughably loud, my near dead silent pc has become nearly deafening one bios update and one gpu later, it's a real shame the card suffers from coil whine.. hopefully that goes away.

[Breach]

Duh. Should have thought of that, I knew 900 comes with UEFI only BIOS because I asked on the MSI forum myself The MB BIOS update should bear absolutely no relevance on the GPU fan speed (unless you have some very weird BIOS). I'd suggest you get Afterburner and make sure the fans are in Auto mode with no custom curve (default curve is fine unless you're ocing). There have also been some reports on the MSI forum on the fans being stuck in 100% mode, but I'm not affected so I don't know details.

[billford]

Duh. Should have thought of that, I knew 900 comes with UEFI only BIOS because I asked on the MSI forum myself

Is that specific to the MSI models or does it apply to the card in general?

I was looking at the MSI 980 because I prefer not to use reference coolers but BIOS updates make me nervous

[Breach]

Duh. Should have thought of that, I knew 900 comes with UEFI only BIOS because I asked on the MSI forum myself

Is that specific to the MSI models or does it apply to the card in general?

I was looking at the MSI 980 because I prefer not to use reference coolers but BIOS updates make me nervous

No clue whether that's a general thing or, I was told that: "You are using a UEFI/GOP vbios and there are only UEFI/GOP versions available for GTX9xx series."
https://forum-en.msi.com/index.php?topic=183746.0

I think that normally OEMs get the reference BIOS from nVidia so I'm guessing it's across the board. It would be only an issue if you're using the card in a non-UEFI BIOS system anyway.

[billford]

OK thanks.

I've no idea what the BIOS is in the machine I have in mind, I'll check next time I reboot it.

[Kjetil]

Way reboot? Install this software: HWiNFO64