1

I purchased a used GTX 1080 TI a few days ago to use for 3D rendering. Unfortunately, the 3D render engine I am using (Redshift) crashes inconsistently when I attempt to do renders on this card, which essentially makes it useless. However, it is capable of outputting multiple displays and running Furmark with no issues.

I have an older 1080 TI that I bought last year which does not have this problem. I can render with no issues in Redshift on it.

Both cards are EVGA 1080 TI SC2 cards (the dual fan with the silver color).

To confirm the new card has an issue, I isolated both cards on the exact same machine, with the same NVIDIA drivers, same redshift installation, etc. All I did was swap the cards. The old one I've had for a while works fine, while the new one I just bought crashes unpredictably when rendering my own scenes and even when running the default Redshift benchmark. This NEVER happens on the old card.

I tried this swapping method on both an X99 Intel motherboard as well as an X670 AMD motherboard. Same issue on both.

I am surprised the old card can run FurMark with no issues but not Redshift. The seller advertised the card as working perfectly, so it seems like it probably worked fine for gaming (since that's what most people use these cards for) but not 3d rendering (since few people would test the card for that).

I don't really know what 3D render engines do differently that cause this card to have problems with Redshift. My guess is it is memory related since Redshift shuffles textures between VRAM and CPU ram often. Is it possible that the memory is overclocked somehow and I need to set it back to normal? If Ethereum had been mined on it, could that lead to memory problems? Or is this likely not memory related at all?

I have some comparisons of what I observed in GPU-Z while running Redshift on both the old card (which has no problems) and the new card, which sporadically crashes.

New GPU, with issues, below:

enter image description here enter image description here

Old GPU, working perfectly below:

enter image description here enter image description here

Another observation is that the new card with problems has a Bios version of 86.02.39.40.92 while the old one that works perfectly is 86.02.39.00.90. I suppose it is also possible that downgrading the BIOS to the version of the older working card could solve the issue. But I'm not sure.

I don't really know what to do. Is it a bad card, or are there some settings I can play around with to try and get it working?

joejoejoejoe4
  • 1,438
  • 9
  • 25
  • 48

1 Answers1

0

I ended up solving it by using EVGA Precision X1 and setting the clock on the card to "-40". This completely stopped the crashing. It works flawlessly now.

enter image description here

I guess the default clock setting on the card was too high?

joejoejoejoe4
  • 1,438
  • 9
  • 25
  • 48