OTOY Forums

Hi, one of my 2 GPUs almost always fails in a few minutes after the rendering process begins.
It happens with Poser plugin as well as Octane standalone.
The rendering continues at half the speed after one GPU fails.
Octane standalone says at the bottom of the window “2 GPUs (1 failed)”.

It is always the same GPU that fails, and the other one never does.
My system has two GTX 780 with the same amount of VRAM,
but from different vendors with different BIOS etc. with different clock speed according to GPU-Z.

It happens with a relatively simple scene with even just one Poser figure,
or with any of the sample files that came with the Octane standalone.

Poser plugin log shows the following message:
===
Started logging on 24.01.16 17:40:51
OctaneRender 2.24.2 (2240001)
CUDA error 700 on device 0: Error code unknown.
-> kernel execution failed (pt)
device 0: path tracing failed
===

In case of Poser plugin, closing and reopening the viewport windows brings back the failed GPU to work,
but it fails again shortly after.

Even after the GPU fails in Octane, the Windows device manager still shows it is working properly,
so it appears to be an Octane specific issue.

Is there any known incompatibility issue with a certain type of graphics card,
or is this problem unique to my particular set of hardware?

I posted the message in this forum because I don’t normally use Octane standalone but only use Poser plugin.

Thanks very much for your help,

Masato

Hi Masato. It looks like you potentially have a faulty card, or perhaps one of the cards is overheating (GPU-Z will help determining that). Try underclocking the card that keeps failing.

Paul

If it blows up underclocking, then I'd pulling a card, running each one solo to establish if the card is good or bad.

Yes, and also - there have been issues reported with Nvidia driver 361, so try installing 358.50.

Paul

Maybe it is not plugged properly into the slot?
I had this case recently. Mostly a render fails because of Vram running out.

just an idea (similar to post before), maybe it's PSU? power supply & contacts are one of those reason to check first..

Hi everyone, thanks very much for your advice.

I checked and tried everything but with no success.

Then I thought about running some of the sample graphics apps provided by NVIDIA.
So I enabled only the seemingly faulty card and tested, and none of them even launched!!
They all worked fine on the other card, so I have almost concluded that I have a faulty card.

Theoretically there is still a possibility that the PCI slot is causing the card to fail,
so I will try swapping the cards over the weekend and see what happens.

Masato

Hi, swapping the graphics cards I have confirmed that the graphics card is faulty.
I also discovered that the PCI slot that had the faulty graphics card isn’t quite right either.

I got this system just 2 weeks ago (motherboard + 2 GPUs),
and the shop has agreed to replace them free of charge.

Thanks everyone for your support.

Masato