OTOY Forums

Hello, just looking for some advice, getting failed message from Octane on my GPU during render.

Have GTX Titan Z acting as sole primary GPU (a Dual Core 12GB Vram with 5760 cuda, divided evenly in 2 cores, 6GB Vram and 2880 cuda). Just noticing last day or so that only one of the 2 cores works in Octane. As background, was working fine recently, no changes to pc. GPU still now lists normal/usual status in Device manager as well as in NVidia Control Panel. When in Octane, I can select both as CUDA devices like normal (both listed with identical memory specs, cuda cores, clock, etc.)

Here is my problem now. For purposes of this post, I now refer to the cores as:
(1) GPU core 1 - working
(2) GPU core 2 - not working

If I select only GPU core 1, render proceeds and occurs normally, 6 GB Vram with 1/2 cuda of 2,880.
If I select only GPU core 2, render never starts and I get red text "failed"
If I select both GPU cores 1 & 2, render proceeds, but text is in orange and says "2 GPUs (1 failed)". Render occurs as if only one is active.

Any thoughts, setting modifications, or trouble shooting techniques are appreciated. Thx so much!

Have you overclocked it?

Hi, thanks for asking. I have not over clocked it (if I did, I sure did not intend to, I don't know how to!...).
And, it is not the overclocked version of the card. Since getting the error I was looking into underclocking to see if that would help, and I got confused with BIOS and software like MSI afterburner, so I haven't been able to try it yet. I did run a 'Heaven Benchmark' test on OpenGL and DirectX11, which showed both cores at ~62 degrees, 150 FPS, and memory of 3,330 Mhz. The Directx11 pass had a lot of artifacts and gigantic black stutter marks, whereas the OpenGL passed through with minimal artifacts. On a single core run of both DirectX11 and OpenGL, there were no artifacts, but all load (temp and memory, were being handled by a single core.

I also went into the global and per-application settings of 3d Visual/Management (...or something like that) on the NVidia control panel, and it had CUDA cores set to be used by all cores.

Thx for any ideas!
Win 7, 64 bit, ASRock Extreme 11 X79, 32Gb DDR 3-1600, intel core 7 3930k

Those kind of errors are usually either because of a faulty GPU core or because of OC.

Try youtube on how to use Afterburner

What is your PSU power?

smicha wrote:What is your PSU power?

good question, he needs at least 850 W if not 1000 for safety. It’s usually about 280W required per core at max output as I recall.

Yes. I'd recommend (300W+1)*# of gpus. About 900W PSU for 1xZ.

Hi all,

I have a 1500 watt Thermaltake Toughpower PSU.

I guess from the previous posts I am getting a feel for what a switch to failing could possibly be caused by:
A) PSU fault, or even a MOBO fault
B) overclock setting leading to issue
C) Card fault

I'm hoping its something I can tweak/adjust to restore functionality.
One thing, I did get upon trying again, was a 6/16000 render halt, where the grey background for a render actually started and appeared, but it was almost instantaneously inactive. But the grey background had little black, pink, and yellow speckled lines, which Im thinking would not have been part of the rendered 3d scene.

Thx!

Reinstall nvidia driver (run clean install option).

Hi all,

No success yet, but have made discoveries.
To the last poster, thanks so much for suggesting I reinstall NVidia drivers clean.

Reinstalling drivers had no effect. But rolling back drivers revealed something, in that instead of getting red text that said the GPZu failed, I got "the NVidia OpenGL driver detected a problem with the display driver and is unable to continue. The application must close." When I updated the drive back to current, I just got the red text without crash.

I also rolled back CPU main BIOS, tried more files for testing, and found that some scenes would work (both cores) while others would not. I began to think it was a BIOS issue. But when I reverted BIOS back to current, I now found that the same scenes render, and the same others do not. Then I discovered that the scenes that do render are temperamental, sometimes they work, and other times won't...sometimes get artifacts, other times won't. When trying through demo 2.17, it was noticeably more temperamental, in a bad way (less success). And the scenes that do not work, never work as dual core render. When I made I must not have had the settings active or something, as in I possibly had dual core off.

Here is the super kicker . My colleague let me try his Titan Z, and there were no issues at all. But, this other one is EVGA based, while mine is PNY based. And they have different VGA Bios. So I thought about backing up the PNY VBios and flashing it to match the EVGA VBios, but GPU-Z cannot export the Titan Z VBios.

So, it could be faulty core, or my card's VBios is not playing nicely with my PC BIOS somehow. Leaning toward faulty core, but then now I also found my card works fine in a dual core stress test on GPU-Z, making me still consider my card's VBios.

What do think, or suggest I try. I know others have had temperamental success on some scenes too for reasons related to Octane, bug in my case it's weird that another Titan Z handled them no problem.



Again, thanks so much for feedback and ideas!! Appreciate the help.