Driver API error : 999

[Morne]

Hi

On one of my sites we just got 3 systems with GTX 580 3GB cards.

When using RT in CUDA mode, various scenes crash 8 out of 10 times.
The one machines keeps saying the display driver has stopped working but has recovered, on a nother machines I got a message NVidia OpenGL encountered an error and then in the RT window I get all these errors:


After updating the NVIdia driver to the latest, it still does the same.

Here we're on Win7 Pro, 32GB Ram, i7 3930K, Max 2014, Latest official VRay Build

Any thoughts?

[vlado]

The latest 320.xx drivers cause V-Ray RT GPU (and other GPU software too) to crash anyways, so I do not recommend using them. You will have to go back to a more stable version of the drivers - I will try to find out what drivers we are using on our test system.

Best regards,
Vlado

[Morne]

Yes please let me know as soon as possible please

[Morne]

One machine is using driver version 280.26, and gives the same error!

[3LP]

Yeah

I'm nearly convinced that it would be the Windows UI tasks are not processed on time.

http://www.chaosgroup.com/forums/vbu...ing-on-new-rig

I had the same issue 3 months ago, and fixed it with the TdrDelay registry value at 10 seconds.
Follow Vlado's recommendations but C) had definitely be my holy greal.
I have since then created a reg file that I have used on all the gpu nodes and RT GPU doesn't (nearly) crash any more. (Well not for that matter anyway).

Stan

[vlado]

Ah yes, there was that too...

Best regards,
Vlado

[Morne]

How does iRay and other realtime GPU systems compare to VRay RT? What is the problem here? Does RT just suck or are we 20 years away from actual production use. Everytime I tried RT, since it came out few years back, it always crash after 2 or 3 tries, or something is not compatible. So is it just my system that suck, or what is the problem here?

[vlado]

How does iRay and other realtime GPU systems compare to VRay RT? What is the problem here?

I can't really tell you, as I haven't used them much... However pretty much all GPU renderers have limitations of one sort or another. That doesn't stop many people from using them successfully in production. But I can tell you that all the hardware stuff is getting on my nerves too

Best regards,
Vlado

[3LP]

I haven't used the other one either, but as far as I'm concerned, RT is just so good and sucks so much the power of the GPU that it's usually not RT but windows the problem.
If you stay in the boundaries of what's been implemented and in the ram usage, it's been pretty stable for me for the last couple of months.
And having some specific GPU nodes with 4Gb cards helps a lot to democratize RT through our pipeline.

And like I said, the previews post of Vlado helped a lot to make the node stable...
Stan