V Ray RT GPU Crash after installing Tesla Card

[leehenshall]

I have recently installed an Nvidia Tesla C2050 to run along side my Nvidia Quadro 3800.

Since installing the new Tesla card, Whenever I launch a V Ray RT render using OpenCL (single kernel) my RT Dos render window/log crashes and nothing happens in my render view.

Just wondering if anyone knows how I can fix this?

Also will using a Tesla card improve my V-Ray RT performance?

Tried latest/legacy drivers.

Thanks

[vlado]

Which are the driver versions that you installed for the Tesla card?

Other than that, a Tesla card is not particularly faster than a GeForce card, but is more reliable and has more video memory.

Best regards,
Vlado

[leehenshall]

Thankyou for your feedback,

I have tried 276.14 and 275.65.

Is there a place were V-Ray RT stores a log that I can show you guys?

System is:

Windows 7 64bit Service Pack 1
16gb Ram
2x Xeon E5530 2.40GHz
Quadro FX 3800
Tesla C2050
Maya 2012
V-Ray For Maya 2.00.04

[t.petrov]

Hello,

As far as I know there is no log file, but you can copy and paste the info from
the "Command promp"/console window, where the RT server is running. Do
you see such window? If no, probably you're running the server as service,
which won't work, because the GPU is not accessible from Windows services.

Hope it helps,
Teodor

[vlado]

I have tried 276.14 and 275.65.

Can you try 270.xx? Also make sure you are running the 64-bit version of V-Ray.

Best regards,
Vlado

[leehenshall]

I have just tried 270.90 and I still have a crashing command window.

I have attached an image illustrating the problem:



Edit: I definitely have 64 bit V-Ray Installed.

[vlado]

Thanks for the screenshot - it seems to be because the two graphics cards have different architectures - the Tesla is a Fermi card, while the Quadro is not. You can try disabling the Quadro card from the V-Ray RT OpenCL device selector utility from the Start menu and see if this solves the problem.

Best regards,
Vlado

[leehenshall]

Thankyou for all your feedback.

I have noticed that V-Ray for Maya does not have a "OpenCL device selector utility" In the start menu but V-Ray for Max does.

If I use the V-Ray for Max selector will this also work for V-Ray for Maya also?

Please see the attached image.



On a more general note, If I can only use one of these cards is it even worth using my Tesla card atall?

It seems to me that Tesla cards are more aimed towards processing scientific imaging data rather than media creation and GPU rendering.

My work machine acquired this card because it was going spare, but if it's not really going to help me much should I just take it out?

Since I have 2 quad core processors would it perhaps be better to simply do RT renders on the CPU?

[t.petrov]

Hi,

You can manually choose the device by defining an environment variable VRAY_OPENCL_DEVICE=1 (or the number of the device you want to use).
You can specify many devices using commas.

On the other questions:
1. VRay RT GPU is not using rasterization, but raytracing and thus is not using the GPU as all the games are using, but more like a scientific application.
We used the GPU for computations, so a Tesla card is perfectly suitable for VRay RT GPU.
2. VRay RT GPU works best with Fermi base GPUs, the older ones have limitations and are slower.
3. You should test what is faster with your scenes. Also keep in mind that the GPU has lots of limitations (only image textures, some of the materials don't work, less memory per scene, etc).

Hope it help,
Teodor

[vlado]

I have noticed that V-Ray for Maya does not have a "OpenCL device selector utility" In the start menu but V-Ray for Max does.

If I use the V-Ray for Max selector will this also work for V-Ray for Maya also?

Yes, it will work too.

On a more general note, If I can only use one of these cards is it even worth using my Tesla card atall?

You'll have to make some tests, but I think the Tesla 2050 is a lot faster faster than the QuadroFX 3800.

It seems to me that Tesla cards are more aimed towards processing scientific imaging data rather than media creation and GPU rendering.

Tesla cards are more durable and have more video RAM.

Since I have 2 quad core processors would it perhaps be better to simply do RT renders on the CPU?

Again, it depends for the type of work you want to do.

Best regards,
Vlado