Vray Standalone has stopped working...

[digitalx3d]

Vlado, a couple times I have gotten RT GPU working, but most of the time I get the Windows error "VRay Standalone has stopped working".

My setup:
Vista 64bit
3dsMax 2011
Dual AMD Opteron 2216 CPUs
Video 1: Nvidia GTX570 1GB
Video 2: Nvidia GTS250 1GB
Nvidia Driver: 266.58

The first scene I tested was on a brand new Max/Vray install. I made a scene with several spheres- 4 wide x 4 deep x 5 high, a ground plane underneath and a Vray light. I get this error almost everytime I try to render the scene in GPU mode. The very first time I tried in GPU mode, it just appeared to hang there "waiting for image data". I closed the RT window after a couple minutes. There was no error. Then, I don't know if it was the next try, or several tries later, but after sitting there at "waiting for image data" for several minutes, the RT image finally appeared. After closing this window and restarting more RT renders, the loading time was very fast, like normal. I was switching between OpenCL devices trying to test them individually. Oddly, my results showed the same results regardless of which video card I was using OR even for both! So, that leads me to think that even when I have both selected, RT is only using one of them?? Do you know of any issues with these cards? Or anything else in my setup?

Thanks!!!

[vlado]

Look at the V-Ray RT server window to check if V-Ray is actually using the GPU, or of it is running on a CPU. The first time you run it on the GPU, there is expected to be a delay (up to several minutes) while the OpenCL code is being compiled for the GPU. The GPU code is cached so sub-sequent runs should be a lot faster.

There will be issues (i.e. crashes) if you have several cards with different architectures (e.g. a Fermi and a non-Fermi card), as V-Ray will attempt to load the same GPU code on all cards. If that happens, use the VRAY_OPENCL_DEVICES environment variable to specify which devices you want it to run on, and delete the *.clbin file with the GPU code in your temporary folder before running V-Ray RT again.

Best regards,
Vlado

[digitalx3d]

ooooohh.... well, that might explain it then. Is selecting only one card in the "Select OpenCL devices" sufficient to avoid this conflict? or should I try disabling it in the Windows Device Manager? Or removing it from the system completely? EDIT- I ask this because I have already tried selecting just one card or the other and still received the error. Is that .clbin file in the temp folder the cache file you are referring to? Maybe I have to delete that again and make sure only one card is selected?

Is this conflict (cards with different architectures) something that can be fixed in an updated version of RT? Or will this be a permanent limitation?

Thanks!!

[vlado]

Maybe I have to delete that again and make sure only one card is selected?

Yep, that should be enough.

Is this conflict (cards with different architectures) something that can be fixed in an updated version of RT? Or will this be a permanent limitation?

It can certainly be worked around by compiling the GPU code for each card separately. The issue we had with this is that the OpenCL driver would re-compile from scratch regardless if the cards are the same or not, so for systems with 4 or more GPUs we were looking at 15+ minutes which was unacceptable. But yes, it is something that we keep in mind and will look into it at some point.

Best regards,
Vlado