V-Ray RT Distributed rendering Question

[Syclone1]

So I have 4 render slaves in my server room and all have Quadro 4000 video cards in them, I thought I might be able to use them with V-Ray RT distributed rendering but looks like Im SOL. Just wondering if I am right in thinking that these are not supported or am I missing something. Screen shot of what I was getting on the render slave running RT.

[vlado]

Which driver version do you have? Are the cards listed as available for CUDA in the nVidia control panel?

Best regards,
Vlado

[Syclone1]

I just updated the one mentioned to "347.52-quadro-grid-desktop-notebook-win8-win7-64bit-international-whql" Ill have to find a monitor and run up to the server room to check the control pannel, it wont run in remote desktop mode for some reason.

[Syclone1]

It "looks like" Cuda is up and running when I log into the workstation directly...

[vlado]

Ok, you didn't mention the part about the remote desktop. CUDA is only available from a physical session on the machine, unless you have Tesla cards. CUDA will not work through remote desktop. You can use something like VNC or LogMeIn instead, since they create a physical display session.

Best regards,
Vlado

[Syclone1]

I was just using RDP to open and close the applications, I can get RT-DR running using the CPU's but not CUDA

[vlado]

Yeah, like I said it won't work with RDP.

Best regards,
Vlado

[Syclone1]

OK thanks Vlado, at first I was confused as to why you were blaming remote desktop but I just now tested it as Im running GTX780's and 980's and knew they "should" work lol. but I saw what you meant now where I used RDP to fire up the RT server on my 980 workstation and of course nothing, then went to the same workstation and logged in, shut down the RT server and started again and then tried the RT-DR again and of course it worked. So if you have a moment why does starting the RT server using RDP make a difference?

[vlado]

So if you have a moment why does starting the RT server using RDP make a difference?

When you connect through RDP, Windows creates a virtual desktop environment for applications to use, so that it can capture their output and send it to the client, and also to be able to support multiple connections from different users with different desktops. This virtual desktop environment does not support GPU acceleration of any sort and programs are not able to access the GPUs.

Other remote access programs like LogMeIn, VNC, TeamViewer work by creating an actual physical desktop session, so programs started through them can use the GPU. (The downside being that those programs don't support multiple user connections with different desktops.)

If you have Tesla cards, they have special drivers that are compute drivers, and not video drivers, which allows those cards to be accesible from a virtual RDP session. There have been requests for nVidia to do the same thing for all cards, and it should be doable, but I don't know if/when.

Best regards,
Vlado

[Syclone1]

Ah k got it, thanks much for taking the time to explain that. Greatly appreciated Vlado