Network Slave Not Utilizing CUDA?

[pixelmonkey]

I'm running a Toshiba laptop as a master (i5/Intel HD 4000 graphics) and a desktop with i7/GTX660 as a network slave. I've noticed that when I enable GPU rendering through the Toshiba, my only option is the integrated HD 4000 graphics. I expected the network slave to utilize CUDA through the GTX660, but instead it also defaults to HD 4000 (the desktop mobo also has HD 4000 as integrated graphics). Is there any way to force CUDA rendering through the GTX instead? Thanks!

[lycium]

Heya pixelmonkey (awesome nick ),

One thing it might be is that GPU acceleration doesn't seem to work through remote desktop on Windows; I'm going to test rendering with CUDA over network again now to check.

[pixelmonkey]

Hey Lycium, thanks for the reply! Fairly fond of the nick myself ...
I actually wasn't using remote desktop for the renders, both computers were on a local network and I was using TeamViewer to keep tabs on the i7/660Ti slave. It seems that because the job is sent to slaves in a '1:1' fashion, the slave is forced to use the Intel HD 4000 graphics as that GPU is also present and is the GPU specified on the master. Any validity to that theory? Thanks again for the reply and looking forward to any insight!

[fractal.design]

Yes I've the same issue.

Master (windows) is running GPU acceleration with CUDA.
Slaves (mac) have OpenCL.

In network slave report window it says that slave is running GPU acceleration with CUDA (device the same as Masters GPU device)


Be sweet to fix this (check for available GPU devices, if the Master's device isn't there, use the next available one, or something?)

[OnoSendai]

I think this is a bug in Indigo.
Will fix.

[OnoSendai]

The reported issues should be fixed for Indigo 3.6.12.