Hi All, thanks for the great add-on - it seems that this issue is happening with me too, i have two identical workstations running on 2.8 and on the local (client) machine i have set each node to utilize cuda, which happens on the local machine, but the other it just utilzes and maxes out the cpu - it seemed to create the correct balancing between machines at 0.5 each (as expected), but the other machine is lagging far behind as it's only maxing out the cpu and not touching the gpu at all. The cards are Nvidia Quadro p4000's in case you needed to know that :)

Furthermore i seemed to have an issue syncing them and can only manage to do so if the blender application is open on the external pc.

I've also ensured that after each time i make a change to the settings that i resave and resync the connection so that it uploads again.

Great add-on and it's so close to what i need for these great gpu's! i might try it out on 2.79 when i get a chance!

Hi @Tony Bloomberg Not yet, we'll be investigating this though, trouble we're having is that we have one rig with a GPU here and its an AMD card so it uses openCL and we're not seeing this issue on openCL, it appears to be a CUDA issue, unless you're using openCL. So we're going to have to get our hands on a CUDA card or a system with one in. Might take a week or two until we know what's going on.

Was this ever fixed? i have the same issue. I've set gpu only on local and remote computer (in preferences and crowdrender settings) but still maxes cpu and no gpu usage at all.

Hi Jonathan, ok, that is a bit odd. We usually expect that the CPU will be active on all cores for the first part of the render when the BVH build is happening and the render is being 'setup' to be given to the GPU. This involves intensive use of the CPU, but it shouldn't continue. One thing to perhaps try experimenting with would be to look at the settings on that remote node, does it have hybrid rendering setup to use CPU and GPU?

If it does, would turning that off make a difference?