Has anyone here managed to render a task per GPU in Cinema4D?
I’m testing on a machine with 2 GPU’s, I’m using the -parallel argument to start multiple instances of C4D, and I have set up 2 workers that each use one of the GPU’s.
No matter what I do, I can’t get the workers to both render.
Usually, both will start a task, one of them might render, the other will say it has rendered but saves no frames.
They are definitely only using the GPU assigned to them, but it seems to be impossible to get them both rendering at the same time.
I’m assuming that I’m doing something wrong, so if anyone out there has any tips before I start a support ticket, I’d be glad to hear them…
which version of Deadline, C4D are you using? and which renderer?
are you doing different frames or is the same file being overwritten? what do the logs for each task say?
Are you monitoring the GPU usage with MSI Afterburner or similar? are you able to submit each job individually to check the render works on the correct card and renders out ok?
We have seen the “Render has finished but no frames are rendered” problem now and then. Its usually a C4D licensing problem which is weird as your two C4D instances are both on the same machine
Can you post the logfiles?
I’ve just checked everything again (with Afterburner), and it looks like the individual workers aren’t using the correct GPU at all.
If I send a job to just an individual worker, all of the frames get rendered.
As soon as I want to use both workers simultaneously, that’s when the problems start.
@bonsak: The logfiles for the tasks that rendered and the tasks that didn’t render are identical.
I’ll have to do some cleaning up before I can post anything here, but that is the first place I looked.
I also just tried sending the job to just one of the workers and using the “Select GPU Devices” in the C4D submitter to limit the render to one particular GPU, and that didn’t seem to make much difference either.
The “-redshift-gpu” argument is also being correctly sent in the jobs’ command line…
I just checked all of the Redshift logs, and all renders have always used both GPU’s, no matter what I have set up.
I sent a test job to just one worker, after removing the GPU affinity in the worker preferences, and in the C4D submitter I set the job to only use 1 GPU.
It made no difference and just used both GPU’s.
Update:
I just exported to Redshift proxies and rendered with Standalone, and everything worked as expected. I had 2 concurrent tasks using one GPU each.
Our C4Drender pipeline has been customised quite a bit, and it looks like the problem might lie there.
