I’m reasonably satisfied with Cycles on my nVidia with CUDA, but I would like to play around with OpenCL rendering for a couple of reasons:
- To see what performance looks like running the split kernel vs the monolithic kernel on my current card (even with the nVidia implementation penalty)
- I’m probably going to be adding to 2nd GPU in the not too distant future and want to look at the feasibility in running hybrid GPU rendering (which means either OpenCL or multiple rendering instances)
I saw an older post which suggested export CYCLES_OPENCL_TEST=all on Linux would disable the AMD specific check but that didn’t give me an OpenCL option in the system prefs. Is it possible to force-enable OpenCL with the ‘standard’ Blender build or do I need a custom build to try this out?