I wanted to try out the Radeon Pro 5300M of my MacBook Pro 16 for the Rhino 7 Cycles but then I realized, that there was no OpenCL option available, whereas in the Rhino 6 version it is visible. I read here, that the AMD cards are not really supported with OpenCL for Mac in Cycles in Blender. Is it the same case in Rhino?
I am currently addressing stability issues, and since I haven’t had proper time to do good OpenCL testing after our update of Cycles this month I decided to disable it for now and re-enable when I’ve verified everything still works on OpenCL side. Indeed upstream Cycles no longer has OpenCL enabled on MacOS, the same version we are currently using. But I don’t know yet if that impacts us too or not.
Alright, thank you a lot and good luck with your work!!! Also respect for what has been done so far!
I guess that means, that Cycles for Rhino is an adapted version of the original Cycles for Blender? Will developments in Cycles for Blender appear parallel in new versions of Rhino as well?
Do you think it would make sense to use the Metal framework for the macOS version of Rhino?
I try to keep the Rhino version of Cycles in sync as much as possible. We have some customizations that we’ll be posting as patches for upstream at some point. Most of the changes we do are to ensure Cycles works nicely together with Rhino - drawing into Rhino viewport, slightly changed control over sampling loop, Rhino-specific texture coordinates / mappings, and basic clipping planes. Other than that it is pretty much the same.
I just tested Cylces in Rhino 6 a bit and got some interesting result with different rendering modes activated. I was testing a very simple filleted surface and comparing how long to took to get to 30 samples:
OpenCL with AMD Raedon Pro and Intel UHD combined: 100s
OpenCL just with the AMD Raedon Pro: 26s
CPU: 23s
Does this mean, that my CPU is just much better suited than my GPU, or that the OpenCL is not optimized yet?
It might be that we still compile in a bit more into the OpenCL kernel than necessary. I know from an upstream developer that enabling code that deals with voronoi shader node can slow down OpenCL rendering significantly, even if it isn’t used in the scene currently rendered.
I’ll be looking in more detail into OpenCL when my current task is done.
Very interesting, havent heard of a Voronoi shader before, but I´ll dig a bit deeper into these topics meanwhile… Would be really awesome if Cycles could make more use of the AMD card.
Much success with your current task then!
I was surprised because the WX 8100 was faster than the WX 9100. But yes, also in general, the OpenCL performance is not as nice as the CUDA performance.
Ok, I guess we have to be patient… AMD has joined the Blender development fund, so maybe we´ll see something soon.
Are there any news regarding the CPU rendering performance? I did some nice renderings with Cycles recently and it is just a pity to know, that they could be rendered 3x faster…