M1 Pro GPU: 6:45 (this was done without the benefit of Metal support in Blender Cycles)Īnyone tried running some LLM text generation on any of the M3 generation (M2 would be interesting as a reference point)? Anyone care to? * Default scene is set to 50 samples, but the test for the classroom scene has been upped to 300 samples as practiceĢ560x1440 samples * This isn't the default setting, so change these settings please.īlender 3.3 with OptiX for the Nvidia and Blender 3.3 with Apple M1 Pro GPUīlender 3.5.1 latest build as of last week with metal support for the Apple M2 Max I have unplugged. The Junkshop as downloaded (Nvidia using the OptiX) - Blender 3.6:Ĭlassroom scene 300 samples* Blender 3.5.1: So I wouldn't bother executing the python scripts and just run the render test. I ran the same test without executing the script and the render time was 4:25 minutes. * I ran the python scripts even tho I was given a warning dialog box it was a security risk. I suspect render times across all systems will likely see similar increase in render times. The Blender Cycles engine in Blender 4.0 has had some changes to improve the quality of rendering including subsurface scattering and glass energy conservation, etc, etc, (all available on developer notes for anyone curious). With Blender 4.0 Beta on M2 Max (no hardware ray tracing), Mac OS Sonoma 14.2 public beta They have results for Blender 3.5 rendering the Barbershop* scene in minutes compared to M2 Max GPU Blender 3.6 beta (similar render times to 3.5):
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |