Blender 2.80, Cycles, GPU-Rendering

A few days ago I had my first meeting with Blender 2.80. Apart from all the other big changes Cycles changed a lot as it is now possible to use CPU and GPU at the same time for rendering (https://wiki.blender.org/wiki/Reference/Release_Notes/2.80/Cycles#GPU_rendering), A good rreason to take a small scene and compare render times. These are the rsults:
A few things are remarkeable.
Of course, combined rendering is the star of this show. Almost twice as fast, without spending any money on additional hardware.
BTW.: Forum software should provide a way to create tables. I never wanted to be an ASCII artist.
- Code: Select all
CPU: AMD Ryzen 7 1800X, 8 physical cores, 16 threads
GPU: Nvidia GTX 1070
All results in minutes
| CPU, Tilesize 32x32 | GPU, Tilesize 512x512 | CPU+GPU, Tilesize 32x32 | GPU, Tilesize 32x32
----------------------------------------------------------------------------------------------------
2.79 | 7:32.30 | 5:34.29 | |
----------------------------------------------------------------------------------------------------
2.80 | 5:56.64 | 6:17.17 | 2:58.15 | 4:52.77
A few things are remarkeable.
- CPU-Rendering got faster with Blender 2.80.
- GPU-Rendering got faster, but be aware of the next point.
- Small tile sizes work better than big ones with GPU. In 2.79 it's the opposite.
Of course, combined rendering is the star of this show. Almost twice as fast, without spending any money on additional hardware.
BTW.: Forum software should provide a way to create tables. I never wanted to be an ASCII artist.