Hey all,
I'm working on a render comparison between the top render engines and I've put together a sort of litmus test. The original scene was created in Blender and rendered with Cycles and now I'm going through the process of setting up the same scene in each of the other engines (using Max).
Now, the original scene in Cycles uses about 8gb of Video memory which is fine for my 1080ti testing rig and the memory usage has remained consistent(ish) on the other engines but with Octane I'm getting about 15gb+ so I can't fairly compare.
To try and narrow down what the problem might be I disabled most of the scene except some SUPER basic geometry like walls and other basic shapes and set it to clay render. Even then it's using 11GB of memory with a relatively low resolution HDR.
Any ideas on this? It's so strange. I can provide additional information if required of course.
Any help is hugely appreciated.
Bill Barber @ Poliigon/BlenderGuru
Memory usage vs other render engines.
Forum rules
Please add your OS and Hardware Configuration in your signature, it makes it easier for us to help you analyze problems. Example: Win 7 64 | Geforce GTX680 | i7 3770 | 16GB
Please add your OS and Hardware Configuration in your signature, it makes it easier for us to help you analyze problems. Example: Win 7 64 | Geforce GTX680 | i7 3770 | 16GB
Good question. I heared some opinions too about Octane high memory usage that grows from version to version.
Architectural Visualizations http://www.archviz-4d.studio
Hi,
first of all, the Max Sampling value in the Kernel settings has an impact in speed, but also in VRAM consumption.
In v2 was set to 1, in v4 to 8, and now 16 by default.
Also note that you need to set the Image texture nodes for Bump/Roughness/Amount textures, and in general for all grayscale textures, to Float/Grayscale, instead of Normal/RGB, to save up to 3x memory per texture.
From some versions you can also use texture compression, to save even more VRAM memory.
ciao Beppe
first of all, the Max Sampling value in the Kernel settings has an impact in speed, but also in VRAM consumption.
In v2 was set to 1, in v4 to 8, and now 16 by default.
Also note that you need to set the Image texture nodes for Bump/Roughness/Amount textures, and in general for all grayscale textures, to Float/Grayscale, instead of Normal/RGB, to save up to 3x memory per texture.
From some versions you can also use texture compression, to save even more VRAM memory.
ciao Beppe
Beppe, Can we expect in near future some further NVlink optimizations regarding to memory pooling for textures too, not only for geometry or it is not possible ?
Architectural Visualizations http://www.archviz-4d.studio