ChrisHekman wrote:TBModel wrote:ChrisHekman wrote:You can check the log in the OctaneGUI.
In the octane gui, go to the top bar and click "windows"
Then select "craete log window"
Then you should see the logs of this error. Could you share them with me?
If you are using interactive mode and unreal viewport rendering, it might be that the first frame just takes a while for heavy scenes. Have you tried waiting?
Caustics are supported. Although they require a lot of samples. You can increase the caustic blur to make them show up faster (although they become more blurred)
It is also advised to use octane materials for glass surfaces, as the conversion of glas isnt always perfect. Using octane materials can also help caustics appearing and renderign faster.
We have confirmed that this is a problem caused by RTX acceleration. We are rendering a scene of 28 million triangles. There are many vegetation in this scene, and the number of triangles including instances should be around 600 million. This scene can't use RTX acceleration at all. I see your other post, and also try to reduce the parallel samples to 2, but still can't use RTX acceleration. This scene The rendering of is really stuck.
After a lot of tests, if we slightly delete some polygons, for example, reduce to 20 million, the rendering speed can be increased from 0.5m to 5m, or even higher. I don't know why this happens? Is this a bottleneck?
In addition, if you simplify a part of the content, RTX acceleration can be turned on occasionally, and the speed can reach 10 or higher after turning on RTX acceleration. I don't understand why there is such a big difference. It seems that RTX can't be turned on because RTX doesn't support RAM memory yet? But why is there such a big speed difference when only 8 million triangles are deleted??
Could you share your device memeory? In the octane gui, go to file -> preferences -> devices.
Preferably a screenshot.
The reason there is a big footprint when using RTX, is that the acceleration structure that RTX creates effectively duplicates most of the mesh data in the scene in order to trace it.
It should support instancing, so it should not be a problem to have a lot of vegetation instances.
Note that unreal also consumes a lot of vram when using large scenes. So one solution could be to export your scene via the ORBX Exporter in the unreal Sequencer.
And then render out your ORBX file in octane standalone.
The text I posted above is the content displayed in the log. It prompts CUDA or insufficient memory, but I use 64g RAM as additional memory. When the number of polygons in the scene is small, it can render the scene faster, and RTX acceleration is basically normal, but when I render a scene with more than 30 million non instance triangles (the total number of triangles The RTX acceleration can't be used at all. The above errors have been repeatedly reported;
Moreover, even if I use normal rendering, the rendering speed is terrible. In the past, I could render scenes at a speed of up to 10-20 m (using the default parameters). Even when rendering large and complex scenes, it can be guaranteed to be around 5-10m, but the rendering speed of this scene (non instance triangle) with more than 30 million triangles is only 0.5m!! It's really slow..
In addition, when rendering this scene, RTX acceleration becomes completely unavailable. I have tried to simplify some polygons in the scene and control triangles (non instances) to about 20 million. At this time, the rendering speed becomes normal, about 5-6m, and RTX acceleration can be turned on normally. After turning on, the speed can be increased to 10-20m;
Is this caused by the speed difference between ram and VRAM? Or are there other bugs? The rendering speed of 0.5m is really too slow. This is just a 30 million triangle (non instance) scene. I can't believe what will happen to the larger scene and whether the speed will drop to 0.1M or even lower.