Complex scene is rendered empty

[Micha]

Hi,

I try to render a train exterior, 8.0000.000 polygons. During Octane loading the scene there is no error, anything looks fine, but finally the scene is rendered empty.

I stopped the rendering, hide a big part of the scene, start again - it works, very low memory usage.

New test - the half is hidden -> one single, fragmented part of the scene is rendered only.

What can I do?

-
Micha

[face_off]

I suspect you are running out of VRAM or CPU RAM. What do you have Max Number of Polygons set to in the Settings? Octane 2 can handle approx 19.6 million triangles (halve that number if you have quads). How much VRAM do you have? When it renders - what are the exact memory usage numbers pls? If you click on the Octane Log - does it show any errors?

Paul

[Micha]

I updated my signature with the machine specs. Rhino is using 4GB RAM, approx. 8GB are free. VRAM 1GB of 6GB are used. Octane Log started - no errors, not any additional info there during starting the rendering. Max Polygon limit at 20.000.000.

Is the scene written to a temporary directory? My window temp dir is on a 4GB RAM drive.

[face_off]

If you send me the scene - I will take a closer look.

Paul

[Micha]

I can't send you this scene, but I will send an other, smaller scene, where a few parts are rendered only.

[face_off]

Thanks - it is throwing a cuda error 719 for me - so something is definitely wrong. I am investigating....

Paul

[face_off]

Micha - I think the issue is that there are 29,885 materials in that scene, and that is exceeding some sort of internal Octane limit. I have escalated this to the Otoy team, and also posted on the McNeel forums (http://discourse.mcneel.com/t/problem-g ... etry/24515). Whilst the Materials panel is showing 44 materials, the Rhino API is reporting 29,885 different materials - so there is a problem with that scene.

If you select the whole scene and assign a single material to it, the scene then renders correctly.

Paul

[Micha]

In the past at VfR there was a problem with Rhino file sizes. The solution was that a cleaning code was implemented that fix the duplicated materials. Looks like you need to do it too.

[face_off]

This is looking like a Rhino bug - but I will wait to see Andy's response. I wouldn't think de-duplicating would be possible because there would be no way to determine which materials were original's and which were duplicates. The best option would to manually select the geometry items and assign a new material.

Paul

[Micha]

But not the best option for the user. User doesn't like to think on workarounds and like smooth workflows. Cleaning a complex scene isn't nothing for users, it's something for an automatic running code.