From Lightwave to RNDR

[mikefrisk]

Maybe it was because it was just a small test (100 frames) but my export seemed pretty seamless. It didn't take long to export at all. Make sure you check the file in Octane Standalone to make sure it loads and renders properly before uploading the ORBX file. The ORBX exporter for C4D is updated quite a bit (so maybe it's a LW thing, sadly). I'd hear more complaints than I do if it were such a big problem. I do hear complaints, because I'm on the RNDR network (mostly as an operator, but artist on occasion as well) and keep tabs of both the artist side and operator side).

How many polygons in your scene and do you have any animation/deformations?

Each frame of an ORBX file is saved as essentially a point cloud afaik. If you have a scene with a lot of movement it basically has to bake the entire scene out frame by frame, which results in massive ORBX files that take hours upon hours to export creating massive files.

[juanjgon]

Yep, I suspect that it is the same that happens in Houdini when the users are trying to create .orbx files including hundreds of frames from scenes that have objects with non-constant topology, volumes, complex particle systems, etc. The .orbx files become so huge (tens of GB) that at some point everything seems to fail. This is why probably almost all the other renderers export the scenes to single-frame files, I mean, a scene file for each frame. I know that .orbx is another concept of scene file, including inside the animation, textures and so on, but at some point all this information plus the animated meshes/volumes/etc. can create files so huge that are hard to manage.

-Juanjo

[NemesisCGI]

Yeah, RNDR is a really poor setup. Really the best system would be for the host plug-ins to see RNDR as GPUs or be able to send first the scene data and then just update the scene frame by frame rather than packing a full scene.
Personally, I'd rather see the RNDR GPUs as virtual GPUs on your own system. So you will only see one GPU but RNDR would under the hood use lots to get the render done. The bottleneck for this would be your connection speed though.
Another option would be you use it like most other online render farms where you upload your native scene & content to a server. So in the case of LW you'd upload a .lws & its content.
The problem again if you have large data sets like VDB animations or massive water or particle sims. These can go right into the TB size.

[mikefrisk]

I'd rather see the RNDR GPUs as virtual GPUs on your own system.

Absolutely. This would be the best approach. I don't know exactly how the render farm works but something tells me this wouldn't be possible. The GPU's are farmed out in queues meaning they are not always available and allowing them to be accessed remotely might bog down the server (but I have no idea).