I am curious to know what are the use cases of the “Subdivision” mode of the vray displacement modifier. From the little information I was able to read on this, it looks like it can be used in place of the standard mesh smoothing modifiers like TurboSmooth and OpenSubdiv. I did a couple of quick tests and it takes more time and memory than using the TurboSmooth modifier, but maybe I’m not using it the intended way?
Hi I think you are not using it the intended way. Maybe you could share your project with the forum or explain a bit further what are you trying to do?
It can be useful to replace turbosmooth. If I remember correctly it was a suggestion from Vlado for a car model I have
which I wanted to try with GPU. It wouldn’t fit using turbo but did if I switched to using displacement set to 0, so no actual displacement, only smoothing.
I am just trying to figure out what it is used for. So I’m not really doing anything with it at the moment. Help docs weren’t really clear about its use cases, it just says it can smooth the mesh similar to turbo/meshsmooth. However, in my quick tests, it used more memory than turbosmooth modifier. Maybe it is useful only with particular types of meshes that are extremely heavy such as HD car models similar to what @fixeighted suggested? Does it make sense to use this on any mesh that I currently use turbosmooth on? Again, I’m unclear about when exactly to use it, if that makes sense.
It can subdivide taking into account the edge length in pixels.
To compare it to turbosmooth, ensure you match the subdivisions (leaving TS to 0 for viewports. Both need to perform subdivision right before rendertime, this way.), set edge length to 0.0 (i.e. subdivide until max subdivisions are reached) and set the maximum number of subdivisions to the same as TS (i.e. 3, 4. down form 256.).
You’ll not want to leave edge length to 0 and max subdivs at 256.
It’s generally not useful to have an edge length below 1.0 pixels.
For example: i subdivide a sofa from Cosmos, from ~910k polys to 3642830 triangles.
Then render only the LC and not the final image.
The *sequence* time is the time to check.
An openSubdiv modifier takes 22.5s, a Turbosmooth takes 9.3s, a vrayDisplacementMod takes 8.5s.
It should be noted the job TS does is potentially inferior to OS and the disp mod, as the latter two are able to take creasing into account, and offer multiple subdivision schemes.
Notice that at rendertime both will then perform identically for speed (i.e. the triangle intersections) as both will be static geometry by default (the vrayDisplacementMod can be configured to stay dynamic. That may save some RAM, but it can also be quite a bit slower.).
Very interesting! Thanks for this nugget of knowledge, @_Lele ! I will make sure to try this out.