I have been doing some tests rendering a sphere with motion blur and different methods of smoothing the model.
Doing these tests I have noticed that using sub division conversion on the maya model makes the render time 3 times slower.
Is this normal or am I missing something here.
If I use the maya smooth option it renders much faster than the subD alternative.
Render times
subD = 46 sec
smooth = 19 sec
no smooth or subD = 15 sec
I would say this is normal. When you use are using the Maya smooth node ( I think it is called polySmoothFace), V-Ray treats the triangles differently compared to using the subd option.
The subd option will cause V-Ray to use a special BSP tree, that can dynamically load/unload geometry when needed. Subsequently this tree is slower than the tree used when you have normal meshes without subdivsion (since their triangles are known in advance, the tree can be better optimized).
Also depending on your settings, using the subd option may be outputting a lot more triangles compared to the Maya smooth command.
Thanks for the reply, what you say makes sense it just seemed like a huge difference.
I am only rendering a sphere, so it makes me think how long will a character take to render with subD’s like we typically do with prman.
by default the max sub-d tessellation is set to 256… which is really high. You can try adding sub-d attributes and limiting the max tessellation which should speed things up.
I will try this.
I see the option for controling the number of subdivisions.
I assume you add both attributes to the object, one to enable and they other to control the level?
Or does the subD displacement node do both?
Ok a couple of things I have noticed.
If I enable mblur on the camera shape under Extra Vray Attributes, but have it turned off in the render Settings I still get motion blur.
Also I noticed on a simple sphere the following
subD and 6 Max substeps = 49sec
subD and 256 Max substeps = 51sec
how much memory do you have on your system, and what is your memory limit set to in the globals?
By default its super low… like 500… we have 16 gigs of ram on our machines so we set it to the max of 8000.
Displacement and sub-Ds and proxies all use a dynamic memory system that works really fast (assuming you have enough memory, and have allowed vray to access it all).
Setting the max. subdivisions to 256 does not mean that V-Ray will always subdivide it 256 times, it is just a maximum value. There should be visible difference between lower values though (e.g. 1 and 8).
Anyways, I agree that this is one area where we can optimize things further. The internal builds already have some optimizations in that regard.