Maxwell Render | Official public forum

My question is this.

Some render engines seem to like more 3D detail than others. (High poly counts vs low). How does Maxwell Render prefer things? For a given render should I try to keep the poly counts low on things lets say in the distance, or is it ok to give it what you have and go?

Also, lets say I'm using a flat surface for a simple floor (no displacement), should I just use a plane with like 2 triangles, or is it better to have it subdivided more?

I apologies if this is answered someplace, I didn't find the answer yet on the forums, but I may not have searched for the right terms yet. And I don't remember seeing anything in the manual mentioning this (yet) there is a lot to absorb so maybe I missed it, and I'll find it as I go over things again.

It's just a question of how much RAM you have, Maxwell doesn't really care how many polygons are in the scene, it won't cause the render time to increase. It's just the initial voxelisation stage that will take longer but we are talking seconds or minutes.

If it's a floor and you're not using displacement on it, there's no reason to subdivide it in more than 2 triangles.

Ok, sounds good.

Thanks

In my experience, it would be better to make models as detailed as you can for rendering with Maxwell.
But you don't have to have tens of triangle for a complete plane. Only two triangle will do.

That's true, and you should try to work by these famous words.. "The devils hides in the details" which means that your mission is to bring out the devil from your images by adding as much detail as you possibly can do.

Show no mercy to those poor polygons

/ Max

Yes, slice them polygons up! (for curved surfaces)

i usually create the 'master' scene with tons of polys. And then -basing on camera position- i hide everything that is not visibile directly and INDIRECTLY (backfaces, reflections, shadows and so on..), creating so a different and optimized scene for each camera.

this helps to obtain clean scenes in less time. useful for big renders where time machine is greater than my scene optimization time.

Yes for sure but this optimization is nor completely correct for an unbiased rendering... And, sincerely, somebody try to render the same scene with for example 700k polys and the same with 450k, i am quite sure the bench will be quite similar.

In relation to this topic, i had a question: how does maxwell behave with NURBS? I mean, will a Rhino maxwellized object be heavier due to meshing in respect with a model already built with polygon modelers or not?

Wouldn't you have control over the mesh detail with NURBS surfaces, for when they get tessellated (if that is the right term) into triangles?

Yes, but when you manually mesh from Rhino (at least when i do it, beeing not exactly an expert) if you have, let's say, a wall, that wall will be described with 24 polys, 4 for each face while in a poly modeler you would describe it with six quadrangles, one for each face...am i right?

6 sides yes, or 12 if they are split into triangles. But if the number of polygons isn't a huge issue, (unless you are low on RAM) then I wouldn't worry to much.
Don't know much about Rhino though.

Ok, next question along these lines. How critical is it to not let objects intersect?

On solid objects, I can imagine a misplaced item passing into a translucent object could mess things up...

Something that slows things a lot is crossing many clipped planes.

I think the main rules about intersecting geometry are:
Do not intersect SSS objects.
Do not intersect emitters.

The rest work just fine. I believe even the clipped plane issue has been fixed.