very very strange question :)

nooby

Hi everybody, recently I was arguing with my colleague at work at a certain point, so the question is , what is more performance consuming for most 3d packages - total number of polygons in the scene or number of objects displayed at the same time? this might sound stupid but anyways I'd like to know the truth

[HP]

Isn't it the same? Since objects are made out of polygons?

Or you do you mean, what's more of a performance hog, polygons or draw calls? Quick answer is they both are.

It's all a matter of balance, it's not a matter of 0 and 1.

#Edit: I just realised you were asking regarding a 3d package, I thought you meant inside a 3d engine.

Mark Dygert

With 3dsmax its better to have a bunch of separate objects instead of one object that is a bunch of elements.

The reason for that is "adaptive degradation".

With it turned on it can degrade the visual appearance of individual objects when the viewport starts to chug, switching the material off, switching to wireframe mode and even dropping down to displaying just a bounding box.

It can't do this on sub-object "elements" inside of one object only on objects. So if you do something like scatter rocks on a path, or create a bunch of high poly bricks you're better off leaving them as separate objects.



It tries to degrade things farthest from the camera and smallest on the screen, first and will degrade what you have selected last. So even if you have a bunch of separate objects if you have them all selected it thwart Adaptive Degradation.

nooby

sorry me, I mean that if for example there are 10 objects with 100 polygons each and 1 object with 2000 polygons, what will affect mostly on frame rate ?
@[HP] - thanks for reply

nooby

@Mark Dygert - thanks a lot for explanation

Ark

Well having multiple parts would be well suited if its going in a game engine as the engine would be able to cull the sections that are not visible whereas if it was one big piece, the culling wouldn't happen until the whole mesh was hidden. Ofcourse in multiple parts, they might add more geometry, but it's all a balancing act like HP said.

Ghostscape

I've actually found that 3ds Max seems to have larger issues with 200 objects at X triangles total than 1 object at X+ triangles.

Something else that is useful for scene management (particularly high poly bake and builds) is making sure that imported sculpts are editable mesh, not editable poly - for whatever reason editable mesh objects are a lot easier for 3ds Max to manage than editable poly by a sizable factor.

It can also help to collapse turbo/mesh smooth down to an editable mesh, but that has the obvious disadvantages of making the mesh unworkable (but it's good for baking).

poopipe

the modifier stack obviously has an effect and so on but if we're talking simple mesh objects...

a million tris in a single object will perform considerably better than a million tris spread across a thousand objects.


I'm in two minds about adaptive degradation - it depends on the scene but i've sometimes found that the act of degrading slows stuff down just as much as displaying the full res objects in the first place. cunning use of layers and hiding is the only sure way to keep performance tolerable in big scenes

nfrrtycmplx

Mark is correct in theory:

But I find Ghostscape to be correct in practice:

Which leads me to believe "it depends"...

Depends on your system spec and version of the software IMO... On my machine 1000 objects at 100 polygons is way worse than 1 object at 100000 polygons...

The only way to know for sure is to do the test yourself on your machines and the conversation is no longer a debate but rather something that can be observed...

Ace-Angel

Smoothing groups, shaders and number of vertices also matter. I remember one time, I ended up giving all of my models 1 sg in Max, and the performance sky rocketed to Neptune and back.