Scene DAG question...

[Steve Davy]

Here's a (maybe) interesting question for the technical peeps...

If you have a scene that contains hundreds or thousands of individual objects (say cubes), is there a performance advantage such as with rendering if these are combined into a single mesh, even though the number of points and faces remains the same?

How about with Maya itself, scene saving etc.? Any advantage to one over the other or is it the same amount of data that has to be crunched either way?

[matt estela]

My vague recollection from way back when talking to RnD at AL was a single million polygon object is cheaper than having a million objects each of 1 polygon.

There's always exceptions and whatnot, but broadly speaking each object requires a transform, plus the xyz position of each vertex. For any rendering operation, either offline or in the viewport, it has to multiply the vertex positions against the object matrix; its more efficient to multiply a million vertex positions against a single transform matrix, than it is to multiply a million vertex positions against a million transform matrices. 

This gets worse if you have hierarchies, as now each group also needs to have its transformation matrix concatenated and multiplied against your vertex positions.

--
You received this message because you are subscribed to the Google Groups "maya_he3d" group.
To unsubscribe from this group and stop receiving emails from it, send an email to maya_he3d+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/maya_he3d/BYAPR07MB5861DD0F735DCEBBD4D15E45DD639%40BYAPR07MB5861.namprd07.prod.outlook.com.

[joiec...@gmail.com]

Completely agree with that comment.

[Steve Davy]

Intuitively that's what I thought, nice job explaining it though Matt.

---

From: maya...@googlegroups.com <maya...@googlegroups.com> on behalf of joiec...@gmail.com <joiec...@gmail.com>
Sent: Thursday, April 20, 2023 5:39 AM
To: maya_he3d <maya...@googlegroups.com>
Subject: Re: [maya_he3d] Scene DAG question...

 

To view this discussion on the web visit https://groups.google.com/d/msgid/maya_he3d/530649a1-d60f-44b8-b575-fb534af3d2e1n%40googlegroups.com.

[ryan harrington]

If you are rendering with Redshift, it's possible to push the render time from seconds and minutes to hours with one big mesh, or kill your render entirely.  That is about managing the swapping of geometry and textures in and out of the available vram. it's obviously better not to have to swap constantly, and to have headroom for multiple assets to be loaded in parallel for each bucket.
So seperating one large mesh into multiple smaller ones is often preferable.  
AS the GPUs have got bigger, this is less brittle, but still worthy if you like fast renders.

Once Redshift is out of core, performace is off a cliff and everything sucks. This can also be aided by thinking about how your geometry is displayed in Maya, unless you have a seperate GPU for your display, you probably don't want everything drawn in the veiwport when doing a GPU render in the GUI.

I do a lot of environment work, and diciplined use of instancing rules my world....right now I'm loving MASH's distribute node's "initial state" mode...you can plug locators or groups into the node to drive an instancer, (easier to track than Maya instances + instacer node's display options).