View Full Version : Which way is faster?
BuckGFX 08-02-2004, 05:13 PM Hi, guys!
I have a building model. It consist of many meshes and two materials. First material is assigned to one part of meshes, second is assigned to another part.
And I have two cases:
1. All meshes are attached to one whole mesh.
2. Meshes are attached by material. So we have two meshes with different materials.
Question is: which of these cases gives best FPS? Which way is faster?
Any propositions?
|
|
Supervlieg
08-02-2004, 07:52 PM
By my limited knowledge it doesnt make a whole lot of difference. The amount of polies and textures to be rendered by the engine remains the same in both cases.
It's probably down to the preferences build in the game engine. But I doubt there will be a noticable difference between the two.
the rookie
08-03-2004, 03:26 AM
I have to agree with him there, I believe some mentioned about that, maybe if you can (which is a trick to save a poly count) is make on mesh and one texture if you can pull it off, if the engine doesn't have that huge of a limitation, then I wouldn't worry about it, but I believe there is a section that speaks about this as to what happens when you do this, look under the 3D game art section? I think, and it's under the texturing at the beginning, I think they did touch upon it
Tom Pawlik
08-19-2004, 12:58 PM
If you attach after the materials it should give you a speed-boost at a certain point. Texture switching is a state-change is very often time consuming. The same for shaders. Try not to switch textures often and this is the case if you do not sort your geometry by state-changes.
Today i tend to sort my geometry by shaders/number of lights to achieve an optimal framerate, but i also had some levels for my engine, where no sorting at all was faster (Little rooms with only little geometry throughput).
cu
Tom
P.S.: State-changes are time consuming in OpenGL, i don't know for DX, but it is a hardware limit as far as i know, so they should behave the same.
Sashelas
08-21-2004, 12:21 AM
I'd guess two seperate meshes. Specific shaders per mesh would probably be fastest since you could knock the inner loop instruction count down if the materials differ. Also the shader would pass the material over the GPU.
If one mesh is out of view, it doesn't have to be considered giving you a performance boost.
It depends on your mesh complexity as well too. If it is just huge, I'd chop into into pieces. If it is usually on screen, then one mesh might be better.
It is the most flexible as well. Do you have different characteristics for the materials?
CGTalk Moderation
01-18-2006, 10:00 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.
vBulletin v3.0.5, Copyright ©2000-2012, Jelsoft Enterprises Ltd.