I'm wondering what is the standard complexity that a model should have so as not to lag in Irrlicht. Presume the context would be that there will be at most 10 of these characters animating around at any time on a culled terrain.
Would the complexity be based on the number of faces, or the number of vertices that the model has? [tri vs quad] I'm guessing an okay number of vertices would be 1000 verts. Am I anywhere near the ans?
Another qn: If I want my character to have interchangable outfits, do I switch outfit meshes in-game and try to attach it to the parent body, or do is it simpler to switch an entirely new character mesh altogether?
EDIT: The 2nd question, I decided it'll be better if I try parenting meshes to fixed locations on other meshes myself first. So you can ignore tt qn.
Thanks in advance.
Model complexity
Model complexity
Last edited by Cardinal4 on Mon Aug 06, 2007 2:59 pm, edited 1 time in total.
Same question here
Good question! I have the same...
And I'm also a .NET advocate.
May I ask you what type of model you use (.3ds? .x?)?
And I'm also a .NET advocate.
May I ask you what type of model you use (.3ds? .x?)?
Your firsth question is hard to answer. It depends on lot of diferent things:
Howe fast you want your app to run?
Howe good users hardware is suposed to be?
Performance does not depend only on polygoon count. Performance is influenced also by number of render orders ...not only each scene node is rendered separately but each mesh buffer does. Means if your node use lot of materials number of render calls is equal to them.
You can simply test efect by creating mesh of 10000 polygoons and then compare rendering 1000 meshes of 10 polygoons each. Number of polygoons will be the same but the second example will kill your FPS quit a lot.
So my suggestion is: make seweral tests and make your choice based on them.
Howe fast you want your app to run?
Howe good users hardware is suposed to be?
Performance does not depend only on polygoon count. Performance is influenced also by number of render orders ...not only each scene node is rendered separately but each mesh buffer does. Means if your node use lot of materials number of render calls is equal to them.
You can simply test efect by creating mesh of 10000 polygoons and then compare rendering 1000 meshes of 10 polygoons each. Number of polygoons will be the same but the second example will kill your FPS quit a lot.
So my suggestion is: make seweral tests and make your choice based on them.
it also depend on machines you aim to.
My guess is...if you want blazing speed response for fps fast shooting, the fewer the better , if your artists can fake higher count in 1000 tris, ok...but there's a clear jump in quality between 1000 and 1500 /2000. (ie, fingers, etc )
If is for rol game seen from a long distance from 45º above, 1000 would suffice.
It also depends on a lot of matters: fx, sound, Ai, terrain, objects, can sink it alone, more if you add the models. Yep, do test in low machines. Not very low if your target audience is fps players, they wont have all a 8800, but perhaps is good to aim from gf4 ti 4200 to above....
edit-- sorry, didnt read arras post...
My guess is...if you want blazing speed response for fps fast shooting, the fewer the better , if your artists can fake higher count in 1000 tris, ok...but there's a clear jump in quality between 1000 and 1500 /2000. (ie, fingers, etc )
If is for rol game seen from a long distance from 45º above, 1000 would suffice.
It also depends on a lot of matters: fx, sound, Ai, terrain, objects, can sink it alone, more if you add the models. Yep, do test in low machines. Not very low if your target audience is fps players, they wont have all a 8800, but perhaps is good to aim from gf4 ti 4200 to above....
edit-- sorry, didnt read arras post...
Finally making games again!
http://www.konekogames.com
http://www.konekogames.com
Re: Same question here
@arras + vermeer:
Thanks for answering. I completely forgot that render speed has to do with the client's hardware too... Actually I'd needed a rough gauge of polycount before I decide how to model my chars (makehuman.org or manually). Guess I have to test the hardware settings when I get more media items to load into Irrlicht.
@tmighty:
.b3d, .obj have worked for me so far. I'm hoping to get .x to work soon [modeling-side error]
Thanks for answering. I completely forgot that render speed has to do with the client's hardware too... Actually I'd needed a rough gauge of polycount before I decide how to model my chars (makehuman.org or manually). Guess I have to test the hardware settings when I get more media items to load into Irrlicht.
@tmighty:
.b3d, .obj have worked for me so far. I'm hoping to get .x to work soon [modeling-side error]