Polygon limit in Game Engine????

I know that in the QUAKE2 engine for example,
you should stay under 1300 polys…so what’s
the limit/recommended limit in the Blender Engine?


Depends on your target system, but I’d wager around 10,000?

I’d argue that 10 000 is really only playable on a computer with (working) hardware acceleration. Let that be a factor as well


10K polys for a single character?
Isn’t that excessive…?
Did you mean 10K for the entire
game? :o

wow 10 tousand. how come when i make a simple game and it has around 1k polys then i tex it. it runs so slow, i even have hardware accelerater and a decent comp. but on some of the blender game book examples they have a game with over 3,000 polys with tex and it runs fine! why is that?


Did you mean 10K for the entire
game? :o

He means it for the whole game…

There is no real limit for a single character but recommend around 650-800 triangles. That should be enough for a good character. The less the better (faster). The Blender game engine is kinda slow on some hardware.

When modeling characters you wanna keep them under 1k, although constantly aim for 500 polys and do use blenders wieght paint to your advantage it will always help to get cleaning movement with out more polygons/armatures


so how would the Blender Engine compare to
the Quake or Quake2 engines?

Lol, I didnt mean for a single character, I meant for the whole game. And without a decent hardware accelerator, does blender even RUN? If it does, I wouldn’t want to see it…

The quake and quake 2 engines are much faster because they have CULLING, which wont draw stuff that you wouldnt see anyway.

The quake and quake 2 engines have much more special effects, and better realtime lighting.

The quake and quake 2 engines are slightly more limiting if you plan to just use them as-is, but with coding you can do whatever you want. Blender is more free-form.

It doesn’t
it is another arena, being so flexible, yet not as optimized.
so I guess it does, and I just compared them

Yeah, 800 or 600 polygons is still very big for a character. I made a unreal or quake looking game character to test my game modelling skills, you can view a demo of my character at http://home.hetnet.nl/~jdvhulst/elysiun/human.zip it’s around 389 vertexes and 490 polygons. :stuck_out_tongue: Also add some extra’s, use keys 1,2,3,4 to chance camera. :smiley:

Well, when speaking of polys I’m meaning triangles not quads- and you’ve got over 700. :wink:

Saluk, I thought Blender has this function in it’s game engine. Can’t it be set with the view limit of the cameras? Or are you referring to Background culling, where only what can be seen by the camera is all that is rendered?


Quake and quake 2 use binary space partitioning, which has to be precalculated, but its fast. The only culling blender has, is, if object is too far away, dont draw it. Very inneficient, it still draws objects that are behind you, and it still has to go through all the objects in the whole scene every frame.


On this page you can find a python script for Blender for Background Culling. :smiley: :smiley: :smiley:


unfortunately thats no use for real-time (sorry sonix, thanks for the link though :slight_smile: )

OpenGL semi-automatically handles backface culling if you enable it, unfortunatly that has to be done at the basic(ie source) level in the proceedure that actuallys draws the scene.

i think if something like Octrees or other visibility system were implemented then Blender would compare quite well with Quake 2, its certainly more versatile.


The opensourced engine Crystal Space, has culling, it’s very fast even without hardware acceleration and many other features. We can work on some kind of CrystalSpace based Ketsji-compatible game engine/game player, so we can play our .blends with that player without need of modification.

That looks like a useful script for rendering still images but it wouldn’t be of any use in the game engine. :frowning:

I think it would be great to get proper culling workin in blender. Sure it would mean having to wait for a game to be compiled before you can run it, but it could always be made optional. It would be like a last thing you’d do before releasin a demo.

Unfortuately the human race will be long dead before we see that kind of development in the game engine. :frowning: :frowning: maybe the super-evolved worms would find it useful tho. :slight_smile:

Keith. 8)

Meditating on this, I have been.

I have not had chance to test this idea but;-

Say you parent a very large cube to the camera. Large enough to accomodate everything you want to see, but small enough to exclude the objects that don’t need to be rendered.

Make the cube an actuator, one that makes Blender start to render the objects only within the cube.

The camera could be situated very near the vertical backface of the cube, to give ‘behind’ culling; rotate camera which rotates the cube.

As I say I haven’t had chance to test this theory.

Any good though?


thats the kind of idea, the only question is whether going through all the objects to see which are within the cube would be faster than just using the normal clipping distance on the camera. I guess you could do a similar thing using a radar sensor on the camera, but again the performance increase might not be enough to offset the added cost of checking every object. One of those ‘try it and see’ situations i suppose… :smiley: