3000-5000 polys is a LOT. Any competent modeler can make a quality humanoid in 1000-1500 (or less), which should be sufficient for something like an FPS where enemies and such typically aren''t real close to the camera.

i would think that targeted polygon count would also very greatly depending upon your targeted platform. if you plan on targeting next generation hardware only, then high polygon models/terrain, lod use, texture size, texture layers, etc., might not be as big an issue as if you''re planning on targeting previous generation and current generation hardware. in other words, this sounds like pre-optimization to me. i''d worry first about getting the engine/models/terrain done. you can always downsize the models/terrain pretty late in the game, can''t you?

Halo 1 actually used a fair amount of bump mapping, but 3000 polys is Unreal Tournament 2003 quality, so if your engine can do that, go right ahead, but expect to be on lower resolutions if you want to have say, 10 on screen at once, im no genious, and i know nothing about rendering engines, but would it be good if the renderer only rendered the polys that are showing? i mean, the backside of a character is still rendered(just not displayed) I think... itll probably make you dumber if you read my ignorant posts

I''''m American and damn proud of it.