Consider this image by the wonderfully talented
Stephan Martiniere.
Now I'm thinking about trying to create this in the UDK for portfolio purposes. I will, of course, credit Martin for his concept and provide links etc.
The issue of a triangle count is a bit confusing though. If I were producing this for a game I would obviously be LoDing the hell out of most of the static meshes. There are also a lot of repeating shapes and the Unreal engine supposedly can draw repeating meshes at only marginal increase in resources to the first (instancing, I guess?).
Doesn't this make a traditional hard polylimit kinda obsolete? Is there a better way to produce a scene to be friendly to framerates?
Replies
never use more than one material per object and put as much geometry in one object as you can. but still make it reusable if possible and don't include some tiny piece that is far away in it just because it has the same material ... what i mean by that don't unnecessarily increase your bounding box size beyond the actual dimensions of the majority of your object other wise it might screw with the culling and the object might draw even if it is off screen
hardly any game these days is vertex transform bound meaning if people played their cards right with pixel cost of each shader and over draw from transparent shit ( like large dust particle clouds in the air in front of your face) they could probably all easily push twice the amount of geometry than what they currently do !