Hey people, I was wondering if some people here have some knowledge of hardware and/or experience with GeForces (I don''t). I''ve been checking a lot of info on dynamic lod reduction. Some of the argos are pretty kewl and effective, but they are also pretty expensive, CPU-wise. I was wondering: is it worth doing poly-reduction with the CPU, if your GPU''s got extreme poly-pushing power? Please elaborate...

To answer this question you have to ask yourself if your poly-reduction algo is only for your own purposes or is a part or an engine which you want to be used by another people as well. I think it is still worthy doing it with CPU especially in the second case. If you want another people to run appliations based on your engine you must think about those people who doesn''t have GeForce cards yet. However it''ll be interesting doing it in hardware. If you will decide to implement poly reduction with GPU please let me know. I would like to see the difference.

Actually, I didn''t mean do LOD reduction in hardware. What I meant was: just don''t reduce LOD at all. Make the GPU do all the hard work. This might actually be faster than dynamically lowering the poly-count.

Tell me if I''m wrong, but here''s my thought:
Low poly count means less transformations (which is hardware accelerated on GeForce, and soon will on any board, I guess). Fill-rate won''t be drastically altered (but that''s not the point). What I mean is: would it be beneficial to just let the GPU do all the hard transformation work in parallel to your CPU doing other gamecode, instead of using the CPU to reduce the polycount (needed transformations) and let the GPU do less transformation --which might be less than it actually could do in the time it took the CPU to reduce polycount.
Confused? I don''t hope so...

there was a decusion recently over on flipcode about it.
bascially for GPU''s it all depends on the app. eg with flight simulaions you have to use LOD even with a geforce3

http://members.xoom.com/myBollux