Does anyone know a good source for information on optimizing matrix, vector, and general 3d math for these instruction sets? Does it really even make a difference? If I use them to build the math functions for collision detection and frustum clipping, will the overall speed improve noticably? Thanks

Check the manufacturers'' websites; IIRC, both AMD and Intel offer vectorized math libraries with vector and matrix routines for their architectures.

How appropriate. You fight like a cow.