better way to do this?

i was looking at bump mapping and i found out that its all about changing normals to back things look bumpy
i was wondering would it be costly to performace if i subdivided a mesh and sorted the normals out on that then just slaped a texture on top??
or is there a better way to do this?
thanx

Bumpmapping essiantaly is a tradeoff between fillrate and geometry complexity. Geometry complexity is lowered, thus skinning and all other mesh deforming algorithms that work on the vertices can be executed faster. You also save bandwith when transfering the geometry to the (v/g)pu. The bumpmap can be preloaded. However bumpmaps only fake complexity. That’s why they thought displacement mapping would be a good idea. It applies geometry complexity after the geometry arrived in the (v/g)pu, thus it means not faking complecity and saving bandwith (per frame) anyways.
There are some other issues too…