Right now I am planning on using one large static VB for all static geometry. However, what about when you have some static geometry using one vertex type, and one using another? Should I split them up into multiple VB's and incur the penalty of SetStream, or should I just use the greatest common vertex type and incur the penalty of unused and extra memory? The only example I can think of where this matters right now is tangents. Non bump-mapped geometry only need position, normal, and tex coord. Bump mapped needs an additional tangent in the vertex type. If I include tangent in the vertex type, even for non-bump mapped textures, it's unused wasted memory. If I split them up, I then have to manage multiple static vertex buffers, and switch between the streams. Trying to decide which option is better. (Or a third option) Thanks.

Check this out:

Here

Sorry...

http://www.ati.com/developer/gdc/D3DTutorial0_Intro_And_Features.pdf