Since the performance overhaul project, there should be essentially no difference between rigged meshes and non-rigged as far as vertex buffer management is concerned.