Optimized Rendering Loop

I''ve once read that the best rendering sequence in a loop is : loop : game/logic process render the current frame present the previous frame goto loop does that mean I have to render my frame first on a different surface ? what about the flipping chain ?? How can we achieve this design in Direct3D ?? A friend told me by doing this we can make sure that CPU and GPU can work paralelly which makes it optimize. is it true ??