If I understand you correctly it would seem you want to have to surfaces, one for the play window and one for the interface which is overlayed on top of the play surface and then this is all flipped to the user. While this would work and isn't too much of a drain on memory it would mean blitting a very large image (the play surface) over to the screen when it could have been drawn to screen in the first place. Since this will happen every frame I'm pretty sure you will lose a lot of frames just doing this. I'll endeavor to write a test program that demonstrates the framerate differences between the two methods.