This implementation is greatly simplified compared to the older implementation where framebuffers were attached to output targets. Instead, the graphics pipeline object will create new framebuffers on the fly. The suggestion is still to use dynamic rendering for best performance.