Videos often are often letterboxed or pillarboxed, which means that some color (generally black) needs to be displayed outside them on screen. This means that DWM needs to continue to display the RGBA main plane onscreen just for them, which requires memory bandwidth and therefore additional memory controller and memory power.

The DCLayerOverlayProcessor could detect if all the quads outside the overlay are solid color filled with black. Currently videos need to be converted from NV12 to YUY2 to display onscreen, so this gives us the opportunity to create the YUY2 texture to match the screen's aspect ratio. The area that's not filled with the video could be filled with black. 

We use DXGI_SWAP_EFFECT_FLIP_SEQUENTIAL, so after the first two buffers are presented it wouldn't need to fill that in in the buffers any more, so there would be no increase in work to create the buffers, and they should save a lot of power in scanout.

Ideally we could convince DWM to not display the main plane at all. According to https://docs.microsoft.com/en-us/windows-hardware/drivers/display/multiplane-overlay-hardware-requirements the area outside the video would display black, but I don't know a way to get WDDM not to draw to that area.