There must be some magic going on here. Some reconstructions have nothing but pixel mush for the internal res and yet we get a very clear and accurate reconstruction. Example: the five small vertical stripes on the helmet or the five small dots on the thigh armor. I wonder if DLSS allows for a temporary/selective higher internal resolution if the image is stationary to improve upsampled output.
https://youtu.be/_ja-31bYFTs?t=735







