So the problem with FSR 1.0, was it was useing a Spatial upscaling technique, limited to just the current frame.
In praktis this ment it looked fine while standing still, but things in motion, or movement would cause shimmering.
Thats delt with, when you use a temporal solution (like nvidia did long ago).
Another thing, is the image quality is vastly improved, since your able to use more information, instead of those just limited to the current frame.
So image quality sees a vast jump, in quality compaired to FSR1, at only a slight loss to performance (still vastly better than running native).
This is a step, that takes FSR much closer to DLSS than it previously was.
Hardware unbox's take: