It is not easy to "benchmark" video playback, because there is no single parameter to determine ranks. If you find such a review, there will be inevitably tester's subjective opinions mingled together with the data. Plus, video cards are so powerful these days (even cell phones can output 1080p) thus a trouble usually stems from drivers, codecs, source material, and player, etc., not the hardware.
There are, of course, those basic cards that do not have adequate horse power for modern material (especially if you have sensitive eyes). Often times you can fall back on the CPU in such situations unless drivers or players are being stubborn.
AMD and NV are at feature-parity for the moment. (Can't speak for Quicksync because I haven't bothered to use it) Once you go past those laundry list of feature sets, NV cards usually do better handling something newer or less known (read: torrents, foreign stuff,.. and porn, maybe?), often at the cost of higher power consumption - those tends to rely on GPU shaders instead of video processing unit on Geforce. Kepler, however, did shaken things up significantly, in part with better video processing engine and in part with its all around modest power profile.
Trying to guage the performance by CPU utilization is difficult because modern CPUs have robust power-saving mechanism that changes various clock domains in milliseconds. Total power consumption might be one way to measure performance differential but as long as you test with "legal" material, the difference between similar-grade cards will be so minimal that there isn't really a point.
I personally prefer AMD's native color space and its out-of-the-box post processing, but NV is definitely more versatile as of now, thanks to Kepler. NV had been usually a step behind in this area, or used to do things in brute-force ways. (Fermi was the biggest offender) But "now" is what counts anyway, and Kepler's video processing engine gets the nod for its versatility and low power consumption.