Yep, it is better.
If you use GeDoSaTo, you can choose which scaling algorithm you want to use, so you can downsample with Bicubic downsampling, or Lanczos downsampling and get visuals that are identical to what most people get resizing afterwards in Photoshop.
Also, GeDoSaTo does multi-step downsampling, meaning resolutions above 2x your screen res get downsampled more than once.
Ex:5120x2800 -> 3840x2160 -> 1920x1080 w/ Bicubic filtering
as opposed to
5120x2880 -> 1920x1080 w/ Bilinear
Which provides better results. Along with all the other benefits it offers, if it works for a game, it's basically a straight upgrade!
Regular driver downsampling is limited to a useful ratio of 2x2 generally.
And it only uses a bilinear filter.
However, GeDoSaTo doesn't really have a resolution limit, and better filters to choose from.
But, downsampling a game with noAA at the downsampling resolution still is going to look very aliased. Regardless of Driver based or GeDoSaTo. Even Nvidia's built in OGSSAA modes (1x2,2x2,3x3,4x4,etc) suffer the same issues in a lot of games.
If you downsample, you generally need to combine other AA methods in conjunction to get a worthwhile result (IE: MSAA,FXAA/SMAA,other post processing AA). Unless you are playing your game at insane ratios that are unplayable.
Very interesting. I might have to start using it then.
Thanks for the answers.