Sadly most of the links to read-world images (the wettransfer ones) just give me "Oops - file expired" 
.
jwcdds wrote in post #15449162
1.5 stops improvement. Though without having 2 files to compare side-by-side, and exif stripped, it's not very easy to tell which camera took what photo.
So if it matters to you, as the photographer, then get the 5d3 over the 7d. Otherwise, just go out and shoot and be happy with the gear that you're shooting with.
Having done a fair bit of reading up recently, I'm more than happy the 5D3 is going to be the better camera (in terms of AF, noise etc.) but I wanted to see how it fairs with identical real world shots instead of test charts.
What raw files I have managed to get (5D3, no matching 7D shot) seem to have a much nicer grain structure than those from the 7D, and having pushed a couple of images with shadows, they appeared to take much more punishment (vs the 7D) before banding started to appear. Obviously not remotely in the same league (in terms of detail/lack of noise) in shadows as a Sony-sensor-equipped Nikon, but I've accepted that a change to the dark side isn't practical.
EDIT: I downloaded some of the raws from the imaging-resource link from the other thread (thanks Broomer). Difficult to make definitive statements given they're being examined in isolation, but those, and some "test card" type comparison images (from dpreview IRC) do point to the 5D3 being less susceptible to the dreading shadow banding than the 7D, and yea, maybe 1.5 stops noise improvement at best.
