http://www.dpreview.com/reviews/nikon-d800-d800e/18
So in this review of the D800/E, the link above takes us to the ISO Noise comps. It's close on RAWS (and open to processing app) but on jpegs, the 5d3 betters the D800 overlall and in some cases even bests the D4... Sort of reverses the gloom and doom that the D800 dished out earlier on in reviews...
Actually if we look at RAW only (at least for me Jpeg are not relevant since I will only shoot RAW), until ISO 3200 the mkIII and the D4 are real close but from ISO 6400 till 51000 the D4 is better.
The mkIII does look better then the D800 with non normalized file. Now personally I never understood why reviews normalized the file size (or MP) when comparing cameras like the mkIII and the D800...until I shot with the D800 myself and was trying to compare it with my mkIII file.
When I took the same shot as the mkIII (same composition, focal lenght, exposure, etc), if I crop 100% on both file the D800 file will give me a more magnified view (much closer) which can give the impression that at higher ISO the mkIII is better. Now if instead I look at both file with the same visual magnification, the noise on the D800 is much less then at 100% crop and the output more closely match the mkIII.
They are both amazing, but I am really puzzled how Nikon manage to control noise so much with 36 MP.