LUMINANCE SNR MID-TONE:
Looking at DxO the luminance SNR seems to be between 1/3 and 2/3rds of stop better than the 7D, a decent improvement. And basically right up there with the D7100.
7D was relatively well behaved when it came to banding at high ISO so I don't know if you will notice as big of an effective help in that regard above what the SNR tells you as you did comparing 5D2 vs 5D3 though.
I haven't carefully looked into it, but it appears that the JPG engine for in cam developed photos might handle high ISO noticeably better than the 7D.
The DR seems to be a trace worse (although mostly to a small enough degree to not be able to notice real world) than the 7D (although perhaps nearing 1/2 stop worse at ISO200 unless that is just a kink in the graph and within the error bar) until very high ISO where it is a tiny bit better than the 7D.
DR is vastly worse than the D7100 at low ISO and still noticeably worse even at high ISO this time too, in the past Canon usually at least kept up with DR at high ISO, not this time.
It is a shame that Canon is still living way in the past in terms of dynamic range with this sensor.
Once again the measure the masked area thing I did a few weeks back on it proved to be a reliable indicator of DR performance.
Early reports hint that it may have less banding than the 7D which may make the actual effectively usable DR somewhat higher than with the 7D even if it doesn't measure any better for engineering DR though.
COLOR STUFF, CHROMA SNR:
Color discrimination appears to be fairly close to the D7100 (a touch worse under outdoor lighting and a hair trace better under tungsten) and about the same as the 7D for outdoors and somewhat better for tungsten lighting.
Color noise appears to be much worse than with the D7100 and about the same as with the 7D (at least under natural lighting, there is a chance it might be noticeably improved compared to the 7D under tungsten lighting but I'm not sure DxO tests that, I have to check, but I think their color sensitivity only tests daylight conditions? and you can't necessarily extrapolate metamerism index to chroma SNR differences, although often they have tracked, the 70D doesn't necessarily seem to though so it's an open question).
The finer details of what produces better color is a very complex and tricky subject and may depend upon precisely what the question is (say skin tones, vs ocean tones, vs fall colors and whether lighting is outdoor or tungsten, or fluorescent or weak or strong, etc.). Someone who tried to look into things a bit claimed that in the past Canon has tended to avoid nasty peaks or other issues in skin tones for people with certain tones of skin better than Nikon under certain types of lighting often used for shooting people but also tended to be less accurate for colors and not distinguish quite as many overall especially in yellows/oranges. But it is a tricky topic and it surely varies model to model.
PIXEL LEVEL SHARPNESS:
Have not looked into personally, but I know the 7D was a trace on the softer side due in large part to heavily split greens on the CFA. Early reports hint that the 70D might have a bit better acutance than the 7D.