I just checked out ISO 25,600 read noise and compared it to the 70D and 7D2 on DxOMark, and it seems to be same as the 7D2 at the pixel level, but potentially slightly better at the image level with more pixels.
Thanks for the news. It looks like there is about a stop more pixel DR at ISO 100 than previous Canons ; not exactly exmor-like, but a good step forward, and the noise character looks excellent, like the 7D2. Pushing the black cloth in the mug as if it were ISO 4000 or so, It looks mostly black with vague hints of color noise, and I can see the individual rows of threads in it.
I find myself slightly disappointed that there's not more DR at low ISOs - but even a 4 stop push on the ISO100 raw didn't show any signs of banding (unlike a 7D MkI image). I guess that's partly down to the noise quantity not being that much smaller, but the noise character being much better (than the 7DI)?
The ISO3200 and 6400 images looked a fair bit nicer than equivalent 7DI shots; much less objectionable noise, but I wouldn't like to call it "n stops better" (whatever n may be).
In any event, I've got one on order as I'm sure it'll be a significant upgrade vs my partner's 60D. I have a suspicion (indeed, expectation, based on the specs) that the AF will better my 7DI.
This reminds me that I'd really like to see some qualitative DR testing of the 1Dx II