The 2080 Ti won't be 80-90% faster than the 1080 Ti, if the 2080 is 20-30% faster than the 1080 Ti. That then puts too large a gap between the 2080 and the 2080 Ti.
Expect a similar jump, or slightly less, than the jump from 9XX series to 10XX series. i.e ~70% per segment.
This is 16nm -> 12nm, which is minor, and then (supposedly) a new architecture. They'd need to have made substantial arch improvements (and/or substantial die size increases) to manage a 90% increase per segment when the process node change is so minor.
7nm is where the very big gains will come.
...
28nm to 16nm was almost exactly a doubling of density (or 100% increase), and also a 35% clock increase (plus the extra clock increase from arch modifications in Pascal).
So the 1080 Ti is effectively like a 952mm2 28nm chip with ~40% more clockspeed. Bearing in mind the 980 Ti was 601mm2
So ~58% more effective die size and ~40% more clocks resulted in 60-70% more performance.
12nm is reported to be roughly 15% more density and 15% more clockspeed. That's a huge difference compared to 100% more density and 35% more clockspeed.
So in order to be an equivalent jump from 980 Ti to 1080 Ti, the 2080 Ti would have to be ~640mm2 on 12nm, and clock to ~2.8 GHz.
And that would hypothetically yield 60-70% performance again.
In order to be better than that, and/or be a smaller die and clock lower, the new arch would have to be substantially better than Pascal per core and per clock.
EDIT: It's also worth noting the Titan V (i.e. the GV100 core) is only 1.6% denser than the Titan Xp core (i.e. GP102). So the consumer cards may be less than even that small 15% more dense.