Those specs don't make any sense at all - Maxwell was already build mainly with high frequencies in mind, Pascal build upon that even further, and both were extremely successful because of that, so it would be really hard to believe NV all of a sudden cut the clock speeds in half basically, just to to build those massive GPUs, just to gain... 1-2TF literally, compared to 2080/2080Ti.
Which speaking of, a 2080Ti already does even 100-150FPS at 4K, and you need one of those 1000$ displays to actually utilize all that raw power, so again, that's would be the point of making such big chips for little-to-no performance gains? I understand that going to 7nm process will allow them to pack more CUDA cores on the same die size, but not for such a clock penalty, if anything, they would opt to keep the high clocks and just slightly increase in SPs over Turing GPUs, resulting in more performance but on smaller dies, which translates into more profit. Seriously, an OCed 2080Ti already puts out 16-17TF, this rumored design looks like a serious step back.
My bet would be using the 7nm to seriously boos the RT performance, because that's what's seriously lacking at their current GPU lineup, going from those mentioned 100-150FPS at 4K to barely 60 at 1080p means nothing less that the RT performance needs to be boosted roughly 4-5 times, and I'm not sure the 7nm process node can even allow for that TBH.