RTX 4090, AD102-300-A1, 16384FP32, 384bit 21Gbps 24G GDDR6X, 450W, base 2235 boost 2520 actual max >2750;
RTX 4080, AD103-300-A1, 10240FP32, 256bit 21Gbps 16G GDDR6X, 420W,
RTX 4070, AD104-275-Kx(x is a number)-A1, 7168FP32, 160bit 18Gbps 10G GDDR6, 300W.
https://videocardz.com/newz/nvidia-geforce-rtx-4090-to-feature-2520-mhz-boost-clock-almost-50-higher-than-rtx-3090
https://twitter.com/kopite7kimi/status/1543844605531418625?
The only one to be bought for me is Rtx 4070 and that if it has a good price,the Gddr6 memory will allow not that much of overheating and fine tuning its clocks for mining I think we can easily be sure this card to reach near 85-100 Mhsh for just 180-200 watt of power consumption making it one of the most efficient cards,just like the Rtx 3060 ti is the best so far in terms of efficiency.
As for the other two,no matter if they will have a better hash rate dealing with Gddr6x memory and those power consumption is asking for trouble,not finding those interesting for mining unless you live in Alaska or some other cool place.
here is something to think about
density is efficiency, rigs will be capped by PSU, board or connections. it is going to get capped one way or the other, if you add (managing) more rigs/cards--it is your man power (labor+time) who will get capped, now if you hire an extra hand, is that efficiency?
there is no escape regarding heat, burning or consuming watts the by product is heat. if you use 5000w of gddr6 compared to 5000w of gddr6x they will both heat the same room about the same.
the gddr6x issues back then were thermal pad quality and cooler designs, if those are to be fixed in the next gen gddr6x is the way to go. actually in the 3xxx series if you choose the right models properly you can do okay.