So I just made a quick Port Royal benchmark with my 338w 2080 Ti and compared it with Gamers Nexus 3080 FTW3 420W OC.
The GPU-Z Tool showed power draw stats of chip, board and MVDDC (memory I guess)...
So the chip of the 2080 Ti consumes 261W, while the Board does use 315,9W. (chip does use 82,6%, MVDDC 37W, 14,18%, rest 3,22%)
The Chip of 3080 does use 233,2W, while the board dues use 404,2 W (chip does use 57,7%, MVDDC 96,5W, 23,9%, rest 18,4%)

So the GA102 chip seems to use less watt at higher board power draw vs TU102 chip at lower board power draw.
So thats probably the reason, why the chip does not get really hot.
 
Thats also imo the reason why Samsung 8n is no problem, the chip does not get to hot. It is the rest of the board.
 
I would also say with GDDR6, low clocks and some tuning the GA106 will be pretty efficient in laptops, because some people already memed about Ampere in Laptops.

might be the connections on the pcb to transfer stuff fast?

Sajin
Hmm, so the 2080 ti is using 350w while the 3080 is using 430w.

Looks like it's a 2080 Ti running at 315w vs. a 3080 FTW2 with the 400w pre-release bios, maxed.  The TU102 is pulling more power and current at a higher voltage.

watts:fps is not exactly linear

yes thats the strange thing for me: when the rest of board sucks so much power, undervolting the chip should not do so much in power, because rest is still high.... so somehow the rest of the board scales together with the chip? Interesting

Check frame chasers YouTube channel, he has an in depth Turing vs Ampere voltage/power draw chart that he tested himself.

GDDR6X for the 3080 was a bad idea. Nividia should have just made it with 12GB of cheaper GDDR6.... same performance, more VRAM, less power draw.