Power is a problem and is scaling far less than the node scaling. Raw single threaded performance basically isn't scaling at all. Going from 45nm to 7nm we on

Power is a problem and is scaling far less than the node scaling. Raw single threaded performance basically isn't scaling at all. Going from 45nm to 7nm we only gained about a factor of two in terms of power consumption. Most of that is due to the fact that interconnect and SRAM aren't scaling. The upshot is that it is worth doing almost infinite amounts of calculation in order to avoid moving anything from point A to point B in a compute unit. If you must use memory, DRAM is death so put big amounts of SRAM on chip and make the buses as wide as possible.