Taky mě to zajímalo. Cituji někoho z Arduinum fora, překládat to nebudu. Ve zkratce: ATmegy by měly být na 350 nanometrech, PICy na 150 nanometrech a ARM pro běžné embedded zařízení cca 90nm. Atmel arduino like atmega chips are mostly 350nm process, some maybe 210nm. Microchip does 150nm with their pic32mx series, stm32f407 (discovery kit) are 90nm. The smallest the nanometers number the higher frequency the chip can run. The bigger the nm number the cheaper the process. So one chip to produce at 350nm is maybe 500x cheaper to produce than 32nm one. With 32nm you can go @10GHz, with 350nm maybe 100Mhz (ideal situation, not taking into account the chip schematics and wiring too much). The smaller the process the less power it takes at the same frequency. Therefore ie stm32f4 @168Mhz has got lower power than pic32mx795 @80Mhz.If atmel decides to go 32nm with atmega328p, they can reach 2GHz easily, but it would cost them a LOT of money. MCU makers would tell you to sell their chips is very difficult today, so they are not too much keen on such technology moves.. :) Btw I have several 8pin chips in my junkbox running internaly @5GHz.. :)