Imagine how much better it would be with four high performance, and four energy efficient, cores, with a unified memory architecture that didn't require predicting diddly as it juggles data through the bus slowing things down significantly. And if it used TSMC's 5nm process (or it's upcoming 3nm...