"One day cellphones will probably be as powerful as today's fastest supercomputers"

Not sure that's true anymore. The fastest supercomputer today can achieve 1.5 petaflops. I could not find exact information about this, but I would expect a phone like the iPhone to be rated at around 0.5 gigaflops (it performs more or less at the level of a 400MHz G3 computer).

Anyway, that's a difference of 3000 times, in other words, you'd have to double the performance of the phone 11-12 times to get it up to the performance of today's fastest supercomputers.

How many generations do you think Moore's law has left? Certainly not enough to double performance 11-12 times. If we're lucky we'll be able to squeeze out two-three more generations of smaller chips before Moore's law is no longer able to be satisfied. We're approaching the physical limits here.

So, supposing someday we do have cellphones with today's super-computer power, this will 1) take much, much longer than it has in the past 2) be prerequisite on changing the fundamental way our computers work.

 Reply

Oops, I just realized. That's not a difference of 3000 times, that's a difference of 3,000,000 times. Such a difference could never be made up for using computers that work on the same physical principles as the ones we have today.

 Reply

Everyone has known that there is a physical limit to the speed/size of individual processor cores for some time (Remember the high end PIVs?). But that doesn't directly translate into the failure of Moore's Law.

 Reply