iPad 2 would have bested 1990s-era supercomputers

Don't count on having the power of today's supercomputer in a hand-held device in 20 years. From 1985-2010 you have had roughly 12 2-year Moore-units. It takes 2 Moore units to get a halving in feature size, so device sizes are 2^6 smaller than 24 years ago (32 nm today vs. 2 um then). Will devices see another 64x reduction in size? Could you see a 0.5 nm transistor gate length? Maybe, but your channel would be 5 Silicon atoms long, which puts you a lot closer to quantum effect neighborhood.

So much of computing's advances have been wrought out of raw scaling, but that method advancement will necessarily end. You can't make a switch less than one atom wide, and we're getting closer to that limit every day.

Also note that the PowerVR SGC543 in the iPad 2 can run OpenCL (although Apple probably haven't released an iOS API for that yet) and thus could boost the benchmark figures considerably over running it on the ARM's Neon unit.

This doesn't even account for the SGX543MP2 GPU either, which is OpenCL compliant, and therefore capable of billions of calculations per second by itself.

If you really want a cool Apple-based parallel computer, have a look at the AppleCrate: http://home.comcast.net/~mjmahon/AppleCrateII.html

Won't beat any supercomputer, except for coolness factor! :-)

So, the A5 chip could possibly beat out most supercomputers of what'd be essentially the late 80s through to the very early 90s had it run the Linpack test using both cores. Impressive yes but we are talking 20 year old supercomputers, Moore's law of complexity would have put them out to roost long before the 90s were out.

Doesn't the Moore's law say that, if it toke 20 years to accomplish those comparations, by now it will take 10 more years or less to an iPhone's computational capabilities reach the same comparative performance? :P