A few years ago, when Intel announced its change of direction in chip design, there were a lot of stories written claiming Moore’s Law was dead.
I disagreed. I said Moore’s Law has always gone in multiple directions, that the concept applied to more than mere line density on a silicon substrate. As my own book attests, there are Moore’s Law effects to be seen in radios, in fiber transmission, in hard disk storage, and in optical storage, among other places.
Well right on schedule we’re seeing Moore’s Law effects in Intel’s direction. Having started with 1 processor per chip in 2004, Intel delivered 2 processors per chip in 2005 and this year (wouldn’t you know it) we have 4 processors per chip.
Parallelism, which was developed at the Sandia Labs 20 years ago, is
just one of many factors that have impacted technology design since
then and kept the impact of Moore’s Law going right along. (The idea
of Reduced Instruction Set Computing, stripping down the different
operations a chip could do so it could do more of them, also dates from
this time.)
But parallelism has had a longer play. It started by hooking up some
mainframes together and writing software that would allow them to solve
problems faster. It moved on to racks of Macs, then to the "distributed
processing" period of systems like SETI@Home, and now we have it
on-a-chip.
Best of all, we’re now at the number 4. We’re starting over. And there
is a long, long way to go from here. Just as there is a long long way
to go in terms of energy efficiency, and all the other great inventions
made possible by the original Moore’s Law.
Moore’s Law means exponential growth. Moore’s Law continues onward.
