Was Moore's law, now something else, parallelism

[bogstad-e+AXbWqSrlAAvxtiuMwx3w at public.gmane.org (Bill Bogstad)]

On Tue, Jul 13, 2010 at 12:52 PM, Richard Pieri <richard.pieri-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org> wrote:
> On Jul 13, 2010, at 9:48 AM, Mark Woodward wrote:
>>
>> Any thoughts?
>
> I think that you're stuck on clock speeds. ?The original Core architecture breezed past Pentium 4 at lower clock speeds and significantly lower power consumption. ?Each successive iteration of Core has been faster out of proportion to the clock speed increases that you've been conditioned to expect. ?There have been, and are, comparable developments with POWER and ARM architectures, and that's not even beginning to touch on what's happening with GPUs.

Faster computing and "2x faster computing every 18 months" are not the
same thing.  Of course, it was officially never 2x faster computing
anyway it was more about # of transistors.

I don't know exactly how much faster i3/i5/i7 CPUs are on single
threaded code vs. the original core 2 CPUs.  However, everything I've
read in the last ten years, supports the thesis that single-core
performance increases are not occurring at the same rate as in the
past.   Whether it's because increasing the clock rate would fry the
CPU or the designers can't figure out a way to use more transistors to
increase the speed of that multiply instruction isn't really
important.   The result is that programmers can no longer assume that
their code will be speed up due to hardware improvements without them
requiring any changes in the program itself.

Bill Bogstad