Intel's downs and ups

willbill wrote:

correct me if i'm wrong; IIRC each of Intel's past
major CPU transitions weren't without problems

nothing like being a guinea pig with
new technology (whether it's h/w or s/w)

Well, Intel clearly went down a technological dead-end with their
"Netburst" architecture, with it's design goal of "performance via
high clock rate". I think that everyone agrees that their more-recent
designs, from the Pentium M up to the Core 2 Duo, are designed much
more intelligently. I think there's a couple points to be made beyond
the obvious "they were getting their butts kicked and needed to do
something".

1) Netburst just had to be the result of Intel marketing's demand to
have the highest GHz numbers. I can't believe that they thought that
it was really the optimal engineering solution, especially when power
requirements are factored-in.

It seems the world is now over this "faster clock = better" nonsense.

2) In the past, a new Intel CPU architecture was expected to last
three process generations, and, because of that, it seemed that their
new architectures didn't really "hit their stride" until the second,
die-shrunk generation. This resulted in first-gen products that were
hot running and mediocre in performance. With the Netburst CPU's the
third generation proved to be a bust as well (which was their wake-up
call).

It seems that Intel has now accepted that, in order to be competitive,
they need to redesign more often, so that their new designs work great
right from the start and so that they are not stuck with old designs
that are past their use-by date.