Re: 11 most influential microprocessors of all time

["Michael J. Mahon" <mjmahon@aol.com>]

Eric Rucker wrote:
> On Sep 2, 11:06 pm, "Michael J. Mahon" <mjma...@aol.com> wrote:
> > PA-RISC shook up the industry in the 1990s, by pushing performance
> > harder than most thought possible.  And it was the major factor in
> > catapulting HP into the server and high performance workstation markets.
> >
> > Unfortunately, having done its job, it's largely retired now...but
> > it was a heck of a run!
>
> More to the point, how did PA-RISC influence the marketplace in the
> long run?

Personally, I wouldn't nominate PA-RISC into that club, either,
but I appreciate an honorable mention.  ;-)

HP was certainly not alone in advancing the performance curve, but
we were a major player, with several "firsts" that helped propel the
industry "performance wars"--all of which led to our present position:
single-processor performance pushed to the practical (power) limit.
Certainly we would have arrived here with or without HP and PA-RISC.

As Matt noted, HP was the first to incorporate SIMD instructions for
operating on pixel data (though the capability was first introduced
in PA-RISC to support fast BCD arithmetic for legacy applications,
and my first exposure to the idea was in MIT's TX-2!).

Of course, there was a fair amount of independent invention going on,
but the PA-RISC architecture permitted the "multimedia" instructions
to be implemented with almost no impact on silicon area or timing.
(If they had required any significant implementation investment, I'm
sure that their implementation would have been delayed.  ;-)

Some significant software technology was also inspired by PA-RISC.
A couple of things that come to mind were the combination of software
prefetching and "dismissible" loads, and what I think is the first
application of dynamic optimization to an operating system kernel.

The problem with proprietary technology is that it has relatively
few progeny compared with "volume" technology.  ;-)

> In the long run, PA-RISC got killed in favor of Itanium, and it was
> all but forgotten about.

Much of PA-RISC lives on in Itanium.  HP's compatibility strategy
was always predicated on code translation and dynamic optimization,
so all the semantics of PA-RISC are close at hand.

The transition from PA-RISC to Itanium was motivated by the foreseeable
exponential increases in fab costs with each "node", coupled with the
potentially huge advantage of moving to high-volume chips.

> Every single chip on that list except for possibly COSMAC and the 8088
> strongly influenced the marketplace in some way. 4004 for being the
> first, 8080 for making microprocessors practical at all for home
> computers, 6502 and Z80 for making them cheap, 68000 for making
> personal computers with GUIs practical, Pentium for marketing the CPU
> rather than the computer, and ARM and PPC for ramming themselves
> deeply into the embedded space. The 8088 was lucky (whatever was in
> the IBM PC would have probably dominated the market, therefore, the
> 8088 wasn't what was influential,) and the COSMAC... yes, it was in a
> well-known embedded application, but what else did it really do for
> the market?

I agree with your assessments.

...we now return to your regularly scheduled Apple II newsgroup...  ;-)

-michael

NadaNet 3.0 for Apple II parallel computing!
Home page:  http://home.comcast.net/~mjmahon/

"The wastebasket is our most important design
tool--and it's seriously underused."