NetNews Usenet Archive 1992 #26/NN_1992_26.iso/spool/comp/sys/mac/hardware/21250

Text File | 1992-11-10 | 4.6 KB | 92 lines

Newsgroups: comp.sys.mac.hardware
Path: sparky!uunet!spool.mu.edu!yale.edu!yale!gumby!destroyer!sol.ctr.columbia.edu!usc!cs.utexas.edu!oakhill!matth
From: [e-mail] (Matthew Holle)
Subject: AIM Alliance (was RE> 68050)
Message-ID: <[e-mail]>
Sender: [e-mail]
Organization: Motorola Inc., Austin Tx.
References: <[e-mail]>
Date: Tue, 10 Nov 1992 16:21:59 GMT
Lines: 80

[e-mail] (John L. Coolidge) writes:
> At the time of the alliance, the Alpha was either not announced or had
> been just announced. It probably was never in the running.
> Politically, it would probably be impossible for Apple to base its
> next hardware generation on Intel processors. The R4000 is a nice
> chip, but at the time there were grave doubts about MIPS' survival. In
> addition, MIPS (even now) has nowhere near the resources to pour into
> chip design that IBM does.

Actually, it's even better than that.  IBM and Motorola are working
together on the design for PowerPC.  In the case of microprocessor
design, the sum of the two is be better than each taken seperately.
(I say this considering the cost to design a single generation of
a microprocessor architecture...).  Combined, the design resources of
IBM and Motorola have a large momentum.

> The Sparc might have been a decent
> alternative, but in mid-1991 the Sparc was mired in uncertainty. Even
> now, the Sparc line is pretty sluggish comparatively. Finally, HP's
> Precision Architecture line might have been interesting -- it's fast,
> it exists now, and it sells pretty well. The decision to go with IBM
> rather than HP certainly had marketing/political considerations;
> technically, they're both not too bad.

Actually, Apple was pretty much locked into another architecture
(or so the other company thought) before the AIM alliance was announced.
This other RISC architecture was pretty much built to Apple's spec's
and was much cleaner/faster than the RS6000 was back then.  Apple's
decision for a derivative of IBM's original POWER architecture was
SERIOUSLY based on "marketing/political considerations".

And rightly so.  In this bussines today, no one company can recreate
the PC frenzy of the mid 80's.  The industry today is blood-thirsty.
However, when three of the industry's biggy's get together to make
a standard, people pay attention.  Not IBM alone, not Apple alone,
not Motorola alone could have created the wave that's comming down
the line.

As for HP's line, I may be wrong, but PRECISION is also a multi-chip
implementation (given that they have always used off-chip cache).
It's hard to get multi-chip implementations on the desktop for cost
reasons.

> The real question is going to be: how fast can Apple/IBM/Motorola
> scale the PowerPC line compared to how fast Intel can scale the 80x86
> family (P5, P6)? The 601 should be about the same speed as the P5, but
> first shipping machines will be about a year later. However, there's
> hope of the 602/3 happening before the P6. The 620 should happen
> before the P7. The second and third generation PowerPC chips should be
> significantly faster than the respective Intel chips.

Your PowerPC part numbers may be a little off here, but your point is well
taken.  There are three big factors that influence how well/fast an
architecture will scale (in my humble opinion):
1. The architecture itself and the baggage it's carrying at any time.
   Intel will loose on this point as compared to PowerPC,
   if for no other reason than the x86 line is much older and baggage
   accumulates over the years.
2. The quality of design teams involved.
   No doubt that Intel has some of the best designers in the industry.
   I also have no doubt that IBM/Motorola also have some of the best
   designers in the industry working at Summerset on PowerPC.
3. The resources that are available to pour into design efforts.
   Again, Intel will lose on this point as compared to PowerPC.
   In my humble opinion, it has always seemed like Intel has spared
   little expense in getting designs going as early as possible and
   keeping the schedule moving alone.  Unfortunately, Intel is now
   competing against *both* IBM and Motorola's resource pools.

> The next few years in the CPU business are going to be interesting...

You can say that again...  I think end-users will be pleasantly suprised
with the power of RISC in a desktop Mac.

--------------------------------------------------------------------------

#include <std_disclaimer.h>             Matt Holle
                                        Motorola RISC Applications
                                        [e-mail]