PowerPC in your future?

[Peter Davies]

In the interests of reading a lively rebuttal I forwarded a copy of
Mr.Hartley's original posting to the above lists (courtesy dgh) to John Mashey
of Silicon Graphics. I was not disappointed and he consented for a wider
distribution of what follows. I would like to see this discussion on comp.arch
but it would make no sense to forward the following with out the original
material being posted to that forum.

peter.

From: mashamash.wpd.sgi.com (John R. Mashey)
To: peteranetcom.com, dghavalidgh.com, 70214.1036acompuserve.com
Cc: mashamash.wpd.sgi.com
Subject: Comments from Craig Hartley
Date: Sat, 9 Oct 1993 19:20:56 -0700

This got to me via Peter Davies via one of DGH's lists.

Comments:  (I DON'T WANT TO BE MEAN, BUT I THINK CRAIG IS PASSING ALONG A
BUNCH OF COMMENTS WHERE HE DOESN'T HAVE ALL OF THE DATA ... followed by
some opinions, to which he is entitled, but sometimes look like reaching...)

1) I believe PPC will be important, and I think it will have interesting market
share ... Only time will tell to see if Apple ships 1M-1.5M in 1994.
(Looking at costs, and knowing how APple does things, this seems slightly
unlikely to me, given the cost sensitivities in Apple, and cost
difference between 68Ks and PPCs.)

2) Quoting what analysts think the future will bring to prove it
is NONSENSE [I talk to these analysts, I know where their numbers come from,
and I've seen them turn cartwheels over the years ... remember some of
them thought that a) the Clipper would win, b) the i860 would win,
  c) the Moto 88K would win .... etc, etc.]
Anyone who places strong belief in any one such analyst is being very
naive, or just doing marketing by picking the ones who have the right numbers.
3) Recall that Tom Whiteside is running MIPS Technology, as of a month
or so ago.  I don't think he thinks PPC is going to take over the world,
or he wouldn't have left that job for here...
4) There are many comments in the mail represented as facts ... and some of
them are simply wrong, and easy to check [not matters of opinion like
what analysts happen to think].  They are sufficiently wrong that someone
in the inforamtion chain [not necessarily Craig] is either ignorant or
lying, and Craig may be passing along this mis-information.
5) Some of this sounds like a return to the "good old days" where the only
thing that counted was what IBM did, and one could make a whole career being
an IBM-watcher without ever caring what anyone else was doing.  In those
days, we could often see press releases announcing some new feature that
someone else had had for years, but was then treated breathlessly as
a new innovation.... :-)

SPECIFIC COMMENTS.

1) Major manufacturers Commitments ...
"It will have by far the largest installed base of any RISC processor in
history.  Indications are that it will pass 1M units shipped in 1994,
making it the first RISC processor to pass 1M in a year"
NONSENSE, DEMONSTRABLY.
In 1992, according to RISC Management (but cross-checked elsewhere,
and fairly consistent), unit shipments were:
Intel i960:	2,095K
AMD 29K:	  849K
SPARC		  320K
MIPS		  291K

(Now, all it takes to find out such numbers is to ask anybody who actively
tracks the chip wars; the numbers have appeared in various public sources.)

That was 1992. In 1993, I'd expect i960, 29K, and MIPS to be over 1M
(the latter due to the OKI printers, at something like 100K/month).
I don't know about SPARC's expectations.  i960s, on the strength of
HP laserJets,will certainly ship >2M.

Note also, it said "largest installed base", not just unit shipments;
if PPC does 1-1.5M in 1994 ... that will put them at 3rd or 4th on
1994's shipments, I'd guess ...  and in fact, it's quite possible they
will *never* be #1 in installed base, but it's really unlikely in 1994.

" As if being the standard
processor for both IBM and Apple weren't enough,"
B.S.
It is not the "standard processor" for either of these companies;
If you don't believe that, ask Boca Raton when the IBM PC business switches
from 486s and say so publicly.
Last I heard, Apple expects to build 68K-based systems for a long time, which
they *must* do to maintain their low price points; those folks care about
every dollar, and there's no way to make PPCs as cheap as 68030s for some
time...

"it has also been
adopted as the standard embedded processor for all Ford autos (maybe
two per car)."

a) Are any other RISCs are getting into cars?
	A: yes.
b) Do you know how long the design cycle is for micros in cars, i.e., have you
ever sat in meetings with auto companies going thru the schedules (I have)?
	A: Many years, so volume in next few years is zero.
c) Is there any chance other RISC design wins in cars preceded PPC in Ford?
	A: yes, and for sure, because the FORD win was previously an 88K win,
	so they can't have changed their minds before the PPC plan came.


"It will also find its way into TV cable boxes and sets
through alliances between Apple and Motorola and various media
companies."
Probably .. but I'd point out that there may be other companies already
as far, or further along this track, and unless you are heavily involved
in the deals, you probably have no clue.  Just saying "cable boxes"
doesn't make it happen, and especially doesn't make it #1.

Meanwhile, I'll also point out that Nintendo has sold 100M+ hardware games,
and they'll start having MIPS chips in 1994 (arcades), and late-1995
(home games). I'd suggest the likely volume ramp for that one is faster,
and more than, Ford's, by a ways.  The chips used there are also likely to have
other applications, UNLIKE the Ford ones, which will be fairly special-purpose
versions [as they've generally been so far.]

It takes a while to get embedded control to ramp up [years]; some of the
volume ones are very cost-sensitive;  There is already a $15 MIPS chip
(IDT's R3041), and a $50-%70 (volume) version coming very soon (R4200).


"2) It Can Beat the Competition in Price"
Beating a Pentium is no big deal.  The statements about "DEC and MIPS are
trying hard..."  is B.S., at least for MIPS. Anybody who puts DEC and MIPS
in the same category on discussions of fabrication cannot know very
much about how this is done [i.e., DEC Alphas are done in a fab doing
Alphas and VAXen, with a tweaked-up process;  MIPS are done in multiple
industry-standard fabs that run huge volumes of other parts;  the PPC 601 is
done in one fab (Burlington) using an unusual process that nobody else does,
which is why Moto can't make them.]

Prices have little to do with real costs: 486DX2's make incredible profits
at $500 apiece.

" 3) It Is More "Open" Than Any Processor on the Market"
This is outright marketing B.S. . "Open" gets used way too much.
Some of us actually have multiple sources, right now, for our chips;
several RISC chip families actually have multiple substantial vendors
usingthe chips, and ABI efforts a lot further along than POWERopen.

"5)  IBM is Aggressively Marketing PowerPC to Asian Clone Manufacturers"
Good ... but might be a little late.

"6)  PowerPC will be the Platform of Choice for the First Object
Oriented OS to Hit the Market, Taligent's Pink"

This is amusing, especially if you live around here [Silicon Valley].
Talk about vapor-battles...


"7)  PowerPC Systems may be More Usable than those of the Competition"
"This crude figure will give
you the idea (and will serve as a classic example of the tool
limitations of the current state-of-the-art system I'm using to write
this; new OSs and standards will likely offer graphics, sound, and
video in their e-mail environments)."

IT IS INCREDIBLE VAPOR TO CLAIM THAT PPC WILL BE MORE USABLE.
PPC is a microprocessor;  usability depends more on software.
Try an IBM PowerStation and an SGI INDY side-by-side:
ONE of them has bundled:
	a video camera, and several other video inputs
	various audio inputs and outputs
	a good presentation tool, that includes 3D graphics, images,
		audio, hyperlinks
	smart icons, so that if you stick a CD with [software, audio,
		or Kodak images] the icon shows which it is and invokes
		the right reader program if you click on it.
	meida mail, for including graphics ,etc.
ONE doesn't have t5hese things.

Which company, would say, might be trying harder on user interface and
digital media??

I'm running out of time, so I'll stop here.  The fundamental thing is:
1) PPCs ought to do OK.  Embedded control will take longer than you think,
because PPCs cost way too much for most embeddded applications.
If you think most embedded designers would consider a $200 PPC603
"low-cost" than you haven't talked to them much.
2) It is *not* the old days, when IBM could just do whatever it wanted;
you may have noticed this in the press.
3) When you see analysis mostly coming from one side, in some cases that
side may be rather blind to what other people have been up to for years,
and you can get very distorted views that way.
4) I don't mind if somebody says
	"XYZ will ship N units next year"
		that's their prediction.
I don't mind if somebody says:
	"I like XYZ the best"
		because that's their opinion...
BUT, If somebody says
	"XYZ will ship N units, making them #1"
and whoever says that has no clue of other people's volumes,
and there is a lot of factual evidence that the statemetn is wrong,
then either the person is being ignorant or is telling lies...
I *think* Craig was being ignorant;  I suspect other people earlier in
the information chain knew better....


------- End of forwarded message -------