Apple and CPUs

At 1:49 PM -0800 12/7/04, Dan HINCKLEY wrote:
>Has anyone been following this discussion, or do any of you know more than
>this article addresses about Apple and the IBM chips?
>
>http://www.linuxinsider.com/story/34994.html

I suspect it's horse puckey.
The author is spinning conspiracy theories.

Here's his take on IBM not shipping 3 GHz G5 processors (not just to
Apple, but to anyone):

>Whether that's true or not, however, my belief is that IBM chose not
>to deliver on its commitment to Apple because doing so would have
>exacerbated the already embarrassing performance gap between its own
>server products and the higher end Macs.

See any facts there?

--
Marshall Clow Idio Software <mailto:marshallidio.com>

It is by caffeine alone I set my mind in motion.
It is by the beans of Java that thoughts acquire speed,
the hands acquire shaking, the shaking becomes a warning.
It is by caffeine alone I set my mind in motion.

Dan HINCKLEY wrote:
> Has anyone been following this discussion, or do any of you know more
> than this article addresses about Apple and the IBM chips?
>
> http://www.linuxinsider.com/story/34994.html

Just today in the Register and at Slashdot there was some more
speculation about Apple's future:

<http://it.slashdot.org/article.pl?sid=04/12/06/1445209&tid=136>
<http://www.theregister.com/2004/12/06/ibm_apple_speculation/>
<http://www.theregister.com/2004/12/07/ibm_pc_sell_off_power/>

As someone who uses Suns, Macs, and Linux boxes on a daily basis, I hope
that Apple stays away from Sun, since I don't see much of a future for
Sun. (Yes, I know, "imminent demise", but at least it's not about
Apple...) I know we'd get rid of our remaining Suns at work if it
weren't for the fact that there's a piece of software we need to run
that runs only on Suns at the moment. Sun also doesn't seem to be able
to do something as simple as run a security mailing list, and their
software update procedure is awful, no where near the ease of Apple's
Software Update or Red Hat's up2date. (Maybe Solaris 10 will be
better.) A partnership between IBM and Apple, on the other hand, would
be interesting.

--
Paul Schinder
schinderpobox.com

On 12/7/04 at 1:49 PM, dahhq.iucn.org (Dan HINCKLEY) wrote:

> Has anyone been following this discussion, or do any of you know more
> than this article addresses about Apple and the IBM chips?
>
> http://www.linuxinsider.com/story/34994.html

I'm not a chip expert by any means, but from what I have picked up on
the subject (mostly from articles on places like Ars Technica), I don't
trust this article's analysis.

Some points that immediately came to mind after reading the article:

* The author makes claims about the Cell processor by stating "The cell
processor, furthermore, is *confidently* [emphasis theirs] planned for
mass production at 65-nanometer sizes early next year." By my memory,
the G5 was just as confidently planned to be running at 3 ghz by now, a
fact that the author even mentions a paragraph or two above. In other
words, it's a claim that has just as much value - or as little.

* The author also made no mention at all of the fact that the G5 is
based off of the POWER architecture used in IBM's own high-end servers,
or frequent speculation that a next-gen desktop chip is in development
bearing the same relationship to the POWER5 as the G5 bears to the
POWER4. I'm not certain myself how that would fit into the
Cell-vs-Sparc-vs-Intel-vs-G5/PowerPC comparison the author's making, but
it's certainly a relevant and obvious linkage; ignoring it so completely
suggests to me a deliberate withholding on the author's part, which in
turn makes me think he has an ax to grind. This is especially true when
you consider the implication he makes that Apple has no choice but to
move to the Cell if they want to stay with PowerPC, as well as his
accusation that IBM is deliberately holding back on the G5 chip.

* The author makes very favorable claims about the Sun Sparc
architecture, without giving specifics, or indeed any other basis that I
can tell beyond the impression that he likes the Sparc architecture and
the people who work on it. This is at odds with almost everything else
I've read about Sparc; the impression I get from the news sites I read
is that it's a dying architecture, though by now we're getting into
areas where I'm not even an 'informed' layman. :)

* The author claimed that although the Cell processor is broadly PPC
compatible, the lack of Altivec and differences in
graphics/multiprocessor models would actually make it easier to port to
Sparc than to Cell. Again, I'm not even really an informed layman on
this point, but I *do* remember the 68K-PPC transition; from that
experience, I think his claims for the ease of porting to Sparc are
exaggerated at best, and more likely complete bunk.

"In addition, existing Sun research on compiler automation suggests that
multithreaded CPUs like Niagara and Rock could automatically convert
PowerPC and even MC68000 executables to Sparc on the fly -- meaning that
"fat binaries" would not be needed, although a Mac OS 9.0 compatibility
box would probably still make sense." Remind any other old-timers of the
claims for the 68K-PPC emulator, that with the extra speed of the PPC
was supposed to run even old 68K programs faster than a native 68040
processor? IIRC it took at least two generations of PPC Macs to fulfill
that promise, closer to three, and that with a precompiling emulator
instead of the original interpreting one. It's certainly possible that
Sun made enough improvements since then that this time it would work
better, but the only evidence we're given here of that (or any other
Sparc benefits, for that matter) is the author's assertion - and after
the way he fudged other points I do know a little about, I don't trust
him.

In short, the impression I got was that the author was biased in favor
of the Sparc architecture, biased against the IBM POWER/PowerPC
architecture, and picked facts to fit his prejudices instead of
analyzing with an open mind.

>Has anyone been following this discussion, or do any of you know more than
>this article addresses about Apple and the IBM chips?
>
>http://www.linuxinsider.com/story/34994.html

I've not been following this issue any, but I read the article.
Quickly he argues that IBM could've hit 3GHz with the G5 but chose
not to. Instead IBM is more strongly pursuing the cell processor for
Microsoft's X-Box, because Apple currently outclasses IBM in server
speeds, and IBM is embarrassed over this. He goes on and states:

  The bigger issue is that although the new cell processor is a
PowerPC derivative and thus broadly compatible with previous Apple
CPUs, the attached processors are not compatible with Altivec and
neither is the microcode needed to run the thing. Most importantly,
however, the graphics and multiprocessor models are totally different.

It then wanders on to state that Apple and Sun should start a joint
partnership, because it gives Apple a good inroads into the corporate
desktop, and would be a good match for Sun because it would allow
them to showcase their throughput computing initiative.

In general I consider most punditry about Apple hogwash. Most of it
comes from non-Apple types who seem to not realize that Apple is
strongest when its most independent, and not tied up in "joint
ventures." (Yes, the current processor platform came out of a joint
venture of Motorola, IBM, and Apple, but remember the lack of success
of the Common Hardware Reference Platform, which was the long term
goal of this project.)

You know you get Gems that IBM is selling their PC unit, and will now
be able to set up a joint venture with Apple to sell Apple's products.

<http://news.com.com/An+Apple+rumor+a+day/2061-1016_3-5482136.html?part=rss&tag=5482136&subj=news.1016.5>

This isn't completely out of the ballpark, but it doesn't play to
IBM's strengths, and Apple really doesn't need IBM selling their
computers.

Nick

Wow, that article is orthogonal to everything else I've read or
heard. I don't read Linux Insider, nor have I heard of that author
before but my guess is that he is just trying to stir up s**t. Kind
of takes me back to the old days, before the recent rise of Apple's
coolness quotient, when every article about Apple that you came
across claimed that bankruptcy was just around the corner.

My understanding is that (1) the G5 line is not going away anytime
soon and (2) the Cell processor is not intended as a CPU replacement
for general purpose computers but as a co-processor. It is also
being hyped for the gaming market.

Best,

Ken P.

I agree with the other comments.

The whole premise of the article is questionable. There's no
evidence that Apple is going to be stuck with the G5, just as there's
no evidence that the x86 is at the end of the road. Intel in
particular has a lot of resources it can bring to bear on improving
the x86. (The PowerPC was supposed to easily surpass the x86, but
that didn't happen.) Also, you can probably find more analysts that
feel that the SPARC platform is doomed than feel that x86 is doomed.

The issue about Altivec is wrong. Mac OS X and most OS X apps
already support CPUs without Altivec (aka the G3). Correctly written
apps should test whether Altivec is present and optimize accordingly.
So if the Cell processor follows the basic Power PC architecture
spec, but has a different kind of attached processor, then there
shouldn't be a significant compatibility issue. Apple and third
parties can optimize their software for the new CPU at their
convenience.

If the Cell processor doesn't follow the basic PPC spec, then there
might be compatibility issues, but these can't possibly be any worse
than switching to SPARC. If the differences are small, then Apple
can make up the differences with emulation, or IBM might be willing
to make it compatible in exchange for Apple adopting the processor.

Despite what the article says, it would take more than a recompile
for Apple to move to SPARC. And the idea that SPARC is better suited
for Mac OS X because "PostScript display support is well established"
doesn't make any sense. Also, Sun has taken over the title of
"beleaguered" from Apple, so I find it hard to believe Sun could end
up being Apple's saviour in any sense.

--
Larry Rosenstein
lsralum.mit.edu

On 9/12/2004 3:39 AM, "Travis Butler" <tbutlerbirch.net> spake thus:

> would actually make it easier to port to Sparc than to Cell

He also seems to be under the impression that Mac OS X still uses Display
PostScript, as he mentions that Sun's PostScript imaging architecture would
make porting Mac OS X to SPARC so much easier.

The guy clearly doesn't know what he's talking about on important details.
To my mind, this entire article is an obvious troll.

--
Nigel Stanger, Dunedin, NEW ZEALAND.
http://public.xdi.org/=nigel.stanger

My personal feeling is that IBM backed off going for speed increases as
quickly because both AMD and Intel have done so too. Intel will not be
shipping a 4 GHz Pentium 4 processor and supposedly will begin emphasizing
other features, especially mobility features, over pure GHz.

<http://arstechnica.com/news.ars/post/20040313-3528.html>
<http://arstechnica.com/news.ars/post/20041014-4311.html>

Kevin