David Kanter (dkanter@realworldtech.com) on 2/2/03 wrote:
---------------------------
>As of 1:42am, CST, Paul DeMone's new article, The Battle in 64 bit Land, 2003 and
>Beyond is now up. It's a great read, and I hope everyone enjoys it as much as I did.
>
I bet his fellow Intel shareholders will enjoy it a lot ...

But first, credit where credit is due. Paul has done great
work as usual: the article is complete, detailed, based on
sound data, and very well written.

However, it is not unbiased. The bias is subtle, though,
and I don't want the following observations to be regarded
as an accusation.

Observation 1: the author is very selective when giving
benefit of doubt to anyone. It is actually one of his
strong points to always stick to hard facts. He breaks his
own rule when he speculates about Intel's failure to reach
their own projected frequency target for Itanium-2.

Observation 2: the author is very careful not to spread
Fear, Uncertainty, and Doubt. It is actually one of his
strong points to always stick to hard facts. He breaks his
own rule a few times, for instance when he conjures up
images of an agonizing AMD.

Observation 3: the author sometimes uses very different
presentations for very comparable situations. For example
he goes to great length to explain conceptual limitations
of IBM's Power4 core, which he explicitly relates to
limitations of VLIW architectures. When he does present
VLIW processors as sold by Intel, though, he merely names
in passing their much stricter conceptual limitation of
being in-order-execution designs.

Another example is the paragraph where the author laudes
the Alpha design team (now employed at Intel as he
mentions) and muses about the mythical improvements these
demigods of the semiconductor world will bring to future
implementations of Intel's 64 bit architecture. The other
half of ex-Alpha designers, who happen to work at AMD, are
doomed to oblivion.

Yet another example can be seen when the author talks about
the practice of leveraging existing chip designs for future
derivatives. Both HP and SGI are presented as slow and
undedicated, because they were using the same basic CPU
core design for a sequence of products. The author even
subtlely ridicules SGI by calling this timespan 'half a
decade'. Intel, however, is not at all blamed for the fact
that their roadmap uses the Itanium-2 core for four years.

I could make a few more minor points, but those would have
to be about things so subtle as to be virtually
nonexistant. As I said, this is not an accusation; I know
very well how hard it is to notice and avoid one's own
bias (the few who know me can tell I am biased as well to
PowerPC in general and Motorola's chips in particular).

I would expect, though, that the author's credibility
would be helped a lot if he'd choose to mention his
interest in Intel stock value in some legible footnote.
Especially considering the reach of publications like RWT.
Otherwise, more suspicious people than me might devalue
the author's competent writings and regard them merely as
elaborate pieces of Intel marketing.

Holger Bettag