Can we have a moment of silence (or several million dollars) . . . please?

Douglas Eadline deadline at
Mon Jun 25 17:38:50 PDT 2001

On Mon, 25 Jun 2001, Bob Drzyzgula wrote:

> On Mon, Jun 25, 2001 at 03:21:32PM -0400, Thomas Lovie wrote:
> >
> > it's unclear which definition of 'better' will be used...
> I believe that 'more profitable' is the only definition
> of 'better' with any relevance to the situation.
> Clearly, without competition, Intel, HP and now Compaq
> would be able to just about set whatever prices they happen
> to like, while trickling improvements out at any rate that
> happens to be convenient.
> Luckily, competition exists, although it is not so
> varied as what had existed previously. I just checked
> with a recent copy of Microprocessor Report. In their
> "Chart Watch" for "Workstation Processors", they list
> nine currently-shipping CPUs: 833MHz Alpha, 1.33GHz
> Athlon, 552MHz PA-8600, 450MHz Power3-II, 1GHz PIII,
> 1.7GHz P4, 400MHz R12000, 480MHz UltraSPARC-II and 900MHz
> UltraSPARC-III.  From an SPECcpu2000 (both  performance
> standpoint, only five were competitive: Alpha, Athlon,
> PA-8600, P4 and UltraSPARC-III.
> So now Alpha and the Precision Architecture chips will be
> phased out in favor of the Itanium; MIPS is dead as far
> as Compaq is concerned; other than Compaq and SGI does
> anyone else still use them at the high end (i.e. not in
> embedded designs)? AMD has the Hammer due late next year.
> UltraSPARC-II and the PIII are dead ends.
> At this point, it seems as if AMD remains the sole leading
> competitor from a raw performance standpoint, with Sun,
> IBM & Motorola giving a strong showing but with a slightly
> different emphasis (i.e. no one but AMD seems interested
> in slugging it out in the commodity CPU arena; Sun, IBM
> and Mot all seem more interested in selling integrated
> systems or partnering with major OEMs -- they are addicted
> to fat margins and appear to have no ability to transition
> to a volume-based strategy).

Like it or not, the commodity CPU arena is what pays for CPU
development.  Ask yourself these questions:

1) How many Unix Workstations are sold a year by Sun, HP, Compaq(DEC),
   IBM (combined SPARC, ALPHA, PA, Power) ?
   (about 1 million)
2) How many x86 (AMD/Intel) desktop systems/notebooks are sold each year.
   (about 120 million)
3) Take the cost to design and fab a cpu (>$1 Billion) and divide
   by number of units sold and see how much you need to make
   a living.

The fact that the modern CPU fab costs upwards of $2-3 billion
(yes that is "b") is the problem. You need volume. Why do you think
HP got behind the Merced?

Interestingly, I heard this story at a Convex Users Group meeting
(1996 I think) from some big wig at HP. They predicted only three
CPU families would survive. It was the same meeting where they
had some guy named Tom Sterling talk about this thing called Beowulf and
some graduate student who wrote an x86 Unix and a some guy named Don
who was writing Ethernet drivers for it....

> Is this this enough competition to give Intel the kind of
> heartburn AMD's been giving them in the ia32 range? Is there
> something else waiting in the wings? Or will real competition
> come only out of left field?

Who knows. I get scared thinking of all Coke and no Pepsi worlds (or all
Pepsi and no Coke).

Paralogic, Inc.           |     PEAK     |      Voice:+610.814.2800
130 Webster Street        |   PARALLEL   |        Fax:+610.814.5844
Bethlehem, PA 18015 USA   |  PERFORMANCE |

More information about the Beowulf mailing list