Re: Hardware Progress: $397 (fwd)

From: Eugene Leitl (Eugene.Leitl@lrz.uni-muenchen.de)
Date: Wed Mar 27 2002 - 22:48:48 MST


-- Eugen* Leitl leitl
______________________________________________________________
ICBMTO: N48 04'14.8'' E11 36'41.2'' http://www.leitl.org
57F9CFD3: ED90 0433 EB74 E4A9 537F CFF5 86E7 629B 57F9 CFD3

---------- Forwarded message ----------
Date: Wed, 27 Mar 2002 18:45:44 -0500
From: Bob Drzyzgula <bob@drzyzgula.org>
To: Robert G. Brown <rgb@phy.duke.edu>
Cc: Christoph Wasshuber <wasshub@ti.com>, Beowulf@beowulf.org
Subject: Re: Hardware Progress: $397 (fwd)

A couple of years ago I stared at Intel's historical price
and performance numbers numbers until I started to hallucinate
and believe that I saw some patterns there. I did a few
calculations, some of which many of y'all might find
interesting. I put it up at

http://www.drzyzgula.org/aicpd.pdf

A couple of notes:

 * I pretty much just use an arbitrary 18-month Moore's
   Law doubling time without much comment. However, I had
   sufficient data to float this doubling rate and have the
   least squares pick it for each pricing level from $150
   to $800. When I did this, the doubling period was more
   like, IIRC, 15-17 months. Interestingly, although not
   surprisingly, the doubling period was shorter for less
   expensive processors. As a result, the model showed
   faster (in the SPECint sense) processors ultimately
   being less expensive than slower processors. To some
   extent, this makes some sense in that really expensive
   x86 processors tend to have larger caches which drive
   up the price without adding to CPU speed. This doubling
   skew was a hassle for the rest of the paper, however,
   so I fixed it a single value across all pricing levels.
   I tried a 16 month period, but it didn't change the result,
   so I went back to 18 months because the economists I
   was trying to reach were more familiar with the 18 month
   mythology.

 * I wrote this paper for internal use at my office.
   Although it's not sensitive or anything, I probably
   would have done a bit better job with references and
   such if I had planned on publishing it. A large portion
   of the data was collected by Micro Design Resources
   and/or the SPEC consortium. I supplemented this through
   other numbers I pulled off the web from things like
   the Chip Directory and Intel's press releases. This
   all should have been credited.

 * Internally, I had an appendix that contained all the
   numbers and the Octave program I used to do the
   calculations. This appendix isn't available because much
   of the data collection is copyrighted. However, if anyone
   is really interested in playing with it -- or especially
   updating it -- let me know and I'll see what I can do.

 * I'm not a statistician. This isn't my area of expertise,
   so there could be some egregious errors. I had some
   real statisticians read it, though, and they didn't
   barf or anything. Still, if anyone spots anything
   suspicious feel free to let me know.

--Bob

On Wed, Mar 27, 2002 at 05:20:54PM -0500, Robert G. Brown wrote:
>
> > I would think a cost reduction by 2-5x maximum
> > in 2006.
>
> More practically, microprocessor manufacturers and other tech component
> manufacturers don't generally focus on reducing consumer costs, they
> focus on increasing speed (and other overall performance measures) at
> constant consumer cost and relatively high margins. They can do this
> because tech profit margins depend to a significant extent on demand in
> a fairly inelastic market with very limited competition.
>
_______________________________________________
Beowulf mailing list, Beowulf@beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:37 MDT