Definition of supercomputing

Martin Fouts fouts at garth.ingr.com
Mon Aug 21 07:02:58 AEST 1989


Before the marketing folk got their hands on the word, supercomputer
was a superlative adjective that was time sensitive.  At any one time
there was only one supercomputer and it was always the *fastest*
system around.  Also, the only thing being measured in those days was
flops, since supercomputers were only for engineering calculations.

Life was so much easier when I was young.

Now, anything that is fastest in its class by some marketing criteria
is a '* supercomputer' with * replaced by such words as desktop and
departmental.

Having spent six years playing with supercomputers by the earlier
definition, my own personal definition is:

1) If costs more to run it for a month than it costs to buy the average
   system sold in the industry

or

2) If more space is required for support hardware like chillers and motor
   generator sets than for the computer system itself

its this year's supercomputers.

Some US machines which were considered supercomputers in their time,
but aren't anymore, include the IBM 7090 (Stretch,) which may have
been the first; the CDC 6600 and 7600, the Cray 1, X/MP, 2 and Y/MP (in
order of introduction,) and the CDC Cyber 203 and 205.

Amazing how many of those machines were designed by Seymour Cray.

Alternative definition:

A supercomputer is a machine designed by Seymour Cray; or a machine
designed to compete in the marketplace with a machine designed by
Seymour Cray.

Marty
--
Martin Fouts

 UUCP:  ...!pyramid!garth!fouts  ARPA:  garth!fouts at ingr.com
PHONE:  (415) 852-2310            FAX:  (415) 856-9224
 MAIL:  2400 Geng Road, Palo Alto, CA, 94303

I'll say something clever, as soon as I get some caffeine in my blood stream
-- 
Martin Fouts

 UUCP:  ...!pyramid!garth!fouts  ARPA:  garth!fouts at ingr.com
PHONE:  (415) 852-2310            FAX:  (415) 856-9224
 MAIL:  2400 Geng Road, Palo Alto, CA, 94303

I'll say something clever, as soon as I get some caffeine in my blood stream



More information about the Comp.org.usrgroup mailing list