Number Crunching

James A. Landau JJJRLandau at AOL.COM
Fri Dec 17 17:50:44 UTC 2004


"Number crunching" meaning heavy-duty numerical computations is a strained
metaphor.  It might have arisen from "numeric data processing" with the image of
"crunching" for "processing".  I suspect that it is one of those "off the
wall", "straight out of left field" metaphors that become popular because they
are so far-fetched---the average user doesn't think of such an expression, and
then is so pleased with such novel imagery that s/he can't resist using it.

I have no hard evidence but I am fairly sure "number crunching" became a
popular term, and may have been coined, circa 1972.  Why?  The answer will require
some detailed computer history.

The first generation (1948 - 1958) and the second generation (1958-1964) of
computers fell into two categories: "business" (considerable data but
relatively little processing of each input) and "scientific" (relatively little input
but much processing).  The third generation (early 1960's onward) computers
were mostly designed to handle both business and scientific applications, that
is, they were designed to be "general-purpose".

Then in 1972 Seymour Cray (whose Control Data Corporation 6600 was one of the
pioneering third-generation computers) left Control Data to found his own
company, based on a new concept.  Cray knew that the most common application in
scientific computing was to solve systems of linear equations (remember the
"two equations in two unknowns" of your high-school algebra.  This is the same,
only "several thousand equations in several thousand unknowns").   So he built
a computer, the famous Cray I, which was designed to handle the problem of
linear equations.   Not only was the Cray I as fast as any of its competitors in
other computer work, it was much faster, thanks to its design, on linear
equations.

The Cray I (which was a commercial success) and a handful of competitors also
designed specifically for linear equation work were soon dubbed
"supercomputers" (a term that is not commonly used nowadays.  Instead a computer like a
Cray I is called a "vector computer".)  The work that they did was referred to as
"number crunching", although "number crunching" never referred specifically
to linear equations but could mean any problem requiring intensive
calculations.

I do not recall hearing "number crunching" before the Cray I went on the
market.

    - James A. Landau



More information about the Ads-l mailing list