[rescue] Personal progressions (was: what to do with a dec alpha 255)

[Carl R. Friend]

    On Fri, 1 Jun 2007, Geoffrey S. Mendelson wrote:

> That's the fault of DEC's sales force. During thew 1970's they got into
> the "battle of the MIPs" (million instructions per second).

    A.k.a. "Meaningless Indicator of Performance Statistic".  Done *right*
it's more of an indicator of clock-frequency and memory-bandwidth than
anything else.  Done wrong, it's "lying with statistics".  I particularly
liked (sarcasm intended) the "VUPs" (an attempt at an IBM-style comparison
between family members called the "Vax Unit of Performance").  Wags very
quickly got to calling it the "Virtually Useless Performance statistic".

> IBM based performance on a standard workload (a set of jobs they used
> to test the performance of a system) and rated the speed of succeeding
> computers in realtion to a specific sustem. In the 360 line it was the
> 360/30 (original model) in the 370 it was the 370/158 (original model).

    That's primarily useful for comparing family members to one another;
it leads to less meaningful numbers once outside the family.  Of course
trying to compare a 370 to a Cray-1 is meaningless as well because the
machines are used for *completely* different tasks.

> DEC created a CPU favoring benchmark [...]

    This is somewhat understandable based on DEC's history.  However,
DG usually beat them at their own game.  See the wag's definition
of "MIPS" above.

    Fudging benchmarks is a time-honoured game in computing and it's
certainly nothing new.  Some of the more brazen ones get immediate
attention (like hiring a consultancy to produce the numbers you want
and then playing it up as "independent research"), but the more subtle
ones usually slip by with noone noticing.

> [...] using the instructions that took the least time to execute (and
> co-incidentaly did the least), and used that to rate their computers
> in MIPS.

    Or VUPs.

> This in the end killed DEC as the 80386 and later chips out MIPSed
> anything they could produce. Again it was not a fair comparison, but
> they had trained their audience to ignore fairness.

    DEC came down for other reasons, most notably investor frustration
with DEC's inability to do consistently well on a quarter-by-quarter
basis.  There was an "investor revolt" and a hit-man (Palmer) was
brought in to sell off all the profitable pieces (for immense financial
gain to the stockholders and executives who had inside knowledge of
what was going on) and leave nothing but a shell that'd collapse and
leave the employees (who'd built up stock over the years) to go bust
when the shell collapsed.  It wasn't pretty.  It wasn't quite an early
prototype for Enron, but it was still a bloody mess.

+------------------------------------------------+---------------------+
| Carl Richard Friend (UNIX Sysadmin)            | West Boylston       |
| Minicomputer Collector / Enthusiast            | Massachusetts, USA  |
| mailto:crfriend at rcn.com                        +---------------------+
| http://users.rcn.com/crfriend/museum           | ICBM: 42:22N 71:47W |
+------------------------------------------------+---------------------+