VAX9000 (was: RE: kda50 manual and sdi cables & questions)
ajp166 at bellatlantic.net
Sat Apr 30 16:15:19 CDT 2005
>Subject: Re: VAX9000 (was: RE: kda50 manual and sdi cables & questions)
> From: Paul Koning <pkoning at equallogic.com>
> Date: Sat, 30 Apr 2005 16:50:15 -0400
> To: cctalk at classiccmp.org
>It's a nice powerful machine. One of its problems was that it was
>originally planned as a water-cooled machine (code name "aquarius").
>That ended up being impractical, so plan B was to do an air cooled
>somewhat slowed down version (code name "aridus" :-) ).
>So a bunch of the technology was done on the assumption that power
>density could be very high and speed would be the maximum possible at
>that time. Also, I think it was the only VAX ever shipped out of the
>"large systems group" which used to be the PDP-10 group before the
>last PDP-10 project got nuked halfway through its development.
One of the interesting aspects is the collision of technology at
that point in time. People that thought they needed a bigger
meaner VAX got it with the 9000. However in the same timeframe
there was increased demand for smaller ligher VAXen and use more
of them. What was happening in the background was the
speed/power product of CMOS was getting better and ECL (ECL100k)
was was poor in the density/power product. So while the 9000
was the fastest vax.. for a while the later CMOS versions was
very close and a smaller plus cheaper was it. Add to that that
a Multiprocessor VAX using chip technology was not only practical
and well supported it offered a better bang for the buck. If
that weren't enough the same ugly dragon that helped kill Jupiter
(PDP-10) was back to haunt VAX. At some point in the near future
it was clear that 32bits no matter how fast was going to be
inefficient and 64bits was a very viable answer, that begat Alpha.
Who'd have though in 1987 that a gigabyte was not enough and
terabytes were only a few years away.
Maybe I have a different view but, every processor technology did
reach a point where speed was not the exclusive issue. At that
point addressing (main memory addressability) would force a jump
in the word size. If that weren't true we'd be running 3ghz Z80s.
More information about the cctech