Tracy Johnson wrote:
> Didn't the Cray II use liquid nitrogen or something like that?

I don't think so.  The "supercooled" technology you most likely refer to
is the Jacobsen junction developed by IBM R&D.  The JJ has incredible
switching speed and can therefore be clocked much faster than current
(pun?) technology.  Recall that a memory/register/whatever "bit" cannot
change from 0 to 1 instantly, there is a transition time.  Similarly,
any logic gate presented with stable input cannot instantly produce the
correct output signal.  These are "switching" limitations.

The Crays, and the preceeding CDC Cybers had a reasonably unique feature
in that they had to be "tuned" during installation to function correctly
at a given clock speed.  Due to the size and expanse of the CPU, and
it's layout of a hardwired backplane, components were interconnected by
miles of wire.  The length of the interconnecting wires was critical to
making things work right.  Too short - previous clock cycle signals
overwritten; too long - previous clock cycle signals persist.

Reminds me of the memorable day I heard Ret Adm. Grace Hopper give an
address on campus.  She started by coming to the lecturn with a piece of
wire in her hand (well, fat telephone cable) about a foot long (knowing
the readership, someone can probably supply the exact length).  She
referred to this as her "nanosecond" since it was as far as an
electrical signal could travel in one nanosecond.  She referred back to
this repeatedly in reference to advances in performance relative to
miniaturization of the components.  Her example of the time was the
basic microprocessor (8008 or 8080, I don't remember).  Put this in
perspective with current state-of-the-art chips.  Connections which used
to be wired across the room now traverse less than an inch.

Well, how's that for an off-topic rant?  Apologies to the disinterested.

Jeff Kell <[log in to unmask]>