16 Nov
2005
16 Nov
'05
3:13 p.m.
Subject: Re: removing parts from PCBs
From: Paul Koning <pkoning at equallogic.com>
Date: Wed, 16 Nov 2005 09:51:50 -0500
To: cctalk at classiccmp.org
I remember that episode. The result was "Aridus" :-), now known as
the 9000.
Also short lived as CMOS got faster..
Allison> It's still packaging. ;)
Yes. And if your circuit density isn't all that high, air cooling
works fine.
That's why Cray actually had to solve these problems, and did back in
the early 1960s (CDC 6600), because he DID build things that could not
sanely be cooled any other way.
The difference is that Cray was pushing tech harder than most and enjoyed
the unique position of faster than most by alot could justify the cost
and complexity. To do that they had to be smaller (the old 1ns/ft problem)
and that was a heat vs volume issue. Refigeration was the only way out.
They would revisit that again with the XMP. However over time the same
scene tended to play out. Those that pushed and resorted to exotic
cooling would be pushed out. Also those exoticly cooled machines
tended to not find a second market making them scarce faster.
I don't ever remember hearing of leak problems in
our Cybers...
Likely not. If there was it would be a really bad scene. I bet any
shutdown for more than minutes would have been viewed quite severely.
My view of that period was large systems ran 24/7 and all maintenence
was to be closely scheduled. I bet the CEs were also checking the loop
and maybe running a pressure hold test on occasion to assure integrity.
But the designers had to have paid close attention to things like assembly
repair, vibration and what not to insure it worked as expected planned.
Remins me of the mainframe mini joke I'd heard years ago. Minis have power
switches, mainframes have BIG RED BUTTONS (emergency power off).
Allison