[rescue] Crimson update: Hmm
Francisco Javier Mesa-Martinez
lefa at ucsc.edu
Tue Apr 1 03:39:44 CST 2003
On Tue, 1 Apr 2003, Jochen Kunz wrote:
> On 2003.04.01 01:15 Dave McGuire wrote:
> > > Liquid nitrogen.
> > It's very cold but it doesn't really carry all that much heat...
> Well. The CPUs of the ETA10 supercomputers (spin off of CDC) where
> liquid nitrogen immerse cooled.
I think ETA used a different approach than cray, CRAY used the liquid in
the supers to conduct away the heat, in order to keep the logic at room
temperature. The ETA on the other hand used CMOS (vs. ECL for the CRAYs),
but they used the nitrogen not to dissipate the heat, but actually to
chill the CPUS (the heat was actually disipated by another circuit -air or
water I think pretty elaborate heat pump design that was quite prone to
failure) to freezing temperatures. So they actually played with silicon's
properties at low temperatures to achieve <8ns clock cycles with normal cMOS
technology (mid 80s!!! Which was quite cool). As a side note, ETA was a
pretty poor choice for a company's name since it shares the same acronym
as the Spanish terrorist group ETA. Much like when the American's use the
term IRA vs. the actual IRA terrorist group.
I think there was a story of a CRAY2 whose fluorinert circuit failed
somehow and the system basically baked in a few seconds (I can only
imagine the smell of that room!!!)
> > water would be much more effective.
> Yes. Water is around 4.2 kJ/(kg*K), oil is around 2, most metals are
> below 1. But other factors are interresting too. Impedance (you can't
> get water 100% non-conductive), dielectric constant, ...
And very very corrosive! :)
More information about the rescue