FW: liquid cooling

Jim.Morton at aventis.com Jim.Morton at aventis.com
Mon Jun 18 15:39:04 PDT 2001



> ->---------
>From: 	Josip Loncaric[SMTP:josip at icase.edu]
>Sent: 	Monday, June 18, 2001 2:20 PM
>To: 	wickert at proteinpathways.com
>Cc: 	beowulf at beowulf.org
>Subject: 	Re: liquid cooling

>Tom's Hardware idea of a water cooled CPU:

> http://www4.tomshardware.com/cpu/01q2/010528/index.html
> 
>Interesting, although they missed the most obvious benefit of today's
>hot CPUs: waste heat could be used to keep your coffee warm.  Now, if
>only someone would start building P4 systems with built-in coffee
>warmers...



A large cluster could make a nice hot-tub heater ?


>Sincerely,
>Josip

>P.S.  Dual use applications will grow.  Combining the traditional and
>the computing meanings of the word 'architect', IBM's Blue Gene
>architects envision an indoor cascading waterfall of chilled water to
>cool their million-processor machine.  IBM'ers will be able to relax
>with the pleasant sound of gurgling water while they compute...


CRAY 2 computers  after serial number one  had a really neat looking 
"waterfall"  of  fluorinert  over the nert resivour.   The "official"
function 
of the waterfall was to degass the liquid.   At NERSC we considered 
adding some colorant to the liquid, but it is so inert that it would not 
mix with anything - you would just get bubbles of color floating around
in the coolant.   We did use a spotlight with a color-wheel for a while
though!   The CRAY 2 series was always a hit with people because they
could actually see the cooling liquid circulating around the electronics in
the machine.   The YMP  was also liquid cooled, but just looked like a
big box - it was more boring.

For troubleshooting the CRAY2 modules,  there was a workbench with 
a small  - maybe  8" X 11"  x 3" deep  basin which the bench could circulate
fluorinert through for full-power testing of a module.  For most
troubleshooting,
the techs would operate in "pulsed" mode -  with power and drive signals
applied
to the module for a few miliseconds at a time  and lots of "off"  time to
cool down.
These tests were run dry, making it simpler to move probes around in the
module
stack  ( CRAY 2 Modules were stacks of 7 PC boards with berg pins making 
connections vertically through the stack as well as connections to the
backmat)
When working with CRAY 2 modules,  a few milliseconds is a lot of
instructions :)
The test computer was a northstar-derivative CPM 80 machine with a special
I/O
card  to drive the CRAY modules.

>-- 
>Dr. Josip Loncaric, Research Fellow               mailto:josip at icase.edu
>ICASE, Mail Stop 132C           PGP key at http://www.icase.edu./~josip/
>NASA Langley Research Center             mailto:j.loncaric at larc.nasa.gov
>Hampton, VA 23681-2199, USA    Tel. +1 757 864-2192  Fax +1 757 864-6134


  Jim Morton





More information about the Beowulf mailing list