Pleas ID this IBM system....

Chuck Guzis cclist at sydex.com
Wed May 22 03:24:33 CDT 2019


On 5/22/19 12:49 AM, Christian Corti via cctalk wrote:
> On Tue, 21 May 2019, Patrick Finnegan wrote:
>> Plumbing (unless you're doing aisle containment or RDHx) shouldn't run
>> through the IT space in the data center.
> 
> So how exactly do you attach a modern water cooled rack system to your
> cooling water system if not using plumbing?

So how are data centers cooled with water now?  Does the water cool
coldplates directly?

I've had only a couple of instances where cooling water was used.  In
the case of CDC mainframes, it was used to cool the condenser coils in
the refrigeration units (located in the mainframe).  I believe that Cray
initially used the same guy that CDC used to fabricate the cooling tubing.

I recall visiting the Honeywell plant in Phoenix not long after they
took it over from GE and the engineers there were tinkering with a
direct water-cooling setup--water circulated in each rack (connected by
what was probably vinyl tubing, I don't recall, only that it was
translucent), with copper diaphragms serving as the interface between
the water and the semiconductors.  I recall from comments made that
algae was a problem and adding an algicide to the cooling water tended
to corrode the copper diaphragms.

To the best of my knowledge, this was a test setup--it certainly had an
impressive instrumentation unit with multiple CRTs and neon
thermometer-type displays.

The most extreme example I ever ran into of cooling was the ETA-10
supercomputer--cooled with liquid nitrogen, supplied by a cryostat.  I
don't recall what cooled the latter.

--Chuck






More information about the cctalk mailing list