Running Computers Cold
ethan.dicks at gmail.com
Wed Dec 9 00:58:22 CST 2009
On 12/9/09, Zane H. Healy <healyzh at aracnet.com> wrote:
> At 9:55 PM -0800 12/8/09, Chuck Guzis wrote:
>>On 8 Dec 2009 at 21:19, Zane H. Healy wrote:
>>> Okay, this is the first time I've ever had to worry about this. When
>>> is it to cold to run a computer? It's 35F out in the garage, and it
>>> is supposed to get a lot colder tonight. I just shut the dehumidifier
>>> down (to cold to run it) and setup a heater near the computers (and
>>> other stuff I don't want to freeze).
>>It seems to me that Ethan would be the perfect one to answer this.
>>35F is probably a heat wave at the South Pole.
> Except I don't think they run their computers outdoors! :-)
But we do use outside air to cool them. It's free except for the
power to push it around. (Oh... and +35F is never seen at the Pole -
the record is +7.5F, and I've personally been around for +7.0F).
As for the extreme case, we've had computers malfunction when outside
access doors were left open and -80F air came in directly, bypassing
the blowers and the louvers. On a day-to-day basis, the room with the
14 racks that was AMANDA (it was shut down earlier this year after a
10+ year run) shed about 35-40kW of heat with indirect access to
outside air with some measure of automatic and manual thermal controls
(covering up open cable panels and stuffing blankets in hatches in
addition to thermostatic controls on air blowers). If we let the room
get over about +55F, the high-voltage supplies for the photomultipler
tubes would go into thermal shutdown (ultra-dry air at 650millibars
doesn't have much heat capacity). OTOH, and more to the point, if we
let the room get much colder than about +35F (say +25F or colder), a
specific rack of digital hardware that was adjacent to the floor vents
feeding cold air to the high voltage supplies would malfunction until
the temp came up to the high thirties to low forties.
In another location entirely, central Ohio, I used to rent the
basement of my mother's typing and typesetting shop. The building was
a late 19th C/early 20th C brick "shotgun" commercial space with a
former storm-cellar-type access to the basement. As such, cold air
poured from the modern back door, down the basement stairs, and into
the space I ran PDP-8s, PDP-11s and a VAX-11/730. One of my jobs at
the time was hacking PDP-11 assembler on an 11/23. The basement would
routinely get to +40F, and sometimes colder if the wind was from the
right direction (the water pipes had electric wraps). I couldn't
personally stand to work in that environment without a heater pointed
at me, but the computers ran fine. The lone device that had problems
was an LA-180 printer I used for listings. It worked down to about
+45F, but colder than that, I speculate that the rail lubricant got
too viscous, because it would blow carriage motor fuses until it
warmed up. I quickly learned not to print on cold nights.
I'd say that if you keep things at or above freezing, you are probably
perfectly fine. Magnetic media is a lot more sensitive than ICs in
terms of cold soaking. One thing to watch for is to not power up
cold-soaked electronics. The current inrush is likely to blow ICs
(the internal bonded wires between the die and the frame, mostly).
I've thawed machines that were left in unheated buildings over the
winter at McMurdo - ordinary temps around -45F or so. Specifically in
that case (ultra cold, powered off), there are known and published
"max rates of rise" of temps to minimize the risk of permanent damage
from thermal expansion. A good rule of thumb is about 2-3 degrees per
hour. What I did with the cold-soaked computers was to throw them
into a lab freezer at -40F for a few hours, then into a lab
environmental chamber at -30F that I would tweak up about 5 degrees
every couple of hours. When the chamber was up to about +20F, I threw
the equipment in a lab refrigerator. The thaw process took two
workdays, but 100% of what I treated that way survived (no hard disk -
these were floppy-booting diskless PCs that ran from a Novell server).
If it gets really cold (+0F, say), I'd bring the disks in the house
and leave the CPUs powered off until the garage temps are back around
+32F. ICs can be stored down to -40 typically, but not operated at
those temps (and especially not put through a power-on cycle at those
So that's my experience and observations of cold and computers. Take
away from it what you will.
More information about the cctech