[Beowulf] /. Cooler room or cooler servers?
james.p.lux at jpl.nasa.gov
Thu Apr 7 20:32:14 PDT 2005
There are a few things interspersed below...
----- Original Message -----
From: "David Mathog" <mathog at mendel.bio.caltech.edu>
To: <beowulf at beowulf.org>
Sent: Thursday, April 07, 2005 5:46 PM
> > From: "Robert G. Brown" <rgb at phy.duke.edu>
> > > Thats dual 32bit xeons, the dual em64t xeon systems are more like
> > >
> > > Michael
> > Wow, that's Hot. That's actually getting up there to the old alpha
> > boxen, per CPU. So 40 1U systems are order of 16 KW/rack? Five tons of
> > AC per rack? $400/year/box for power and cooling (or thereabouts)?
> > Oooo.
> Convert 16KW to horsepower and find 16000 * .00134 => 214 hp.
> The point being that car engines generate this much heat in an
> even smaller space (1/3 to 1/4 of a rack volume, including the
> block and not just the cylinders). Automobile engineers resolve
> this issue by dumping the heat as fast as they
> possibly can. It's interesting to compare and contrast the
> way a car motor is cooled with the way a computer is cooled.
Don't forget that the heat is being rejected from a much higher temperature
in a car engine, and the price of overheating is lower. Going a few tens of
C too high won't permanently damage an engine, but will surely kill that
In semiconductors, the critical thing is the junction temperature, and for
good life, you try to keep that below 105C. There's a fairly high thermal
resistance to the case (perhaps 2C/W... much lower for a big CPU, though..
maybe 0.5 C/W). So the case (which is what, practically speaking, you can
get to for cooling) needs to be <50C or so.
In the internal combustion engine, the reaction is taking place at 1000C
plus. Even though the engine is made of cast iron (or aluminum), there's
plenty of delta T to push the heat to the coolant.
> 1. Car motors are cooled by ambient input air - I've never seen
> a car that prechills its cooling air. Nor is the cooling air
> blowing over the radiator particularly clean.
But the air blowing over the radiator is MUCH cooler than the temperature of
the radiator. Typical for a car would be coolant at 190F (or maybe even
210F) and air at, say, 100F, a delta T of 90-100F. Compare this to a
computer rejecting heat from a heatsink at maybe 50C (122F) to air at 80F a
delta T of 40F, at best... To get to a delta T of 90F, you need to precool
the air quite a bit...
> 2. There have been smallish air cooled motors but I can't think
> of any car motors over 200 hp that were air cooled. Aircraft motors
> that big may be aircooled but in that case there is typically
> in excess of 100 kph in air flow available. So expect either
> wind tunnel speeds for computer cooling air or water cooling.
yep.. this is just a "mass flow" issue. Water has the advantage of large
thermal capacity (temperature change for a given number of joules dumped
into it) compared to air.
> 3. There comes a point where there is so much heat produced it simply
> doesn't make sense to try to cool the outlet air and run it back around
> to the input. Instead the hot air is blown directly out of the
> building, with care taken to keep it away from the intakes. In which
> case either the computer needs to be able to breathe (what would
> now be considered) very dirty air or the intake system needs
> to filter a lot of particulates out very quickly. For sites
> where the weather get really hot it might still be necessary
> to cool the ambient intake air on some days, since 100F air is really
> too hot to cool a CPU or disk drive effectively.
Interestingly, I worked on some equipment at a test range in Hawaii (top of
Makaha ridge on Kauai) where very little AC was used. At first, I was
concerned that they were going to be installing it in a building with no AC.
But, they had copious air flow. The temperature never gets all that high
(neither does it ever get all that low), so ambient works in that situation.
However, in a place like that, humidity is a real concern (especially if you
have a chiller... cool things too much, and they sweat).
> Perhaps we should be looking forward to chimneys and/or cooling towers
> in future computer room designs?
More information about the Beowulf