RE: DS20L

["Christopher C. Chimelis" <chris debian org>]

On Fri, 31 May 2002, Iwao Makino wrote:

> I don't know what environmental conditions in the US, but over here, it's 
> almost
> impossible to get room about 18C(65F).  Best we can expect is 20C,but most
> of places for them machines including telecom centres, it's about 25C.
> Inside cabinet is little warmer. So I can agree heat problem...

Well, even that is relative.  I've been in many machine rooms in my time
that were 25C, but wouldn't have had a problem with even IA-64
equipment.  The key is usually airflow within the room and especially the
cabinets involved.  Too many of the racks at my current place of
employment still have doors on the front that obstruct ~50% of the airflow
going into the front of the machines contained therein.  Front-of-the-rack
airflow seems to not be considered much in some cabinets...I guess since
they only concern themselves with airflow to the rear of the units where
most power supplies are (that, and aesthetics seem to play the biggest
part of rack design...why, I have no idea).

> I've seen 128CPU IA nodes running at 28C or 30C room temp for just fine...

Apples and oranges comparison here.  IA isn't even a consistant comparison
within it's own architecture class, as evidenced by Athlon<->Intel average
core temperature differences.  Comparing CS20 to another Alpha-based
system, or maybe even a larger HPPA installation, would be better (my HP
A500 is a practically a space heater).

C