[conspire] Another motherboard was _not_ burned out today

[Rick Moen]

Quoting Scott DuBois (rhcom.linux at gmail.com):

> For giggles, I installed "lm-sensors" so I can monitor temps-
[...] 
> LOL! I know, for whatever reason I've always had this paranoia of
> overtemp even though I've _never_ had a system burn up, I don't
> overclock (anymore), and I don't do heavy gaming. I barely even do
> video multimedia work anymore.

I'm guessing you'll probably be fine, even during a hot summer --
especially since you can rev up the two remaining fans to do more.
You're doing the right thing in occasionally checking temps using
lm-sensors, to make sure.

Back in dinosaur days, I remember a really hot few days in South of
Market, San Francisco (yes, it does happen) when the bonehead error of 
working on my running 486DX2/66 OS/2 3.0 'Warp' workstation machine with
the lid open for many hours.  There I was stripped to the waist because
the air was so hot and unmoving, tinkering with what was in those days a
small minitower case stuffed with (if memory serves) two 3.5" 7200 RPM
SCSI drives and a CD-R burner.  I'd placed the drives in drive bays with
an eye to convenience for other things but not for clear airflow, and
the worst error was putting my largest-capacity SCSI drive almost snug
against the bottom 'U' of the drive-cage bracket.

Pathetic ASCII diagram of an 'elevation' front-on view of the drive 
cage (within the larger mini-tower that is itself not shown):

-------------
| --------- |
| | CD-R  | |
| --------- |
|           |
|           |
| --------- |
| | drive | |
| --------- |
|           |
| --------- |
| | drive | |
| --------- |
-------------

Leaving the case cover off for a long time, with the system running, was
the critical error, as it meant that the sinngle fan intended to draw air
through from front-bottom up through the top-rear of the case was unable
to do so.  Air _was_ being exhausted but was being drawn from
any-old-where.  Consequently, there was head _buildup_, and that is the
enemy.

Suddenly my system started behaving very peculiarly.  I shut down the
system.  The bottom HD seemed disturbingly hot.  Uh-oh.

Some painful diagnostics followed, in which I determined that the
drive's HPFS filesystems were semi-readable but showed largely gibberish
entries.  Blowing away the filesystems and recreating them, and even
low-level formatting the HD using the SCSI HBA's firmware utilities, 
didn't suffice to make that problem go away.  New filesystems also read
back with mostly gibberish contents.

My best guess is that the drive's onboard electronics, packed down near
the bottom edge of the drive and thus squeezed against the 'U' bracket,
had been baked so hard that they were permanently damaged, and the drive
was of course a write-off.

The next day, I went to Central Computers and bought a somewhat larger
and more spacious tower case to replace the small one, paying very close
attention to the airflow within the case.  

But that was before the days of super-powered gamer video boards, and 
also the real problem was that I'd simply screwed up with drive
placement and the critical error of running the system with the lid off
during super-hot weather.

This was also before lm-sensors.


_Any_ design where heat buildup doesn't occur is a good one.  You can
get there by not generating much (e.g., low-power CPU, no massive video
thing, and SSDs rather than HDs).  You can get there through _convecting_ 
away heat (airflow).  You can get there through _conducting_ away heat
(heavy metal things in contact with heat-producer, fins).  And a certain
amount of heat gets shed through _radiative_ emission, but you don't
control that.

Less heat to begin with, convection, conduction, radiation:  Those are
the only four methods for preventing heat buildup (the last three quoted
straight from Physics 101).  And heat buildup is the #1 thing that kills
and damages electronics.  (AC power fluctuation is on that list too, and
less trivial than I used to think, but it's a piker compared to heat
buildup.)