In hot water (really hot air, but that's something else)

[Kevin O'Gorman]

I may have done something foolish.  Nothing new.  But I cannot tell
for sure, and I could use advice.

I bought parts to make a dual-Xeon machine, with DDR memory.  It was
to be my main server at home.  It's up and running, partly, and it's
about as fast as I had hoped.  But after a while, it suddenly becomes
S U P E R    S L O W.  Think PC/XT speeds.

I understand that Xeon CPUs have heat sensors, and that when they
detect overtemperature, they slow down.  That's good as far as it
goes, but I don't see why they would be overheating.

The chassis is a tall tower (with a sort of upper part and lower
part).  The upper part has the disk drives and power supply, and
is separated from the lower part by a piece of sheet metal with
a hole big enough for all the cables I'll need.  The lower part
has the motherboard with CPU's, my one add-in card (SCSI controller:
adaptec 39160) and a mess of fans.  

The power supply has an exhaust fan, and there's quite a big
auxiliary exhaust fan just above it.

In the lower bay, there's a similar big exhaust fan, both CPUs
have wind-tunnel head sinks with fans pointing towards the back,
and I've put and additional two fans pulling air in the front
and aimed at the CPUs.  When I touch the heat sinks, there's no
noticeable warmth at all, even right down by the CPU chip.

This whole afair is sitting on my table, next to my monitor.
The first two times it went into molasses-mode, all the skins were
on it.  I've got one side exposed now, and nothing bad has happened
so far, but sometimes it has taken 18 hours of running or so for
mollasses to begin.  This morning it only took about 1/2 hour.

The room is not particularly warm, though it's not air-conditioned.
I wouldn't run the A/C right now if I had it.

Any ideas?  Is there something besides heat that could make these
babies slow down?  Is there something I could test?  Can I make
them speed up without a reboot?  Is that a good idea?

Was it foolish of me to gear up for Xeons?

++ kevin


-- 
Kevin O'Gorman, PhD  (805) 650-6274  mailto:kevin at kosmanor.com
Permanent e-mail forwarder: mailto:Kevin.O'Gorman.64 at Alum.Dartmouth.org
Permanent e-mail forwarder  mailto:kogorman at umail.ucsb.edu
Web: http://kosmanor.com/~kevin/index.html