Date: Fri, 3 Jan 2003 20:24:51 -0700 (MST)
From: Craig Loomis
Subject: RE: 2.5m obslog 20021231 (MJD 52640)
  I bet on a poor kernel causing the system crashes, and so upgraded that
(from 2.4.2 to 2.4.18).
  As for the X server, I didn't do much besides a 2.2.7 -> 2.2.8 server
version upgrade, and send xig a plaintive email. We'll see.
  But after that, we might have found a smoking gun. Landru crashed again
during operations, and I noticed that the CPU2 temperature was reported as 87C
(vs. 35C for CPU1). Both fans are fine, and both heatsinks _feel_ like they
are connected to their CPUs in the same way (I didn't want to take everything
apart during observing). A thermocouple put somewhere near the center of the
heatsinks showed an unconvincingly small difference (CPU1 heatsink hotter @
28C vs 24C for CPU2). Still, this was measured after a cooldown and not right
above the tiny CPU.
  Hoping to get through the night, I booted with a single-CPU kernel. Dunno
how physical CPUs are mapped to logical CPUs, but it might work...
  The machine's identical predecessor died in a similar nasty way: after a CPU
fan (ball bearing) failed, nothing alarmed or shut the system/CPU down. The
CPU had to be removed from the motherboard before the machine would even get
past POST. So if it appears that the thermal bond is poor, I'll remove CPU2
for good.
  Does anyone know of a decent mboard monitor for VIA MVP3 motherboards?
 - craig