Karl Cunningham wrote:
On 4/13/2007 12:53 PM, Lan Barnes wrote:
On Fri, April 13, 2007 10:00 am, Karl Cunningham wrote:
On 4/13/2007 9:46 AM, Lan Barnes wrote:
:'-(
It'll run idle overnight with no problems, but when I load the CPU (top
says 30%), it goes from 38 C to 68 C in about 4 minutes.
I'm taking it back and making the case that the original diagnosis
charge
should still be in force.
When you put the load on the CPU, how much does the temperature go up in
the first 10-20 seconds? If most of the 30-degree rise is in the first
20 seconds or less, it indicates a CPU-to-heatsink interface problem. If
it's a gradual increase over minutes, then it's a problem cooling the
heatsink -- probably airflow.
Karl
Started at 22 C and ran for three minutes in CLI.
Took it into X and ran top -- 1.7% CPU and 22 C for another couple of
minutes.
Started a little looping Tcl script (poorly optimized for niceness) that
took it to 33% CPU -- went to 68 C in 46 sec. Stayed there the rest of
the
time.
What is the likelihood that the BIOS as reported through
cat /proc/acpi/thermal_zone/THRM/temperature
is lying about the temp?
If it got to 68C in 46 sec, did the fan come on then and stop it from
rising more? Unless the temp went up most of the way in <10 seconds it's
probably not a cpu-to-heatsink problem.
Temp in /proc could be lying. I've only used sensors. Many times I've
had to calibrate it though, in sensors.conf, by comparing what sensors
says to what the bios says under similar conditions: Boot to runlevel 3,
check sensors, wait for stable temp, reboot and check bios temp, adjust
sensors.conf and repeat.
I do cpu load tests with:
$ perl -e 'while(1){}'
Takes 100% of one CPU.
Check out cpuburn. There is a reference on freshmeat.net. Be careful
using it as you can fry your system.
Gus
--
[EMAIL PROTECTED]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
