Hi All,

As the weather begins to get warmer here in the southern hemisphere, my AMD XP2000+ system running suse10.1 is tending to shut down for thermal reasons that I think are almost certainly spurious. I checked for installed packages and found nothing under "lm-sensors", but there was a package called "sensors" installed, which has a man page. But I'm still not clear on how this all works in suse.

The sensors command (no options) provided the following results:

~> sensors
via686a-isa-6000
Adapter: ISA adapter
CPU core: +1.78 V (min = +0.06 V, max = +3.10 V)
+2.5V: +1.10 V (min = +0.06 V, max = +3.10 V)
I/O: +3.23 V (min = +3.12 V, max = +3.45 V)
+5V: +4.95 V (min = +4.73 V, max = +5.20 V)
+12V: +12.66 V (min = +11.35 V, max = +12.48 V) ALARM
CPU Fan: 5113 RPM (min = 664 RPM, div = 8)
P/S Fan: 0 RPM (min = 664 RPM, div = 8)
SYS Temp: +68.0°C (high = +77°C, hyst = +72°C)
CPU Temp: +39.4°C (high = +146°C, hyst = -71°C)
SBr Temp: +25.3°C (high = +8°C, hyst = -49°C) ALARM

It's fairly cool today, but I'm assuming the cause of the system shutting down is when the sensor labelled as "SYS temp" above exceeds +77 C.

Apart from some pretty crazy limits being set, it looks like the SYS Temp and CPU Temp labels are reversed. BTW: The bios also has these values back to front, with sys tempreature showing as exceeding cpu temperature. Is it really correct on SUSE to set this stuff up by editing the file etc/sensors.conf? The above temperature limits do not appear anywhere in that file on my system. Should I also be running sensors -s during boot (as recommended in the sensors.conf header)?

Finally, how does the system decide what to do when there is an alarm condition?

Thanks in advance..... andrew.

actually, yes the settings should be in /etc/sensors.conf

use runlevel from YAST to activate sensors properly.

i don't think that the alarms will do anything, unless you set it up.

i'll get back with a tutorial esp. for Suse on sensors, its somewhere on my hard drive



lamb

here it is: http://suseroot.com/suse-linux-tweaks/lm_sensors.php
works on SUSE 10.2

Hi Lambchops,

I've followed the tutorial with no problems, so we'll see what happens the next hot one, but I think you are right when you say:

My guess is that doing this, and fixing the right values in /etc/sensors.conf, is not likely to fix the problem because I was getting a system message (providing a few seconds warning before imminent shutdown) before touching this alarm stuff. Turning the m/c back on immediately showed it was nowhere near the limits set by the bios.

However, it is still great to get this side of things tied down. It may help me figure out what's going on, because I'll either have to fix it in the next month, or backoff my cpu for six months of the year.

Thanks very much for that help. In deference, I shall avoid having lambchops for dinner for at least a week.

regards.... andrew.

ha....actually

the reason why i am lambchops is
my last name is "lam"

Hi lambchops,

Ah good, because I enjoy them too much.

It turns out (at least I'm 70% confident) linux goes down when the cpu temperature exceeds what is setup in the bios as the limit for the system temperature.

eg: I had bios setup with system shutdown temperature 70 C (and cpu temp 90C) the pc boots fine but once I get to desktop linux immediately gives message from root that temperature is critical and shuts down. I changed system shutdown temp setting in bios to 75C, rebooted no problem this time, and find that sensors is reporting cpu temp as 72C. The real sys temp is only 38C.

I think I should google the mobo next, and just reporting this now in case you happen to know where suse might be doing this.

best regards.... andrew.

hello lambchops.

Nice tutorial, but it didnt work for me. I think thats because I had tried to run the program and set it up first before following your guide. Could you tell me the best way to totally remove all traces of the sensors program. Im on opensuse 10.2 and just want to start afresh.

Is there a simple program that makes sure all traces are removed?

Thanks

I don't know

that's not my tutorial