On Wed, Sep 26, 2012 at 06:35:21PM +0200, Jean Delvare wrote: > Hallo Markus, > > On Wed, 26 Sep 2012 12:41:52 +0200 (CEST), rupprecht-admin2, Markus wrote: > > I'm the admin of a school. Our Server ist a Server has a Intel DG33FB Mainboard > > (it's Intel G33) with a Intel Core2Quad processor. > > > > We are running Debian Lenny 2.6.35.9 > > > > While doing a backup with clonezilla there was the message, that the cpu > > temperature is too hot and that the cpu will work slower. > > After restarting the server I draw the four cores in kde systemmonitor (in > > german it's calles KDE Systemueberwachung). > > It shows -0,77 and -0,77 and 100 and 95 °C. I guess the decimal number is above > > 100°C. > > It was more likely an error code which wasn't caught by the software. > Given the output of "sensors" below, this seems more plausible. The > coretemp driver can't report a temperature value above the critical > limit, it is technically not possible. > > > Now I'm unsure, if this could be right. > > > > So I did sensors-detect > > 12:26/0 server ~ # sensors-detect > > # sensors-detect revision 5249 (2008-05-11 22:56:25 +0200) > > You do realize this is 4.5 year old, right? Not as old as your board, > but still, using a more recent version may help: > http://dl.lm-sensors.org/lm-sensors/files/sensors-detect > > (Seems to be down right now, you'll have to check later.) > > > Now follows a summary of the probes I have just done. > > Just press ENTER to continue: > > > > Driver `coretemp' (should be inserted): > > Detects correctly: > > * Chip `Intel Core family thermal sensor' (confidence: 9) > > (...) > > I checked: coretemp is build in. > > I googled "PC8374L" and fount that i have to use lm85. > > You would if the chip had monitoring enabled, but that's not the case. > So you can forget about the lm85 driver. This is unfortunate because > this would have given us a point of comparison. On several Intel boards > of that era, monitoring was implemented in a way Linux doesn't support. > > Does the BIOS display temperatures and/or other monitoring values? > > > 12:28/1 server ~ # sensors > > coretemp-isa-0000 > > Adapter: ISA adapter > > ERROR: Can't get value of subfeature temp1_input: Can't read > > Core 0: +0.0 C (high = +84.0 C, crit = +100.0 C) ALARM > > > > coretemp-isa-0001 > > Adapter: ISA adapter > > Core 2: +100.0 C (high = +84.0 C, crit = +100.0 C) ALARM > > > > coretemp-isa-0002 > > Adapter: ISA adapter > > ERROR: Can't get value of subfeature temp1_input: Can't read > > Core 1: +0.0 C (high = +84.0 C, crit = +100.0 C) ALARM > > > > coretemp-isa-0003 > > Adapter: ISA adapter > > Core 3: +96.0 C (high = +84.0 C, crit = +100.0 C) > > The errors for cores 0 and 1 are worrisome. We've seen these a couple > times in the past, but could never explain them nor fix them. > > > Can this be true? On the cpu is a very huge cooler with heatpipes and a large > > fan. When I touch it, it is not hot. It seams to be mounted ok. > > You did not tell us what exact CPU model your machine has. Different > models can have very different max TDP values. > > The fact that the heatsink is not hot isn't necessarily a good thing. > The heat is generated by the CPU and is then expected to dissipate to > the heatsink, where the fan will extract it, and if the case is > properly designed, the heat goes outside of the system. > > A cold heatsink can mean that the fan is doing a very good job. But it > can also mean that the dissipation from the CPU to the heatsink doesn't > happen, either because insufficient/bad thermal paste, or because the > heatsink is improperly mounted. > > The fact that you got error messages related to CPU throttling suggest > the problem is "real", i.e. not a coretemp driver issue. That being > said, the CPU throttling code is reading its values from the same > model-specific registers as the coretemp driver, so if these registers > are somehow busted in your CPU, both will misbehave. > > You may want to give a try to the latest coretemp driver: > http://khali.linux-fr.org/devel/misc/coretemp/ > I'm not holding my breath though. Another thing worth trying is a live > DVD using a more recent kernel. > > But I think that either you have a real overheating problem (check your > thermal paste and heatsink mounting) or your CPU got somehow damaged. > My guess is that it is an overheating problem. Of course, with the CPU running that hot, it might well be damaged by now as well. Guenter _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
