Hi Earl, Please keep the list included, so that your messages get archived. On Mon, 17 Jun 2013 12:21:14 -0600, earl middlebrook wrote: > Ubuntu did have thinkpad-acpi running. I blacklisted it as suggested by > Henrique. Now the output of sensors is: > earl@ElFuturo[earl] > sensors [11:31AM] > acpitz-virtual-0 > Adapter: Virtual device > temp1: +70.0°C (crit = +105.0°C) > > coretemp-isa-0000 > Adapter: ISA adapter > Core 0: +61.0°C (high = +95.0°C, crit = +105.0°C) > Core 2: +64.0°C (high = +95.0°C, crit = +105.0°C) > > This seems much more reasonable. Many of the temps detected by > thinkpad-acpi were 0C of fluctuating off the charts. > The cores temps are approximately the same as before. They fluctuate > faster though. Something about the update rate I suppose. The coretemp driver updates its values every second, and these integrated digital sensors are very prompt to reflect temperature changes. The values are relatively high but still reasonable: thermal margin is 31°C (not actual degrees Celsius BTW, but close enough.) I assume the above are idle temperatures though, you should worry if the temperatures go too high when the machine is under load. > > (...) > > I had not heard of this before, but searching bugzilla.kernel.org for > > similar cases I found this report: > > https://bugzilla.kernel.org/show_bug.cgi?id=27792 > > > > The good news is that you're not alone. The bad news is that this bug > > report is getting old but the bug is still not fixed, apparently it's a > > difficult one. > > It does look like it is not getting much attention any more... That's right. I've added a comment pointing to your report. > I dont understand much of what is going on but it sounds different than my > problem, at least on the surface. For instance, I have no problems with > the buttons and the temps seem strange even after a cold boot. Yes, I have noticed these differences too, but I still suspect a common root cause. > > I'd recommend blacklisting thinkpad-acpi and doing a full cold boot of the > > box (remove all batteries and AC, push power button at least ten times for > > at least three seconds each, and the last one for at least 15 seconds), to > > make sure thinkpad-acpi is not actually making matters worse. > > Done. It removes non-sense sensor readings. CPU still looks very hot. > The above sensors output was with only firefox up. > > > That said, usually these boxes work through the standard ACPI interface, and > > they _really_ depend on GPU and CPU throttling/clock/voltage control to keep > > cool. Especially if the box has a cracked thermal interface or some other > > defect. > > Is this just a matter of getting in there and having a look? First you should ensure that all power saving options are enabled. For the CPU, these are C states and cpufreq (although the usefulness of cpufreq with good C states can be discussed.) Please install powertop, then check the Idle stats and Frequency stats panels (use the Tab key to navigate between these in recent versions of powertop.) You should see multiple C states in the first and multiple frequencies in the second. When the machine is idle, the CPU should spend most of the time at the lowest C states and the lowest frequency. If this isn't the case then something is wrong, either in the BIOS or in the kernel. You may also check the power management of the graphics chip. But that depends a lot on which graphics chip you have and which driver you use for it. > > We really need a dump of the kernel boot messages to know more. > > Could you direct me to where this is located so I may pass it along? Right after boot, just do: $ dmesg > /tmp/dmesg.txt And share the file with us (send it via private e-mail or post it to some pastebin site.) -- Jean Delvare _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
