Overheating with 2.6.16 kernel

Message

vikwiz · Post by **vikwiz** » Mon Apr 03, 2006 3:14 pm

Hi everyone,

does anyone else experiences overheating reported by ACPI thermal zones, after upgrading to 2.6.16 kernel?

Here it goes up (the number at least) near to 100C (98 was the max) when loaded! After downgrading back to a 2.6.15 series kernel it goes up until 85C max, which was more-or-less normal always. Didn't re-re-reboot so much to proove this theory (it's worktime now;), so I ask the forum about other's experiences. But it's perfectly possible that the kernel reports different numbers, and not the real temp. is high. I hope at least, because 100C is *really* very high, very near to damage.

During first boot there were some acpi module loading problems, so it seems the acpi code was touched. The acpi-cpufreq module I was using could not load in new kernel, so after some experiments I'm using the speedstep-centrino and cpufreq_conservative modules now for freq. scaling., which works. But not sure if this has anything to do with the temperature problem. Freq is scaled OK now, and I'm talking about the high freq., high load situation anyway.

Some information: this is a Compaq nx8220 notebook, with 1.73GHz Pentium M proc. Kernel is sys-kernel/vserver-sources-2.1.1_rc14, which is a linux-2.6.16-vserver-2.1.1-rc14 kernel (why the vserver-sources version numbering does not reflects the kernel version anymore? but this is an other, less important question).

I don't report it to kernel devs or bugzilla before I see someone else has the same problem, because it's possible that the temperature is really high, and the heating goes wrong here. I cleaned with a vacuum-cleaner today anyway. But don't really want to go with it to the repair service either.

But it seems to depend on the kernel upgrade, will do some reboots and checks in my spare time.

Thanks for any answer!

brazzmonkey · Post by **brazzmonkey** » Mon Apr 03, 2006 3:57 pm

100°C !! i didn't know compaq made toasters !

do you have ressource-hungry process(es) ?
i think there's a kernel feature called FSB frequency scaling (or something like that, i can't recall exactly because i don't use it) - is it enabled ?

more generally speaking, 85°C is not a normal temp whatsoever, especially for a laptop (especially for a centrino). either there's something really wrong with your config, or you're talking in farenheit degrees, not celsius, or you live in a very hot country, etc...

vikwiz · Post by **vikwiz** » Mon Apr 03, 2006 4:15 pm

brazzmonkey wrote:100°C !! i didn't know compaq made toasters !

Good idea, at least to keep my caffee warm;)

brazzmonkey wrote:do you have ressource-hungry process(es) ?

Yes, this shows up in case of heavy load, one of my actual tasks is very hungry, but experienced during weekend with emerging something (to be honest was playing with catalyst2, try to gernerate a new system, no success, but this is again a completely different story;)

brazzmonkey wrote:i think there's a kernel feature called FSB frequency scaling (or something like that, i can't recall exactly because i don't use it) - is it enabled ?

This I don't know about, only proc. freq. scaling, which is a good thing, and works well. But anyway, this case I want all possible power, and it scales up to top right.

brazzmonkey wrote:more generally speaking, 85°C is not a normal temp whatsoever, especially for a laptop (especially for a centrino). either there's something really wrong with your config, or you're talking in farenheit degrees, not celsius, or you live in a very hot country, etc...

Unfortunatelly I'm talking about celsius degrees. Before I observerd 85C very rarely, the normal high load generated about 75C, and normal silent mode (not running hungry processes, freq. scales down) was 50C. That's why I found it really strange to have this high temperature! The number must be wrong, otherwise I had to *run* to repair service. Fan works OK, it speeds up in 2 or 3 steps as I can hear. And normal, unload temp is 50C, as now.

And no, the ambient temperature is normal now, 15C on street, must be around 20C in room. Today was nice and shiny anyway, if you are interested;)

I hope there's nothing wrong with the laptop, I'll do a backup immediately anyway.

brazzmonkey · Post by **brazzmonkey** » Mon Apr 03, 2006 4:38 pm

glad your weather was fine, mine was more or less like shit. nevermind.

yes i meant cpu frequency scaling, not FSB frequency scaling. my mistake.

centrino are known not to be power greedy, unlike amd mobile cpu. i suppose there's something wrong in your temp values - idle temp should probably be around 30-35°C. and i think your laptop would shutdown by itself in case you achieve 100°C for real (thermal protection stuff).
where do you get those numbers from ?

but still, that doesn't explain why you get overheating with newer kernels. i must admit i have no further idea at the moment...

vikwiz · Post by **vikwiz** » Mon Apr 03, 2006 5:01 pm

brazzmonkey wrote:centrino are known not to be power greedy, unlike amd mobile cpu. i suppose there's something wrong in your temp values - idle temp should probably be around 30-35°C. and i think your laptop would shutdown by itself in case you achieve 100°C for real (thermal protection stuff).
where do you get those numbers from ?

That's what I thought also, if the temp is really high, it would poweroff, I hope. It never happened with me before. This laptop works nice since half year without any temperature effect.

The numbers are from the gkrellm display, but shows the same value as in cat /proc/acpi/thermal_zone/TZ*/temperature, there are 3 sensors, and the first supposed to be the processor. The other numbers didn't change radically from previous values (TZ2 about 50-60C, TZ3 between 30-40C).

Do you think these numbers are cheating with me? It must be not real to have 100C!

brazzmonkey · Post by **brazzmonkey** » Mon Apr 03, 2006 7:50 pm

i sincerely think those values are wrong... i think centrino max operating temp is 95°C, so @100°C the system would have shutdown.

do you _feel_ the heat ? i remember having an old laptop at work which became really hot in summertime : i could feel the heat somewhere near the touchpad, that was really annoying.

just trying to reassure you : have you tried to reboot after a cpu intensive task and check temps in the bios ? if they are similar to those you read in linux you may carry on worrying. you may also want to check if thermal protection is on

on the other hand, if you see significant differences, stop worrying.

i've never checked temps on linux, i used to do it on windows, though (and believe me, unless you're an overclocking geek, you'll soon realise that checking temps brings you nothing but useless worries...).

swooshOnLn · Post by **swooshOnLn** » Mon Apr 03, 2006 8:03 pm

vikwiz, If you know how to get into your BIOS settings, there usually is some kind of "PC Health" screen that should show the CPU temp. Restart after a huge processor load and see if it shows the same (around 75-100C)

vikwiz · Post by **vikwiz** » Mon Apr 03, 2006 8:04 pm

Really hope that you right;)

Yes, it's hot at the places where it's usually hot, but could not tell you if its's *hotter* or not, than before when processing.

BIOS is a good idea, just didn't came to my mind. Will check it, and that must be true, isn't it? I think today's load is over, but next will come, or maybe I generate simple load, where I can safelly just turn off to get there fast.

No overclocking, of course, just it came to my eyes these strange values...

Thanks for the good idea!

swooshOnLn · Post by **swooshOnLn** » Mon Apr 03, 2006 8:24 pm

you dont have to program a space shuttle or anything, but just play some music or some videos, just untill there is a decent load for a few mins.

the BIOS readings *should* be correct. Theres always that possibility however. I would trust it, unless you know for sure its brroken (in which case, could cause false readings, etc)

vikwiz · Post by **vikwiz** » Tue Apr 04, 2006 12:25 pm

You won't imagine: can't check the temp in the BIOS! Didn't belive it, but my college said also that it's possible on notebooks.

So still unsure if this is a hardware, or kernel/config issue. Now I'm using the 2.6.15 kernel today, to see if there will be high temps. I will let you know if there is something valuable experience.

vipernicus · Post by **vipernicus** » Tue Apr 04, 2006 6:44 pm

100 Degrees Celsius? Do you live in Death Valley?

kamracik · Post by **kamracik** » Sat Apr 29, 2006 9:43 pm

i have same problem with my hp nx6110 :] and i phoned to Compaq service center and ask for that and i heard the temp of 80C is normal :] centrinos processors will shutdown automaticlly after reach 125C :]

that same i read in intel technical documentation of processor

so there is no need to be afraid :] i think..

my processor is celerom centrino 1,4GHz

dgaffuri · Post by **dgaffuri** » Sat Apr 29, 2006 10:25 pm

Don't know if it may be related, but you may want to take a look at this bug, related to overheating with 2.6.16 on Toshiba laptops. It's fixed in 2.6.16.10 and 2.6.17-r3.

Psycho Dad · Post by **Psycho Dad** » Fri May 12, 2006 3:17 pm

I also had the suspicion that temperature rises faster and goes down slower with 2.6.16 compared with 2.6.15.
So I made some tests, beginning with 56°C (the temperature when the fan stops running)
The fan starts at 70°C.
All measured with gkrellm. Max Temperature is the same (94°C).

Code: Select all

100% CPU, 1789MHz (full)

              gentoo-sources-2.6.15-r8               gentoo-sources-2.6.16-r7

70°C       25s                                              13s
80°C       66s                                              53s
84°C       101s                                            78s

100% CPU, 927MHz

              gentoo-sources-2.6.15-r8               gentoo-sources-2.6.16-r7

60°C       75s                                              28s
62°C       126s                                            48s
64°C       170s                                            73s
66°C       230s                                            98s
68°C       300s                                            134s
70°C       371s                                            158s      (fan starts)
62°C       405s                                            227s
60°C       415s                                            299s

72% CPU, 729MHz (with glxgears, same driver)

             gentoo-sources-2.6.15-r8               gentoo-sources-2.6.16-r7

58°C       31s                                              23s
60°C       97s                                              73s
62°C       219s                                            131s
64°C       374s                                            204s
66°C       572s                                            322s
68°C       851s                                            460s
70°C       1196s                                          611s      (fan starts)

So under load the CPU gets hotter more quickly.
I have the Amilo A 7600 (a notebook) with a mobile AMD 2400+
The kernel is configured in the same way (only some differences in the iptables as it is different in 2.6.16)
Any ideas?

Thomas