Hot CPU: system automatically shuts down

fax8 · 2008-12-26 18:40:41

Hi guys,

while running Archlinux and extensively using system (compiling, watching HD flash videos, etc), the temperature of the CPU get's pretty hot. For example I'm currently complying amarok 2.0 and this is my CPU temperature:

[root@gamma ~]# cat /proc/acpi/thermal_zone/TZ00/temperature
temperature:             103 C

The problem is that after some time at 100% CPU, it get so hot that the computer automatically blocks and shut down.. then I have to way 10-15 minutes to be able to boot again.

As I never had this problem under Windows XP I suspect this could be a linux only problem or me having some incorrect configuration somewhere.

Moreover I guess that the system should be smart enough to understand that it is getting too hot and that it should scale down the CPU power to lower the temperature. Isn't it?

I have speedstep (cpufreq) installed:

[root@gamma ~]# cpufreq-info
cpufrequtils 005: cpufreq-info (C) Dominik Brodowski 2004-2006
Report errors and bugs to cpufreq@vger.kernel.org, please.
analyzing CPU 0:
  driver: acpi-cpufreq
  CPUs which need to switch frequency at the same time: 0 1
  hardware limits: 1000 MHz - 2.00 GHz
  available frequency steps: 2.00 GHz, 1.67 GHz, 1.33 GHz, 1000 MHz
  available cpufreq governors: ondemand, performance
  current policy: frequency should be within 1000 MHz and 2.00 GHz.
                  The governor "ondemand" may decide which speed to use
                  within this range.
  current CPU frequency is 2.00 GHz (asserted by call to hardware).
analyzing CPU 1:
  driver: acpi-cpufreq
  CPUs which need to switch frequency at the same time: 0 1
  hardware limits: 1000 MHz - 2.00 GHz
  available frequency steps: 2.00 GHz, 1.67 GHz, 1.33 GHz, 1000 MHz
  available cpufreq governors: ondemand, performance
  current policy: frequency should be within 1000 MHz and 2.00 GHz.
                  The governor "ondemand" may decide which speed to use
                  within this range.
  current CPU frequency is 2.00 GHz (asserted by call to hardware).

My modules and deamons enabled:

MODULES=(e1000 slhc iwl3945 acpi-cpufreq cpufreq_ondemand snd-mixer-oss snd-pcm-oss snd-hwdep snd-page-alloc snd-pcm snd-timer snd snd-hda-intel soundcore)

DAEMONS=(syslog-ng cpufreq @network @acpid @crond @sshd @named dbus hal !fam @httpd @mysqld @postfix @samba)

Any help appreciated.

Thanks,

Fabio Varesano

jacko · 2008-12-26 18:45:48

temperature: 103 C

There is no way that can be correct. If it is, open the case and clean out the PC ffs.

peart · 2008-12-26 18:48:23

I agree with jacko. There's something really wrong there. My desktop's E7300 has never gone higher than maybe 50 degrees, and my laptop's T7300 rarely hits 60. 103 degrees is not hot, it's down right volcanic.

fax8 · 2008-12-26 18:58:55

just checked again.. that's correct or at least that's the value I get from thermal_zone while doing a compilation:

[root@gamma ~]# cat /proc/acpi/thermal_zone/TZ00/temperature
temperature:             103 C

Yes.. that's really hot. Touching the bottom of the notebook is almost impossible.

I don't think I'm going to open my notebook to clean it. It would invalidate the warranty I think.

One thing I noticed is that with such high temperatures the fan seems to still not run at 100% .. I explain..
I mean.. if after some minutes of 100% cpu work I reboot the notebook the FAN really goes faster than before while doing the BIOS boot while it slow down after linux boot..

It could mean some incorrect configuration somewhere which never let the fan goes at 100%?

Cheesebaron · 2008-12-26 19:48:40

What laptop is it?

fax8 · 2008-12-26 20:15:47

toshiba tecra a7

xaiviax · 2008-12-26 21:22:15

jacko wrote:

temperature: 103 C
There is no way that can be correct. If it is, open the case and clean out the PC ffs.

I have a toshiba and it can get to 100 C before it auto shuts down.

fax8 · 2008-12-26 21:36:10

@xaiviax: so you are also experiencing my problem? Did you find a solution? Any suggestion?

xaiviax · 2008-12-26 23:18:04

fax8 wrote:

@xaiviax: so you are also experiencing my problem? Did you find a solution? Any suggestion?

Umm... yeah... kind of...

A while back, I was using openSUSE on my notebook, and it would occasionally just turn off with no errors. After a lot of googling and such, I realized it was overheating and that this was a somewhat common issue with a lot of toshiba notebooks (and other brands). I never noticed it with XP before, and the problem seems to be exasperated with linux. I continued researching and did my own unscientific trial and error studies, and concluded that XP overheats less due mainly to the fact that high cpu intensive applications run less efficiently on XP most of the time, and thus don't heat up the CPU as much. However, XP still does a better job than linux of slowing everything down to a crawl when it needs to (critical temperature time). So I began researching how to get linux to slow down the cpu when it starts getting too hot. (and turn it back up when it cools down) And to this day, this is the one and only thing that disappoints me tremendously about linux. I have yet to find an automatic solution. The cpufreq stuff needs root privileges and I have failed at finding/making scripts or "chowning" acpi stuff, or other inelegant solutions. (and searched through every distro and DE's forums.)

So, what I currently do is: I have a temp monitor running all the time. If I'm running video editing software, or long compiling, or multiple virtual environments, or whatever, and I see the temp is going to get too hot, I just have to run as root: cpufreq-set -u 1.67ghz and then cpufreq-set -u 2ghz when the cpu intensive app is done.

Hopefully this helps, and if you or anyone else can shed more light on a better solution, I would be grateful.

fax8 · 2008-12-27 02:35:12

Ok. Seems that you have exactly the same problem I have. Have you ever noticed the fan issue I described above?

Btw.. I did some researches and I got to http://www.linux.it/~malattia/wiki/index.php/Cpufreqd which seems to be exactly what we are looking for.

I'm playing with it right now.. but I always get a "No cpufreqd socket found" ... will continue investigating.

fax8 · 2008-12-27 03:17:12

mmm... I still did not fixed the "No cpufreqd socket found" problem on cpufreqd .. so I have asked for help on cpufreqd user mailing list..
Let's see if someone is able to help me out on this.

xaiviax · 2008-12-27 03:26:22

Nice link, I'll check it out. And yes, now that I think about it, the fan never runs as fast as it does before linux boots up.

MAQ123 · 2009-01-01 02:18:32

Hello.
I really can't belive in this, someone has same toshiba laptop model as me:)
fax8- You're comaining me in this nightmare

I have exactly the same situation. Laptop is gaining extremely high temperatures. Before Arch, I used to run Debian, with kpowersave, cpufreq which was able to manage a cpu frequency, instead of running on Arch, aticonfig --set-powerstate=1 activated, in normal work (even don't think of playing games for a long time) it has about 70 C.

With Arch installed, I observed that some problems with ACPI has appeared, as I said cpufreq doesn't work which effect among others suspencion and hibernation aren't working in gnome power manager (in Debian did).

I've found a temporary solution, in BIOS, disable core multiprocessing and switch cpu frequency in 'always low', which this action cpu has 80C in normal work.

Fan is working as you said, not with it's full speed, but is it really necessary to work with normal temperature on this laptop?(I mention, one more time, in Debian it sometimes switched in max speed )

We really need some geek to deal this problem.

Greetings.

ps. Please excuse me all of my language mistakes.

Last edited by MAQ123 (2009-01-01 02:22:12)

fax8 · 2009-01-02 21:11:44

Ok.. I'm happy in not being the only one with these problems..

Unfortunately your workaround (set the CPU speed as always low) is unacceptable for me...
We bought this laptop to use its full power, not just a half, isn't it?

I think that the cpufreqd solution discussed above would be the perfect solution but I'm still unable to let it work.
Nobody on the cpufreqd mailing list answerd me.. so I'm still stuck on this..

fax8 · 2009-01-02 21:25:44

ohh.. this is new.. I just found this in the output of dmesg

ACPI: Transitioning device [FAN0] to D0
ACPI: Unable to turn cooling device [f7814b40] 'on'

It might be related. Anyone knows?

Ronin-Sage · 2009-01-02 22:48:08

100 degrees(C)...that's like the boiling point for water o.o

At that rate, you risk some serious hardware damage, I'd imagine(just assuming here).

iBertus · 2009-01-03 05:48:30

If this machine hasn't caught fire, I'm skeptical about that reading. Are we sure this isn't just a misreading from software? It's very possible with the way Intel does onboard DTS that the software you are using doesn't have the correct maximum T_junction for you CPU. This would make the temp vary by perhaps +\-15C from the actual value. Does the BIOS report the temperature?

xaiviax · 2009-01-03 10:23:38

For all the OMG, no way your temp is that high posts, if you go to intel's website and look at processor specs, you will see they list around 60C as the high end for current modern desktop processors, and 100C for the comparable mobile processors.

fax8 · 2009-01-03 11:41:18

my processor is a Intel(R) Core(TM)2 CPU T7200 @ 2.00GHz (got from cat /proc/cpuinfo) .. so I'm not sure which should be its ok running temperature.

What I do know is that once it pass 100 C (read from software) the notebook is really hot: I can't almost touch the bottom side of it.. so I'm pretty confident that it is the correct temperature.

Anyway .. I can I verify that the read temperature from the sw is the correct one? (no, no temperature indicators on BIOS).

Melson · 2009-01-04 09:41:13

xaiviax wrote:

The cpufreq stuff needs root privileges and I have failed at finding/making scripts or "chowning" acpi stuff, or other inelegant solutions.

Cpufreq changes the frequency automatically for me, but I had to configure /etc/conf.d/cpufreq first.

The ranges should be set according to available frequency steps in cpufreq-info. I think this should work for you, fax8.

#configuration for cpufreq control

# valid governors:
#  ondemand, performance, powersave,
#  conservative, userspace
governor="ondemand"

# valid suffixes: Hz, kHz (default), MHz, GHz, THz
min_freq="1000MHz"
max_freq="2GHz"

xaiviax · 2009-01-04 11:06:00

Thanks for the reply Melson, good intentions are always welcome. But that is not the issue, the governors work fine for scaling the cpu freq according to processor demand, but what we need is something that will scale down when the temp gets to a certain point, and then back up when it cools down to a lower point.

Last edited by xaiviax (2009-01-04 11:06:47)

Melson · 2009-01-04 14:56:12

Sorry, I misread your post. But what seemed strange to me was that the current frequency in the first post was the highest available since I had the lowest after the proper configuration. Maybe the temperature-driven method would not have to be used then. OK, just to mention this.

Rorschach · 2009-01-04 16:35:59

I'm having exactly the same problem on my Toshiba A100-773. Temperature always 80°C >+ . Of course this is pretty bad for the lifetime of the cpu and hdd.

The problem is that toshiba's fans are mostly not controllable by linux and the own controller are fucking bad.

My workaround tested with many distributions and always working:

1.) setting the governour to "conservative" and in no case to "ondemand" (this does the same as conservative but way more aggressive which causes way more heat production) or even "performance". In Arch install cpufreq. Add

governor="conservative"

to /etc/conf.d/cpufreq and remove everything else from this file! Especially the min/max_freq values! Remove them because the kernel checks this by himself and incorrect values there (and the default ones are mostly wrong) can make much trouble. Than add cpufreq_conservative module to rc.conf and cpufreq itself to the daemons.

2.) next thing to do is scaling down your graphics-card. on my system the ati mobility x1400 produces a lot of heat if run highest mode which is the default. You need for ati the fglrx driver to do so. Add aticonfig --set-powerstate=1 to your /etc/rc.local

That has done the job for me. Temp ~65 °C now.

MAQ123 · 2009-01-04 22:44:51

fax8 wrote:

We bought this laptop to use its full power, not just a half, isn't it?

Hell right

I cowardly returned back to Debian, but today my laptop went crazy - self shutdown - cause ? overheating

kpowersave is able to control frequency but in practise i doesn't work.

i racalled that before i had a cpu-freq daemon installed too.

look at this how-to (i used google translator to translate it from polish to english for you, despite this, the most important content is bash commands so it should't be a problem:))
http://tinyurl.com/7xnltw

with this daemon, cpu has around 80 C.

fax8 · 2009-01-26 18:31:52

Ok guys.. seems that I made some progress..

I've been able to install and use cpufreqd .. I'm still trying to get a good configuration but it definitely looks like the solution to our problems.

Anyway two days ago, while doing some video editing under windows (I have a dual boot setup) the system shut down due to overheating ...
so I started wondering if there was really something wrong with my hardware.. something related to the CPU fan..

So.. I got my tools and started to disassemble the laptop.. well.. I have to say that it has been pretty difficult to reach the cpu fan as it is just at the bottom of everything. I had to remove almost every piece in the computer.

Once I open up the fan block.. OHOHOH!!! I found almost 3 cm of dust completely blocking part of the dissipator.. The air instead of passing into the whole dissipator it used just an half.. it has to be noted that the half used was the farest from the CPU.. so it was even worst.

I cleaned the dissipator and the fan and managed to reassemble the notebook.. and I've been lucky that everything still works.

But.. the great news is that I'm running at 100% CPU for almost an hour and the temperature is still 68 C .. pretty regular.

I did some pictures of the disassembling process which I will organize in a blog post.. in case someone want to try this.
I was feared that opening the laptop might invalidate its warranty. But I did not find any warranty seal (stuff which says "warranty lost if removed").

I have to say that anyway, the way this notebook has its fan.. it's pretty obvious that it will get full of dust.. The fan suck air from the bottom and pump it to the dissipator but there are not any filters.. so the dissipator which is pretty thin acts as a filter blocking dust on it. That's not a really good hardware design IMHO.

Arch Linux

#1 2008-12-26 18:40:41

Hot CPU: system automatically shuts down

#2 2008-12-26 18:45:48

Re: Hot CPU: system automatically shuts down

#3 2008-12-26 18:48:23

Re: Hot CPU: system automatically shuts down

#4 2008-12-26 18:58:55

Re: Hot CPU: system automatically shuts down

#5 2008-12-26 19:48:40

Re: Hot CPU: system automatically shuts down

#6 2008-12-26 20:15:47

Re: Hot CPU: system automatically shuts down

#7 2008-12-26 21:22:15

Re: Hot CPU: system automatically shuts down

#8 2008-12-26 21:36:10

Re: Hot CPU: system automatically shuts down

#9 2008-12-26 23:18:04

Re: Hot CPU: system automatically shuts down

#10 2008-12-27 02:35:12

Re: Hot CPU: system automatically shuts down

#11 2008-12-27 03:17:12

Re: Hot CPU: system automatically shuts down

#12 2008-12-27 03:26:22

Re: Hot CPU: system automatically shuts down

#13 2009-01-01 02:18:32

Re: Hot CPU: system automatically shuts down

#14 2009-01-02 21:11:44

Re: Hot CPU: system automatically shuts down

#15 2009-01-02 21:25:44

Re: Hot CPU: system automatically shuts down

#16 2009-01-02 22:48:08

Re: Hot CPU: system automatically shuts down

#17 2009-01-03 05:48:30

Re: Hot CPU: system automatically shuts down

#18 2009-01-03 10:23:38

Re: Hot CPU: system automatically shuts down

#19 2009-01-03 11:41:18

Re: Hot CPU: system automatically shuts down

#20 2009-01-04 09:41:13

Re: Hot CPU: system automatically shuts down

#21 2009-01-04 11:06:00

Re: Hot CPU: system automatically shuts down

#22 2009-01-04 14:56:12

Re: Hot CPU: system automatically shuts down

#23 2009-01-04 16:35:59

Re: Hot CPU: system automatically shuts down

#24 2009-01-04 22:44:51

Re: Hot CPU: system automatically shuts down

#25 2009-01-26 18:31:52

Re: Hot CPU: system automatically shuts down

Board footer