SEO

Hi NG,

I found in my var/log/messages entries like;

Apr 8 07:38:35 tux CPU0: Running in modulated clock mode
Apr 8 07:38:35 tux CPU1: Running in modulated clock mode
Apr 8 07:38:43 tux CPU1: Temperature above threshold
Apr 8 07:38:43 tux CPU0: Temperature above threshold

which was fixed by acuuming the dusty Zalman 7000 copper cooler.

But my temperature monitor using the w83627 module and GKrellm, reports
temperatures never above 66C........hot, but not dangerously?.

At what temp does the kernel start putting these warnings in my log?. Can I
disable them?. They don't seem important, as the CPU is a P4 2,8ghz 800FSB,
which should be able to run fine a 66C, and if not, clock itself down.

/Stig
--
* Registered Linux user #291266, with http://counter.li.org.*

Stig Mogensen wrote:

> I found in my var/log/messages entries like;
>
> Apr 8 07:38:35 tux CPU0: Running in modulated clock mode
> Apr 8 07:38:35 tux CPU1: Running in modulated clock mode
> Apr 8 07:38:43 tux CPU1: Temperature above threshold
> Apr 8 07:38:43 tux CPU0: Temperature above threshold
>
> which was fixed by acuuming the dusty Zalman 7000 copper cooler.

That can easily take a couple of degrees off. You may consider
refreshing the thermal compound as well. My CPU temperature dropped
10-12 degrees when I switched to AS5.

> But my temperature monitor using the w83627 module and GKrellm,
> reports temperatures never above 66C........hot, but not dangerously?.

Depends on the chip and the kind of load (some tasks will heat up
specific parts of the CPU more). Some start making errors as low as 75
degrees. Is the 66C at constant high load? How about if you run a CPU
specific cpuburn?

> At what temp does the kernel start putting these warnings in my log?.
> Can I disable them?. They don't seem important, as the CPU is a P4
> 2,8ghz 800FSB, which should be able to run fine a 66C, and if not,
> clock itself down.

And indeed it does. It does it at the temperature that the sensor
driver is set to, which should be configurable in /etc/sensors.conf. At
least that's how frequency scaling works on my system. The action is a
MB BIOS setting, but the temperature level is set with the driver, which
again gets its settings from the sensors.conf file.

--
PeKaJe

Linux has a history of standing on the shoulders of giants, while Microsoft has
a history of trying to break giants' legs. The latter is quite less efficient.

Stig Mogensen wrote:
> Hi NG,
>
> I found in my var/log/messages entries like;
>
> Apr 8 07:38:35 tux CPU0: Running in modulated clock mode
> Apr 8 07:38:35 tux CPU1: Running in modulated clock mode
> Apr 8 07:38:43 tux CPU1: Temperature above threshold
> Apr 8 07:38:43 tux CPU0: Temperature above threshold
>
> which was fixed by acuuming the dusty Zalman 7000 copper cooler.
>
> But my temperature monitor using the w83627 module and GKrellm, reports
> temperatures never above 66C........hot, but not dangerously?.

lm_sensors don't necessarily have to be tuned fro your mainboard, you need to
tune the configuration file so that you are sure that the values are correct.


> At what temp does the kernel start putting these warnings in my log?. Can I
> disable them?. They don't seem important, as the CPU is a P4 2,8ghz 800FSB,
> which should be able to run fine a 66C, and if not, clock itself down.

Intel CPUs are more heat sensitive than AMDs (just comparing x86), most of
AMDs are able to operate at 80 degrees, while Intel's usually should be hold
below 65 degrees. Even if your lm_sensors does report 66 degrees it could be
75 or even more, I would suggest you do reset your machine, enter the BIOS and
check the values in the hardware monitor page (keep in mind that the
temperature will have dropped).


//Aho

>> which was fixed by acuuming the dusty Zalman 7000 copper cooler.
>
> That can easily take a couple of degrees off. You may consider
> refreshing the thermal compound as well. My CPU temperature dropped
> 10-12 degrees when I switched to AS5.

Yep. Just ordered a tube from my dealer ;o)

> Depends on the chip and the kind of load (some tasks will heat up
> specific parts of the CPU more). Some start making errors as low as 75
> degrees. Is the 66C at constant high load? How about if you run a CPU
> specific cpuburn?

The 66C is reached after days of constant SETI@home runs.

> And indeed it does. It does it at the temperature that the sensor
> driver is set to, which should be configurable in /etc/sensors.conf. At
> least that's how frequency scaling works on my system. The action is a
> MB BIOS setting, but the temperature level is set with the driver, which
> again gets its settings from the sensors.conf file.

Arh....back to fiddling with .conf files ;o)

My MB is an Asus P4P800. If a sensors.conf file fits one P4P800, will it fit
all?.


/Stig

--
* Registered Linux user #291266, with http://counter.li.org.*

Stig Mogensen wrote:

>> That can easily take a couple of degrees off. You may consider
>> refreshing the thermal compound as well. My CPU temperature dropped
>> 10-12 degrees when I switched to AS5.
>
> Yep. Just ordered a tube from my dealer ;o)

Good. You ordered some xylene-based cleaning fluid too, right? Regular
alcohol may also work, but it can sometimes be hard to get all the gunk
off.

>> Depends on the chip and the kind of load (some tasks will heat up
>> specific parts of the CPU more). Some start making errors as low as
>> 75 degrees. Is the 66C at constant high load? How about if you run
>> a CPU specific cpuburn?
>
> The 66C is reached after days of constant SETI@home runs.

OK, so you should probably expect the maximum temperature it can reach
(when all kinds of instructions are being used) to be a couple of
degrees higher. SETI@home tends not to stress the CPU completely,
unless you fetch an optimized build.

>> And indeed it does. It does it at the temperature that the sensor
>> driver is set to, which should be configurable in /etc/sensors.conf.
>> At least that's how frequency scaling works on my system. The action
>> is a MB BIOS setting, but the temperature level is set with the
>> driver, which again gets its settings from the sensors.conf file.
>
> Arh....back to fiddling with .conf files ;o)

It's not so bad. It's a two line change ... Of course you need to
*find* them first! Don't worry, it shouldn't be very difficult. Try
examining the output of 'sensors'. It should tell you which chip
configuration is being used. Then you just need to search the conf file
for it and change the max and hysteresis temperatures that follow
somewhere.

> My MB is an Asus P4P800. If a sensors.conf file fits one P4P800, will
> it fit all?.

Not sure, it all depends on the sensor chip used. While it's unlikely
that they've changed it from one batch of P4P800 MBs to another, it's
not entirely impossible. Sometimes one still needs to fiddle with the
temperature calculation formulas, because otherwise identical MBs differ
in some minor component (a different resistor is not uncommon, or so
I've heard).

--
PeKaJe

IBM: It may be slow, but it's hard to use.