[gentoo-user] kernel heat warnings at low temps

[gentoo-user] kernel heat warnings at low temps

[reader]

I get warnings like these:

  Message from syslogd@reader at Mon Jun 25 16:06:35 2007 ...
  reader kernel: CPU0: Temperature above threshold

  Message from syslogd@reader at Mon Jun 25 16:06:35 2007 ...
  reader kernel: CPU0: Running in modulated clock mode


When these warnings occur the hard drive temps are low.  hddtemp shows
them at 33c (91.4 f).

Kind of hard to imagine the cpu overheating when hdd's are so low.

How can I discover the cpu temperature?  And where is the `threshold'
set? 


-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] kernel heat warnings at low temps

[Alex Schuster]

reader@newsguy.com writes:

> I get warnings like these:
>
>   Message from syslogd@reader at Mon Jun 25 16:06:35 2007 ...
>   reader kernel: CPU0: Temperature above threshold
>
>   Message from syslogd@reader at Mon Jun 25 16:06:35 2007 ...
>   reader kernel: CPU0: Running in modulated clock mode
>
>
> When these warnings occur the hard drive temps are low.  hddtemp shows
> them at 33c (91.4 f).
>
> Kind of hard to imagine the cpu overheating when hdd's are so low.

What has the hdd temp to do with cpu tmeprature?

> How can I discover the cpu temperature?  And where is the `threshold'
> set?

Reboot quickly (Alt-SysRq-S, Alt-SysRq-U, Alt-SysRq-B to do this really fast 
when the Magic SysRq feature is build into the kernel) and check the 
temperature in the BIOS.

To monitor it with Linux runing, emerge sys-apps/lm_sensors. A howto like 
http://gentoo-wiki.com/HARDWARE_Sensors might be of help when configuring.

	Alex
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] kernel heat warnings at low temps

[Hemmann, Volker Armin]

On Montag, 25. Juni 2007, reader@newsguy.com wrote:
> I get warnings like these:
>
>   Message from syslogd@reader at Mon Jun 25 16:06:35 2007 ...
>   reader kernel: CPU0: Temperature above threshold
>
>   Message from syslogd@reader at Mon Jun 25 16:06:35 2007 ...
>   reader kernel: CPU0: Running in modulated clock mode
>
>
> When these warnings occur the hard drive temps are low.  hddtemp shows
> them at 33c (91.4 f).
>
> Kind of hard to imagine the cpu overheating when hdd's are so low.

hddtemp and cputemp are completly and totally unrelated.

>
> How can I discover the cpu temperature?  And where is the `threshold'
> set?

emerge lm_sensors
run sensors_detect - this will tell you which sensors/drivers are used/needed.

I don't know about that threshold - since it seems to be independent from 
lm_sensors - I would say the hardware ;)
-- 
gentoo-user@gentoo.org mailing list

[gentoo-user] Re: kernel heat warnings at low temps

[reader]

Alex Schuster <wonko@wonkology.org> writes:

[...]

> What has the hdd temp to do with cpu tmeprature?

[...]

> hddtemp and cputemp are completly and totally unrelated.

[...]

First ... thanks for the other tips..

I think you fellows may have this a bit wrong.  I have three video
editing desktops all running win xp.  On them I use a piece of
software called `Hardware sensors monitor' or Hmonitor.

I've noticed over time (mnths) that when the cpu gets hot, the hdd are
also at elevated temps.  Maybe not critical but well above where the
run normally.

This is on three different midtower boxes, so I have surmized that
although the heating of cpu may not be related mechanically to hdd
temp, in fact they rise and fall together due probably to close
proximity and being contained in same box.

I realize this is not a definitive experiment but for my uses it does
work like that.

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Re: kernel heat warnings at low temps

[Hans-Werner Hilse]

Hi,

On Tue, 26 Jun 2007 09:27:30 -0500 reader@newsguy.com wrote:

> I've noticed over time (mnths) that when the cpu gets hot, the hdd are
> also at elevated temps.  Maybe not critical but well above where the
> run normally.

OK, you made an observation. But that doesn't make it a rule. In fact,
it may be a theory, but never a definite rule. And theories can only be
proved wrong, not true, so your observations are of statistical value
only.

Obviously, your observation is correct if the air flow in the case is
optimal. Then, all parts with proper cooling will have almost the same
temperature. However, if the air flow is broken, or some parts are not
cooled like they should -- your assumption might soon be invalid. If
you want to explicitly test that theory, just dismount your CPU fan.
This exercise will maybe cost one CPU, though :-)

So, first check if you get sensors up and running on that machine. In
addition to lmsensors, there's also the ACPI temperature zone, if
supported.

-hwh
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Re: kernel heat warnings at low temps

[James Ausmus]

On 6/26/07, reader@newsguy.com <reader@newsguy.com> wrote:
> Alex Schuster <wonko@wonkology.org> writes:
>
> [...]
>
> > What has the hdd temp to do with cpu tmeprature?
>
> [...]
>
> > hddtemp and cputemp are completly and totally unrelated.
>
> [...]
>
> First ... thanks for the other tips..
>
> I think you fellows may have this a bit wrong.  I have three video
> editing desktops all running win xp.  On them I use a piece of
> software called `Hardware sensors monitor' or Hmonitor.
>
> I've noticed over time (mnths) that when the cpu gets hot, the hdd are
> also at elevated temps.  Maybe not critical but well above where the
> run normally.
>
> This is on three different midtower boxes, so I have surmized that
> although the heating of cpu may not be related mechanically to hdd
> temp, in fact they rise and fall together due probably to close
> proximity and being contained in same box.
>
> I realize this is not a definitive experiment but for my uses it does
> work like that.

Just as a note on a possible explanation for your observations in your
Windows boxes:

Most likely (CMIIW), when you notice the CPU and HDD temps rise, you
are actively doing video editing - a CPU and memory *and* hard drive
intensive task. When hard drives are "driven hard", they heat up. When
CPU's are "driven hard", they heat up. If the computers main
functionality is a task that tends to "drive hard" both the CPU and
the hard drive, then yes, you will see a correspondence in the CPU/HDD
temperature patterns. However, this does not mean that you *cannot*
drive the CPU hard without driving the HDD hard, and vice versa - just
because the are both being driven hard when you do video editing does
not mean that they are inextricably linked to each other in workload
and temperature profile.

What operations are being performed on your Gentoo box when you see
these CPU temperature warnings?

-James


>
> --
> gentoo-user@gentoo.org mailing list
>
>
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Re: kernel heat warnings at low temps

[Kent Fredric]

On 6/27/07, James Ausmus <james.ausmus@gmail.com> wrote:
> On 6/26/07, reader@newsguy.com <reader@newsguy.com> wrote:
> > Alex Schuster <wonko@wonkology.org> writes:
> >
> > [...]
> >
> > > What has the hdd temp to do with cpu tmeprature?
> >
> > [...]
> >
> > > hddtemp and cputemp are completly and totally unrelated.
> >
> > [...]
> >
> > First ... thanks for the other tips..
> >
> > I think you fellows may have this a bit wrong.  I have three video
> > editing desktops all running win xp.  On them I use a piece of
> > software called `Hardware sensors monitor' or Hmonitor.
> >
> > I've noticed over time (mnths) that when the cpu gets hot, the hdd are
> > also at elevated temps.  Maybe not critical but well above where the
> > run normally.
> >
> > This is on three different midtower boxes, so I have surmized that
> > although the heating of cpu may not be related mechanically to hdd
> > temp, in fact they rise and fall together due probably to close
> > proximity and being contained in same box.
> >
> > I realize this is not a definitive experiment but for my uses it does
> > work like that.
>
> Just as a note on a possible explanation for your observations in your
> Windows boxes:
>
> Most likely (CMIIW), when you notice the CPU and HDD temps rise, you
> are actively doing video editing - a CPU and memory *and* hard drive
> intensive task. When hard drives are "driven hard", they heat up. When
> CPU's are "driven hard", they heat up. If the computers main
> functionality is a task that tends to "drive hard" both the CPU and
> the hard drive, then yes, you will see a correspondence in the CPU/HDD
> temperature patterns. However, this does not mean that you *cannot*
> drive the CPU hard without driving the HDD hard, and vice versa - just
> because the are both being driven hard when you do video editing does
> not mean that they are inextricably linked to each other in workload
> and temperature profile.
>
> What operations are being performed on your Gentoo box when you see
> these CPU temperature warnings?
>
> -James
>
>
> >
> > --
> > gentoo-user@gentoo.org mailing list
> >
> >
> --
> gentoo-user@gentoo.org mailing list
>
>

You can drive a CPU heavily without crunching disk IO, just windows
doen't do that so well ( espcially the case if you run out of ram and
drive into swap,.. which windows seems more predisposed to doing in my
experience ). You want an example of how to do this, then open any
high-level math software, ie: povray, video-encoding, compression,
SETI.

And you can drive hard-drives heavily without your CPU getting hot
too, I think its something related to DMA and the fact we no longer
use PIO ( well.. at least i hope not ), simply by performing
disk-to-disk transfers ( while there will still be a lot of CPU usage,
its still a bit less than you'd get without offloading ), this is
especially the case if you have a real RAID system and your doing a
RAID controlled mirror ( it has its own processor to control that )

The similarity in temperatures tho, may be related to the dynamics of
case design ( ill pretend to know what im talking about, im no
professor on this, but i have worked out how to cut degrees ).  In the
closed case, most of the time ( at least in my experience ) the
majority of hard-drive cooling is passive, relying soley on the
lone-case-fan by the CPU,  or even relying on the cooling fans in the
PSU, and generally, at least in all the tower PCS ive seen, the heat
flows out of the hard drives and over the CPU / Northbridge .

Often, this is a big sodding melting pot of heat, with your GPU just
under the northbridge, the CPU just up from the northbridge, that area
can get a bit heated, and the extra heat from the hard drives I
believe lowers the effectiveness of the CPU cooler somewhat. My
solution was not a very pretty one, but it works like a bloody charm.

I effectively made a breakout-box for my hard drives, ( well, 4 bars
of aluminium with holes in it for screwing them together ) with all
the hard drives mounted in parallel in a 'portrait' position. ( to
allow heat to flow up over the drives unconstricted ) and mounted 2
cooling fans on the sides to blow cool air over the hard drives and
back into the room  ( EM purists look away here .... ) basically
isolating the cooling systems as as not to be so codepenant. To do
this i need to have my case panels off 24/7, i admit, but my case is
so crap that keeping them on is too much effort.

In summary : CPU temp cannot be accurately measured with HDD sensor
probes, ...especially as CPU is up from hard drives in most cases, and
heat .. generally rises.
If console logs are complainging about CPU over-heat, then either its
overheating, your thresholds are too low, ...or whatevers doing the
measurement is broken.

If you cant find out from some in-linux tool what the problem is, ...
you may want to find some sort of alternative way of measuring
temperature ( ... laser thermometer might be an idea ... )

Either way, best of luck .
-- 
Kent
ruby -e '[1, 2, 4, 7, 0, 9, 5, 8, 3, 10, 11, 6, 12, 13].each{|x|
print "enNOSPicAMreil kdrtf@gma.com"[(2*x)..(2*x+1)]}'
-- 
gentoo-user@gentoo.org mailing list