[lm-sensors] sensors show hdd temp instead of cpu temp using

[lm-sensors] sensors show hdd temp instead of cpu temp using

[vajorie]

Hi,

I guess this could be a follow up from my previous email trying to get
lm-sensors working in this laptop. Since kernel 2.6.32, coretemp seems
to support atom cpus, as I found out, but the temperature reported as
cpu temperature (I guess?) is hdd temperature, not cpu temperature.

The bug report in kernel.org indicated that I should report it here.
Here is that bug report:
http://bugzilla.kernel.org/show_bug.cgi?id\x11161

Here's some info that I hope is useful:

Motherboard: I don't know (an Acer netbook using Atom N280 cpu)

lm-sensors v. 3.1.1

kernel 2.6.32 (arch linux)

sensors output:

coretemp-isa-0000
Adapter: ISA adapter
Core 0:      +45.0°C  (crit = +90.0°C)

coretemp-isa-0001
Adapter: ISA adapter
Core 1:      +45.0°C  (crit = +90.0°C)

acpitz-virtual-0
Adapter: Virtual device
temp1:       +26.8°C  (crit = +100.0°C)

(note: 45C is hdd temp --booted to Windows and checked with hwmonitor,
which reports cpu temp as 50C on average)

lsmod:

Module                  Size  Used by
fuse                   51628  2
aes_i586                6880  1
aes_generic            25834  1 aes_i586
ipt_REJECT              1801  2
ipt_LOG                 4383  1
xt_limit                1288  2
xt_tcpudp               1943  7
xt_state                 967  6
ipt_addrtype            1465  4
ip6table_filter          959  1
ip6_tables              9622  1 ip6table_filter
ipv6                  237500  12
nf_nat_irc               990  0
nf_conntrack_irc        2639  1 nf_nat_irc
nf_nat_ftp              1679  0
nf_nat                 12383  2 nf_nat_irc,nf_nat_ftp
nf_conntrack_ipv4       8951  8 nf_nat
nf_defrag_ipv4           795  1 nf_conntrack_ipv4
nf_conntrack_ftp        4896  1 nf_nat_ftp
nf_conntrack           46129  7
xt_state,nf_nat_irc,nf_conntrack_irc,nf_nat_ftp,nf_nat,nf_conntrack_ipv4,nf_conntrack_ftp
iptable_filter          1082  1
ip_tables               8669  1 iptable_filter
x_tables               10432  8
ipt_REJECT,ipt_LOG,xt_limit,xt_tcpudp,xt_state,ipt_addrtype,ip6_tables,ip_tables
cpufreq_conservative     5100  2
ext2                   56463  1
loop                   11399  2
snd_seq_dummy           1099  0
snd_seq_oss            25304  0
snd_seq_midi_event      4452  1 snd_seq_oss
snd_seq                42628  5 snd_seq_dummy,snd_seq_oss,snd_seq_midi_event
snd_seq_device          4313  3 snd_seq_dummy,snd_seq_oss,snd_seq
snd_hda_codec_realtek   179240  1
arc4                    1058  2
ecb                     1545  2
snd_hda_intel          18697  0
snd_hda_codec          56728  2 snd_hda_codec_realtek,snd_hda_intel
joydev                  7768  0
snd_hwdep               5102  1 snd_hda_codec
snd_pcm_oss            33693  0
ath9k                  65185  0
ath9k_common            1993  1 ath9k
snd_mixer_oss          14810  1 snd_pcm_oss
snd_pcm                57479  3 snd_hda_intel,snd_hda_codec,snd_pcm_oss
mac80211              155343  2 ath9k,ath9k_common
usb_storage            33430  0
fan                     2598  0
ac                      2255  0
snd_timer              16117  2 snd_seq,snd_pcm
ath9k_hw              206370  2 ath9k,ath9k_common
battery                 7328  0
ath                     6876  2 ath9k,ath9k_hw
snd                    43847  11
snd_seq_oss,snd_seq,snd_seq_device,snd_hda_codec_realtek,snd_hda_intel,snd_hda_codec,snd_hwdep,snd_pcm_oss,snd_mixer_oss,snd_pcm,snd_timer
soundcore               5007  1 snd
pciehp                 26539  0
snd_page_alloc          5809  2 snd_hda_intel,snd_pcm
cfg80211              111475  4 ath9k,ath9k_common,mac80211,ath
iTCO_wdt                6745  0
uhci_hcd               19092  0
rfkill                 12320  1 cfg80211
wmi                     4083  0
led_class               1933  1 ath9k
iTCO_vendor_support     1453  1 iTCO_wdt
sg                     20887  0
thermal                 9326  0
pci_hotplug            23492  1 pciehp
psmouse                55619  0
ehci_hcd               31434  0
usbcore               118825  4 usb_storage,uhci_hcd,ehci_hcd
acpi_cpufreq            5599  0
i2c_i801                7122  0
evdev                   6970  13
freq_table              1955  1 acpi_cpufreq
serio_raw               3620  0
processor              25790  3 acpi_cpufreq
coretemp                3789  0
rtc_cmos                7504  0
rtc_core               12011  1 rtc_cmos
rtc_lib                 1450  1 rtc_core
ext4                  313234  2
mbcache                 4278  2 ext2,ext4
jbd2                   68146  1 ext4
crc16                   1041  1 ext4
sd_mod                 24101  3
ahci                   29290  2
libata                135162  1 ahci
scsi_mod               78933  4 usb_storage,sg,sd_mod,libata
i915                  263409  2
drm_kms_helper         21971  1 i915
drm                   126588  3 i915,drm_kms_helper
i2c_algo_bit            4219  1 i915
button                  3638  1 i915
i2c_core               15369  4 i2c_i801,i915,drm,i2c_algo_bit
video                  14871  1 i915
output                  1404  1 video
intel_agp              23157  1
agpgart                23331  2 drm,intel_agp

What other info should I send? (Please CC me as I'm not subscribed)

Thanks.

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] sensors show hdd temp instead of cpu temp using

[Huaxu Wan]

On 19:51 Mon 04 Jan, vajorie wrote:
> Hi,
> 
> I guess this could be a follow up from my previous email trying to get
> lm-sensors working in this laptop. Since kernel 2.6.32, coretemp seems
> to support atom cpus, as I found out, but the temperature reported as
> cpu temperature (I guess?) is hdd temperature, not cpu temperature.
> 
> The bug report in kernel.org indicated that I should report it here.
> Here is that bug report:
> http://bugzilla.kernel.org/show_bug.cgi?id\x11161
> 
> Here's some info that I hope is useful:
> 
> Motherboard: I don't know (an Acer netbook using Atom N280 cpu)
> 
> lm-sensors v. 3.1.1
> 
> kernel 2.6.32 (arch linux)
> 
> sensors output:
> 
> coretemp-isa-0000
> Adapter: ISA adapter
> Core 0:      +45.0°C  (crit = +90.0°C)
> 
> coretemp-isa-0001
> Adapter: ISA adapter
> Core 1:      +45.0°C  (crit = +90.0°C)
> 
> acpitz-virtual-0
> Adapter: Virtual device
> temp1:       +26.8°C  (crit = +100.0°C)
> 

You can get the hdd temperature through hddtemp[1]. 

Coretemp get CPU temperature data from MSR directly. The data in
MSR is reported by the on-die DTS (Digital Temperature Sensor), it
reflects the delta between the current temperature and the maximum
junction temperature of the die(Tj).


[1] http://savannah.nongnu.org/projects/hddtemp/

Thanks
Huaxu

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] sensors show hdd temp instead of cpu temp using

[vajorie]

On Wed, Jan 6, 2010 at 8:59 PM, Huaxu Wan <huaxu.wan@linux.intel.com> wrote:
> On 19:51 Mon 04 Jan, vajorie wrote:
...
>> sensors output:
>>
>> coretemp-isa-0000
>> Adapter: ISA adapter
>> Core 0:      +45.0°C  (crit = +90.0°C)
>>
>> coretemp-isa-0001
>> Adapter: ISA adapter
>> Core 1:      +45.0°C  (crit = +90.0°C)
>>
>> acpitz-virtual-0
>> Adapter: Virtual device
>> temp1:       +26.8°C  (crit = +100.0°C)
>>
>
> You can get the hdd temperature through hddtemp[1].
>
> Coretemp get CPU temperature data from MSR directly. The data in
> MSR is reported by the on-die DTS (Digital Temperature Sensor), it
> reflects the delta between the current temperature and the maximum
> junction temperature of the die(Tj).
>
>
> [1] http://savannah.nongnu.org/projects/hddtemp/
>
> Thanks
> Huaxu
>


Hi,

Thanks for your response (and it's truly great to hear from intel, as
a side note). If I got your message right, unfortunately, I'm not
literate at all about the internal mechanisms of hardware monitoring
:( All I know is that the temperature that is reported as cpu's by
coretemp / sensors is actually (I'm pretty sure) that of the hdd. The
crit temps that sensors report (90C) is accurate according to intel's
specs for N280, but the temperature is too low for this cpu (right
now, it's 29C) and it corresponds almost perfectly to hdd temperature
reported in Windows (checked with a reboot from Linux to Windows). The
average cpu temp reported in Windows is 50C+...

Is it possible that what coretemp thinks is MSR is actually not it (I
don't know what I'm saying at this point)?

This may be related to how Acer set (wrongly?) things up with Aspire
One D250 (via their bios??), but I'm not sure about that. Their bios
(?) already has other problems with this cpu[1].

I opened yet another bug report in linux kernel about this:
http://bugzilla.kernel.org/show_bug.cgi?id\x14991 which has one other
user reporting similar experience on the D250. That bug report also
has my output of acpidump, if relevant here.

Thanks again for your reply.
Is there any way I can help debug this?

[1] https://bugs.launchpad.net/ubuntu/+source/cpufreqd/+bug/422858

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] sensors show hdd temp instead of cpu temp using

[Huaxu Wan]

On 00:07 Thu 07 Jan, vajorie wrote:
> On Wed, Jan 6, 2010 at 8:59 PM, Huaxu Wan <huaxu.wan@linux.intel.com> wrote:
> 
> Hi,
> 
> Thanks for your response (and it's truly great to hear from intel, as
> a side note). If I got your message right, unfortunately, I'm not
> literate at all about the internal mechanisms of hardware monitoring
> :( All I know is that the temperature that is reported as cpu's by
> coretemp / sensors is actually (I'm pretty sure) that of the hdd. The
> crit temps that sensors report (90C) is accurate according to intel's
> specs for N280, but the temperature is too low for this cpu (right
> now, it's 29C) and it corresponds almost perfectly to hdd temperature
> reported in Windows (checked with a reboot from Linux to Windows). The
> average cpu temp reported in Windows is 50C+...
> 
> Is it possible that what coretemp thinks is MSR is actually not it (I
> don't know what I'm saying at this point)?
> 
> This may be related to how Acer set (wrongly?) things up with Aspire
> One D250 (via their bios??), but I'm not sure about that. Their bios
> (?) already has other problems with this cpu[1].
> 
> I opened yet another bug report in linux kernel about this:
> http://bugzilla.kernel.org/show_bug.cgi?id\x14991 which has one other
> user reporting similar experience on the D250. That bug report also
> has my output of acpidump, if relevant here.
> 
> Thanks again for your reply.
> Is there any way I can help debug this?
> 
> [1] https://bugs.launchpad.net/ubuntu/+source/cpufreqd/+bug/422858


Following the link[1], get the package. Make and run the `./burnP6 &`
twice to occpuy whole CPU. Check whether the coretemp temperature
increase.

[1] http://pages.sbcglobal.net/redelm/cpuburn_1_4_tar.gz

Thanks
Huaxu

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] sensors show hdd temp instead of cpu temp using

[Jean Delvare]

On Thu, 7 Jan 2010 00:07:28 -0500, vajorie wrote:
> Thanks for your response (and it's truly great to hear from intel, as
> a side note). If I got your message right, unfortunately, I'm not
> literate at all about the internal mechanisms of hardware monitoring
> :( All I know is that the temperature that is reported as cpu's by
> coretemp / sensors is actually (I'm pretty sure) that of the hdd. The
> crit temps that sensors report (90C) is accurate according to intel's
> specs for N280, but the temperature is too low for this cpu (right
> now, it's 29C) and it corresponds almost perfectly to hdd temperature
> reported in Windows (checked with a reboot from Linux to Windows). The
> average cpu temp reported in Windows is 50C+...

You are doing several wrong assumptions.

You assume that the temperatures reported by hwmonitor are correct. As
this program is not provided by your hardware vendor, this may not be
the case. What makes you believe that hwmonitor is right and Linux is
wrong rather than the other way around?

You also assume that the temperature values reported under Windows and
under Linux should be the same. This may not be the case, depending on
the thermal strategy used by each OS. Many things can change the CPU
temperature, including the voltage and frequency at which it is
operated, its load, and the use of C states.

One thing I can say for sure is that the Linux coretemp driver does NOT
report your hard disk drive's temperature. This is simply impossible
technically.

> Is it possible that what coretemp thinks is MSR is actually not it (I
> don't know what I'm saying at this point)?

No, it's not possible. The coretemp driver is reporting the CPU's
temperature. Possibly not accurately, but in no way it can report
something else's temperature.

> This may be related to how Acer set (wrongly?) things up with Aspire
> One D250 (via their bios??), but I'm not sure about that. Their bios
> (?) already has other problems with this cpu[1].
> 
> I opened yet another bug report in linux kernel about this:
> http://bugzilla.kernel.org/show_bug.cgi?id\x14991 which has one other
> user reporting similar experience on the D250. That bug report also
> has my output of acpidump, if relevant here.
> 
> Thanks again for your reply.
> Is there any way I can help debug this?

As this is an Acer machine... Did you try the acerhdf driver already?

-- 
Jean Delvare

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] sensors show hdd temp instead of cpu temp using

[vajorie]

Cpuburn really helped, thank you :)

> Make and run the `./burnP6 &` twice to occpuy whole CPU. Check whether the coretemp temperature increase.

Indeed, it seems that the temperature is reported low but not as hdd's
temp. I started when hhtemp's report was same as sensors, after some
cpuburn:

$ sensors && sudo hddtemp /dev/sda
coretemp-isa-0000
Adapter: ISA adapter
Core 0:      +52.0째C  (crit = +90.0째C)

coretemp-isa-0001
Adapter: ISA adapter
Core 1:      +52.0째C  (crit = +90.0째C)

acpitz-virtual-0
Adapter: Virtual device
temp1:       +26.8째C  (crit = +100.0째C)

/dev/sda: WDC WD1600BEVT-22ZCT0: 40째C

At 52C, the fans sounded a bit panicky :)

> What makes you believe that hwmonitor is right and Linux is wrong rather than the other way around?

I thought rather that Windows software was closer to an accurate
reading. Running both speedfan & hwmonitor side by side (screenshot
http://bugzilla.kernel.org/attachment.cgi?id$478 ) demonstrates that
no one really knows what the temperature of this cpu is :-) The
particular reason why I believe Windows softwares' report in this case
is because I find the reported cpu temperatures on linux simply too
low. If the critical temperature of my cpu is 90C, why would the bios
(?) turn the fans on at 36C or so? And speeds them up at around 45C or
so? (And one more speedup at 54C I think, but I am not sure about that
at all)

As you said in your email, though, this is pure speculation on my part
(especially after seeing speedfan and hwmonitor disagree too, though
the range of temperature they disagreed sounded more logical to me).

> As this is an Acer machine... Did you try the acerhdf driver already?

Yes, the output is "No such device" (its website says: "no acer d150
and d250 support !!!" at http://www.piie.net/index.php?section촦rhdf
). I don't know why "!!!" :)

Thank you very much for all the help. Sorry to bother you all.

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] sensors show hdd temp instead of cpu temp using

[Jean Delvare]

On Thu, 7 Jan 2010 16:24:48 -0500, vajorie wrote:
> Cpuburn really helped, thank you :)
> 
> > Make and run the `./burnP6 &` twice to occpuy whole CPU. Check whether the coretemp temperature increase.
> 
> Indeed, it seems that the temperature is reported low but not as hdd's
> temp. I started when hhtemp's report was same as sensors, after some
> cpuburn:
> 
> $ sensors && sudo hddtemp /dev/sda
> coretemp-isa-0000
> Adapter: ISA adapter
> Core 0:      +52.0°C  (crit = +90.0°C)
> 
> coretemp-isa-0001
> Adapter: ISA adapter
> Core 1:      +52.0°C  (crit = +90.0°C)
> 
> acpitz-virtual-0
> Adapter: Virtual device
> temp1:       +26.8°C  (crit = +100.0°C)
> 
> /dev/sda: WDC WD1600BEVT-22ZCT0: 40°C
> 
> At 52C, the fans sounded a bit panicky :)
> 
> > What makes you believe that hwmonitor is right and Linux is wrong rather than the other way around?
> 
> I thought rather that Windows software was closer to an accurate
> reading. Running both speedfan & hwmonitor side by side (screenshot
> http://bugzilla.kernel.org/attachment.cgi?id$478 ) demonstrates that
> no one really knows what the temperature of this cpu is :-) The
> particular reason why I believe Windows softwares' report in this case
> is because I find the reported cpu temperatures on linux simply too
> low.

Yeah, since the Athlon XP and its incredibly high operating
temperature, users seem to think that 60 or even 70°C is a reasonable
CPU temperature. Thankfully CPU vendors have learned from their past
mistakes and newer CPUs, in particular the Atom, run much colder. If
you don't believe me, just compare the thermal design power of my
Athlon XP 2400+ (62W) and your Atom N280 (2.5W).

A temperature of 52.0°C when the CPU is operating at full speed, as
shown above, seems perfectly reasonable to me.

> If the critical temperature of my cpu is 90C, why would the bios
> (?) turn the fans on at 36C or so? And speeds them up at around 45C or
> so? (And one more speedup at 54C I think, but I am not sure about that
> at all)

You'd need to ask Acer about it. One possibility is that, given the
small form factor and the lack of other thermal sensor (ACPI thermal
zone doesn't seem to work), they consider the CPU temperature as an
indicator of the overall system temperature. Also, for a laptop, the
user might be in close contact with it, and they may not want the user
to get burnt.

Anyway, the only thing where the Linux coretemp driver can be wrong
(and indeed has been for some CPU models in the past) is the high
temperature limit, which itself is used to compute the current
temperature. So, while the current temperature could be wrong, the
difference between the current temperature and the high limit can't.
So, Acer really wants the fan to kick in when the thermal margin is
less than 90 - 36 = 54 degrees C.

My theory is that Speedfan, HWmonitor and the Linux coretemp driver
simply disagree on the high temperature limit for the Atom N280.
Unfortunately neither Speedfan nor HWmonitor display that value, so we
don't know what they use.

> As you said in your email, though, this is pure speculation on my part
> (especially after seeing speedfan and hwmonitor disagree too, though
> the range of temperature they disagreed sounded more logical to me).

At this point I have no reason to believe that the coretemp driver
isn't reporting the temperature correctly for your system.

-- 
Jean Delvare

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors