Is /proc/acpi/thermal_zone/THM0/temperature reliable? - Hardware

This is a discussion on Is /proc/acpi/thermal_zone/THM0/temperature reliable? - Hardware ; $ uname -a Linux 2.6.9-55.EL #1 SMP Fri Apr 20 16:30:19 EDT 2007 ia64 ia64 ia64 GNU/Linux $ cat /proc/acpi/thermal_zone/THM0/temperature temperature: 27 C That looks good. But I've monitored it for a few weeks once per minute on four servers ...

+ Reply to Thread
Results 1 to 3 of 3

Thread: Is /proc/acpi/thermal_zone/THM0/temperature reliable?

  1. Is /proc/acpi/thermal_zone/THM0/temperature reliable?

    $ uname -a
    Linux 2.6.9-55.EL #1 SMP Fri Apr 20 16:30:19 EDT 2007 ia64
    ia64 ia64 GNU/Linux
    $ cat /proc/acpi/thermal_zone/THM0/temperature
    temperature: 27 C

    That looks good. But I've monitored it for a few weeks once per minute
    on four servers that have the same hardware. The temperature is always
    27C. That makes me wonder if the number is really the actual
    measurement. Thanks for any advice.

    Other info:

    $ cat /proc/cpuinfo
    processor : 0
    vendor : GenuineIntel
    arch : IA-64
    family : Itanium 2
    model : 2
    revision : 1
    archrev : 0
    features : branchlong
    cpu number : 0
    cpu regs : 4
    cpu MHz : 1600.024824
    itc MHz : 1600.024824
    BogoMIPS : 2390.75
    siblings : 1
    [one more CPU omitted here]

    # dmidecode
    # dmidecode 2.2
    # SMBIOS entry point at 0x3ee7c000
    SMBIOS 2.3 present.
    18 structures occupying 1168 bytes.
    Table at 0x3EE7C020.
    Handle 0x0000
    DMI type 0, 18 bytes.
    BIOS Information
    Vendor: HP
    Version: 04.10
    ....
    # rpm -qa | grep lm_sensors
    #

    Yong Huang

  2. Re: Is /proc/acpi/thermal_zone/THM0/temperature reliable?

    On Mon, 28 Apr 2008 08:45:20 -0700, Yong Huang wrote:

    > $ uname -a
    > Linux 2.6.9-55.EL #1 SMP Fri Apr 20 16:30:19 EDT 2007 ia64
    > ia64 ia64 GNU/Linux
    > $ cat /proc/acpi/thermal_zone/THM0/temperature temperature:
    > 27 C
    >
    > That looks good. But I've monitored it for a few weeks once per minute
    > on four servers that have the same hardware. The temperature is always
    > 27C. That makes me wonder if the number is really the actual
    > measurement. Thanks for any advice.
    >
    > Other info:
    >
    > $ cat /proc/cpuinfo
    > processor : 0
    > vendor : GenuineIntel
    > arch : IA-64
    > family : Itanium 2
    > model : 2
    > revision : 1
    > archrev : 0
    > features : branchlong
    > cpu number : 0
    > cpu regs : 4
    > cpu MHz : 1600.024824
    > itc MHz : 1600.024824
    > BogoMIPS : 2390.75
    > siblings : 1
    > [one more CPU omitted here]
    >
    > # dmidecode
    > # dmidecode 2.2
    > # SMBIOS entry point at 0x3ee7c000
    > SMBIOS 2.3 present.
    > 18 structures occupying 1168 bytes.
    > Table at 0x3EE7C020.
    > Handle 0x0000
    > DMI type 0, 18 bytes.
    > BIOS Information
    > Vendor: HP
    > Version: 04.10
    > ...
    > # rpm -qa | grep lm_sensors
    > #
    >
    > Yong Huang


    the lm_sensors has to guess at which sensors / ports / things / devices to
    monitor. if it guesses wrongly and you say... yes... that's the
    configuration to use when you run sensors-detect and setup the
    configuration.... then it will monitor the wrong data... you may get
    random counts or counts that never change.

    I'd run sensors-detect again and see if you can select a different
    device to monitor and see if that gives you better results.

    --
    D.A.M. - Mothers Against Dyslexia

    see http://www.jacksnodgrass.com for my contact info.

    jack - Grapevine/Richardson

  3. Re: Is /proc/acpi/thermal_zone/THM0/temperature reliable?

    On Mon, 28 Apr 2008, in the Usenet newsgroup comp.os.linux.hardware, in article
    <1470d781-dab6-462f-9ae8-c6e62a980f46@34g2000hsf.googlegroups.com>, Yong Huang
    wrote:

    NOTE: Posting from groups.google.com (or some web-forums) dramatically
    reduces the chance of your post being seen. Find a real news server.

    >$ uname -a
    >Linux 2.6.9-55.EL #1 SMP Fri Apr 20 16:30:19 EDT 2007 ia64
    >ia64 ia64 GNU/Linux


    cat /etc/*release /etc/*version

    >$ cat /proc/acpi/thermal_zone/THM0/temperature
    >temperature: 27 C
    >
    >That looks good. But I've monitored it for a few weeks once per minute
    >on four servers that have the same hardware. The temperature is always
    >27C. That makes me wonder if the number is really the actual
    >measurement.


    HIGHLY unlikely.

    ># rpm -qa | grep lm_sensors
    >#


    I take that to mean the lm_sensors packages are not installed.

    You'll want to read the Intel documentation on the temperature sensing
    diode in the CPU, and documentation from your motherboard and/or chipset
    supplier to read about the on-board thermal sensors (thermistors).

    BRIEFLY - neither is "accurate". The CPU temperatures are measured as
    a voltage drop across a diode passing a specified amount of current.
    How accurate is the current "set"? How accurately is the voltage drop
    measured? The answer is that nothing is checked, other than a gross "is
    it alive" type of check. Accuracy is dependent on how accurately the
    materials were fabricated. The best you can do is make relative
    measurements - it was 33C at idle yesterday when the room temperature
    was 25C/77F, and today it reads . Further, the CPU temp is
    the local temperature at a point on the die - it's not the hottest
    or coolest point. As far as the other (non-CPU) temperatures,
    the absolute accuracy is even worse, as the thing they are looking at
    (the change in resistance of a blob of metal oxides verses ambient
    temperature) depends on the mixture of those oxides, the amount of
    oxides, the composition of the mechanical coating that is used to
    package the device. The resistance at a reference temperature (usually
    25C) is subject to a 10 to 20 percent tolerance, and the rate at which
    the resistance changes is subject to a similar additional tolerance.

    The manufacturers can only afford to see that the devices are "alive".
    The cost of actually testing the accuracy is far to much for a consumer
    product - never mind actually calibrating things to some specification
    of accuracy.

    Bottom line - don't look at absolute values. Look at changes in the
    reported values.

    Old guy

+ Reply to Thread