Sun Blade 1000/2000 power management - SUN

This is a discussion on Sun Blade 1000/2000 power management - SUN ; I have upgraded recently my Blade 1000 with two 1200 MHz US III Cu CPUs, from 2 x 900 MHz (not Cu). Everything works fine, except that I have observed that the CPUs never enter low power state when the ...

+ Reply to Thread
Results 1 to 17 of 17

Thread: Sun Blade 1000/2000 power management

  1. Sun Blade 1000/2000 power management

    I have upgraded recently my Blade 1000 with two 1200 MHz US III Cu
    CPUs, from 2 x 900 MHz (not Cu). Everything works fine, except that I
    have observed that the CPUs never enter low power state when the
    system is in low power mode (blinking power button green LED).

    This is how I test the system:
    power.conf :
    ---
    autopm enable
    cpupm enable
    system-threshold 5m
    cpu-threshold 2m
    loadaverage 0.20
    ----
    powerd running.
    I alternate 1 x 900MHz with 1 x 1200 MHz Cu (after power off, of
    course).
    I boot the system without keyboard/mouse attached and I do not log in.

    With the 900 MHz I observe the following (using a Watt meter):
    Immediately after system startup: 200 Watts
    ~3 minutes after startup: 160 Watts
    ~5 minutes after startup: 115 Watts
    No heat coming from the PSU and the CPU is nearly cold. Case fans
    running at low speed, PSU fans stopped.

    With the 1200MHz Cu installed:
    Immediately after system startup: 185 Watts
    3 minutes after startup with system idle and no users logged in: 185
    Watts
    5 minutes after startup still idle: 185 Watts
    PSU overheating, CPU as hot as it were with the system in full power.
    Case fans running at low speed, PSU fans stopped and started at full
    speed after a few minutes.

    What is bad, in the second case, is that the system enters in a low
    power state, I can see the power supply fans being turned off and the
    green light on the power button flashing, however the CPU is still
    draining full power. After a while the power supply overheats and
    starts the fans at max speed for a few seconds, and so on.

    I was wondering if anyone with a Blade 2000 and Cu CPUs could report
    if they observe the same behaviour, or if they see that their system
    actually enters in a low power state reducing the power drain.


  2. Re: Sun Blade 1000/2000 power management

    haydude wrote:
    > I have upgraded recently my Blade 1000 with two 1200 MHz US III Cu
    > CPUs, from 2 x 900 MHz (not Cu). Everything works fine, except that I
    > have observed that the CPUs never enter low power state when the
    > system is in low power mode (blinking power button green LED).
    >
    > This is how I test the system:
    > power.conf :
    > ---
    > autopm enable
    > cpupm enable
    > system-threshold 5m
    > cpu-threshold 2m
    > loadaverage 0.20
    > ----
    > powerd running.
    > I alternate 1 x 900MHz with 1 x 1200 MHz Cu (after power off, of
    > course).
    > I boot the system without keyboard/mouse attached and I do not log in.
    >
    > With the 900 MHz I observe the following (using a Watt meter):
    > Immediately after system startup: 200 Watts
    > ~3 minutes after startup: 160 Watts
    > ~5 minutes after startup: 115 Watts
    > No heat coming from the PSU and the CPU is nearly cold. Case fans
    > running at low speed, PSU fans stopped.
    >
    > With the 1200MHz Cu installed:
    > Immediately after system startup: 185 Watts
    > 3 minutes after startup with system idle and no users logged in: 185
    > Watts
    > 5 minutes after startup still idle: 185 Watts
    > PSU overheating, CPU as hot as it were with the system in full power.
    > Case fans running at low speed, PSU fans stopped and started at full
    > speed after a few minutes.
    >
    > What is bad, in the second case, is that the system enters in a low
    > power state, I can see the power supply fans being turned off and the
    > green light on the power button flashing, however the CPU is still
    > draining full power. After a while the power supply overheats and
    > starts the fans at max speed for a few seconds, and so on.
    >
    > I was wondering if anyone with a Blade 2000 and Cu CPUs could report
    > if they observe the same behaviour, or if they see that their system
    > actually enters in a low power state reducing the power drain.


    I've a blade 2000 (not 1000) with 2 x 1.2 GHz. I've never seen the power
    light flash in the way you describe.

    I do notice when it first switched on the fans run at full speed. They
    then reduce speed after a while. If I switch off (either with 'poweroff'
    or init 5), the fans speed up before the systems powers off. (I assume
    the operating system is supposed to send some sort of 'signal' to a
    power management system. As soon as the OS shuts down, that stops, so
    the fans are set to run at full speed. This is my guess - I do not know
    for sure).

    I don't have anything to measure the power. I do have a 6.5 digit
    lab-grade multi-meter, so could measure the current down to the nearest
    micro amp or so, but unless I hook a scope up, and start measuring the
    phase angle between voltage and current, I can't calculate the power.

    It runs a lot cooler than the Ultra 80, so I guess it is costing me less
    to run!

  3. Re: Sun Blade 1000/2000 power management

    haydude wrote:
    > I have upgraded recently my Blade 1000 with two 1200 MHz US III Cu
    > CPUs, from 2 x 900 MHz (not Cu). Everything works fine, except that I
    > have observed that the CPUs never enter low power state when the
    > system is in low power mode (blinking power button green LED).
    >
    > This is how I test the system:
    > power.conf :
    > ---
    > autopm enable
    > cpupm enable
    > system-threshold 5m
    > cpu-threshold 2m
    > loadaverage 0.20
    > ----
    > powerd running.
    > I alternate 1 x 900MHz with 1 x 1200 MHz Cu (after power off, of
    > course).
    > I boot the system without keyboard/mouse attached and I do not log in.
    >
    > With the 900 MHz I observe the following (using a Watt meter):
    > Immediately after system startup: 200 Watts
    > ~3 minutes after startup: 160 Watts
    > ~5 minutes after startup: 115 Watts
    > No heat coming from the PSU and the CPU is nearly cold. Case fans
    > running at low speed, PSU fans stopped.
    >
    > With the 1200MHz Cu installed:
    > Immediately after system startup: 185 Watts
    > 3 minutes after startup with system idle and no users logged in: 185
    > Watts
    > 5 minutes after startup still idle: 185 Watts
    > PSU overheating, CPU as hot as it were with the system in full power.
    > Case fans running at low speed, PSU fans stopped and started at full
    > speed after a few minutes.
    >
    > What is bad, in the second case, is that the system enters in a low
    > power state, I can see the power supply fans being turned off and the
    > green light on the power button flashing, however the CPU is still
    > draining full power. After a while the power supply overheats and
    > starts the fans at max speed for a few seconds, and so on.
    >
    > I was wondering if anyone with a Blade 2000 and Cu CPUs could report
    > if they observe the same behaviour, or if they see that their system
    > actually enters in a low power state reducing the power drain.


    I've a blade 2000 (not 1000) with 2 x 1.2 GHz. I've never seen the power
    light flash in the way you describe.

    I do notice when it first switched on the fans run at full speed. They
    then reduce speed after a while. If I switch off (either with 'poweroff'
    or init 5), the fans speed up before the systems powers off. (I assume
    the operating system is supposed to send some sort of 'signal' to a
    power management system. As soon as the OS shuts down, that stops, so
    the fans are set to run at full speed. This is my guess - I do not know
    for sure).

    I don't have anything to measure the power. I do have a 6.5 digit
    lab-grade multi-meter, so could measure the current down to the nearest
    micro amp or so, but unless I hook a scope up, and start measuring the
    phase angle between voltage and current, I can't calculate the power.

    It runs a lot cooler than the Ultra 80, so I guess it is costing me less
    to run!

  4. Re: Sun Blade 1000/2000 power management

    Dave,

    Thank you for your reply.

    On 7 Nov, 06:42, Dave wrote:
    > I've ablade2000 (not 1000) with 2 x 1.2 GHz. I've never seen thepower
    > light flash in the way you describe.


    That means that your system never enters in power save mode, which is
    perhaps either because your power saving is disabled or it is set with
    a minimum threshold that is never reached ("loadaverage" default is
    very low).

    Please, could you try to set your /etc/power.conf as shown in my
    previous post, make sure that "powerd" service is enabled, then reboot
    and unplug the keyboard/mouse (to make sure that not even dtlogin
    starts), and leave it there idle for a few minutes without logging in.

    Without a power meter, here is an empiric way to assess what your
    system is doing:
    A few minutes after boot you should see the power button flashing. At
    that point the PSU fans are stopped, because the system has entered in
    low power mode.
    If the CPUs enter in low power mode too you should be able to feel
    that the rear air vent where the CPUs are located goes from warm to
    nearly cold, the PSU (which is the unit with the tall handle) also
    should stay cold and you should not feel any air coming out because
    the fans are stopped (flashing a torch through the rear PSU's vents
    you can see the fans located towards the front).

    In this state the system is still up and running and responsive, which
    means that if you telnet/ssh/log in from the serial console the system
    responds. To save power, the system clock is (supposed to be) cut to
    1/32 of the nominal frequency of 150MHz (at least that is what states
    C.6.1.1 of Sun Blade 1000 and Blade 2000 Service Manual and what
    happens in my case with 900MHz CPUs installed). As soon as you start
    using the CPU, the clock is restored to the nominal speed and the
    power switch LED returns to solid green.

    If instead the PSU gets hot and, the CPU vents remain warm and, the
    PSU fans blast on after a while (because the PSU is overhating due to
    the CPUs still dragging full power), you are in the same condition
    that I have observed in my Blade 1000.


    > I do notice when it first switched on the fans run at full speed. They
    > then reduce speed after a while. If I switch off (either with 'poweroff'
    > or init 5), the fans speed up before the systems powers off. (I assume
    > the operating system is supposed to send some sort of 'signal' to apowermanagementsystem. As soon as the OS shuts down, that stops, so
    > the fans are set to run at full speed. This is my guess - I do not know
    > for sure).


    Tha is normal operation. The case fans are controlled by Solaris, but
    not by the OPB. They are blasted at full speed by the OPB, because the
    OPB does not monitor the CPU temperature.

    HD


  5. Re: Sun Blade 1000/2000 power management

    Dave,

    Thank you for your reply.

    On 7 Nov, 06:42, Dave wrote:
    > I've ablade2000 (not 1000) with 2 x 1.2 GHz. I've never seen thepower
    > light flash in the way you describe.


    That means that your system never enters in power save mode, which is
    perhaps either because your power saving is disabled or it is set with
    a minimum threshold that is never reached ("loadaverage" default is
    very low).

    Please, could you try to set your /etc/power.conf as shown in my
    previous post, make sure that "powerd" service is enabled, then reboot
    and unplug the keyboard/mouse (to make sure that not even dtlogin
    starts), and leave it there idle for a few minutes without logging in.

    Without a power meter, here is an empiric way to assess what your
    system is doing:
    A few minutes after boot you should see the power button flashing. At
    that point the PSU fans are stopped, because the system has entered in
    low power mode.
    If the CPUs enter in low power mode too you should be able to feel
    that the rear air vent where the CPUs are located goes from warm to
    nearly cold, the PSU (which is the unit with the tall handle) also
    should stay cold and you should not feel any air coming out because
    the fans are stopped (flashing a torch through the rear PSU's vents
    you can see the fans located towards the front).

    In this state the system is still up and running and responsive, which
    means that if you telnet/ssh/log in from the serial console the system
    responds. To save power, the system clock is (supposed to be) cut to
    1/32 of the nominal frequency of 150MHz (at least that is what states
    C.6.1.1 of Sun Blade 1000 and Blade 2000 Service Manual and what
    happens in my case with 900MHz CPUs installed). As soon as you start
    using the CPU, the clock is restored to the nominal speed and the
    power switch LED returns to solid green.

    If instead the PSU gets hot and, the CPU vents remain warm and, the
    PSU fans blast on after a while (because the PSU is overhating due to
    the CPUs still dragging full power), you are in the same condition
    that I have observed in my Blade 1000.


    > I do notice when it first switched on the fans run at full speed. They
    > then reduce speed after a while. If I switch off (either with 'poweroff'
    > or init 5), the fans speed up before the systems powers off. (I assume
    > the operating system is supposed to send some sort of 'signal' to apowermanagementsystem. As soon as the OS shuts down, that stops, so
    > the fans are set to run at full speed. This is my guess - I do not know
    > for sure).


    Tha is normal operation. The case fans are controlled by Solaris, but
    not by the OPB. They are blasted at full speed by the OPB, because the
    OPB does not monitor the CPU temperature.

    HD


  6. Re: Sun Blade 1000/2000 power management

    Dave writes:

    >I do notice when it first switched on the fans run at full speed. They
    >then reduce speed after a while. If I switch off (either with 'poweroff'
    >or init 5), the fans speed up before the systems powers off. (I assume
    >the operating system is supposed to send some sort of 'signal' to a
    >power management system. As soon as the OS shuts down, that stops, so
    >the fans are set to run at full speed. This is my guess - I do not know
    >for sure).


    It's all related to picld; as soon as picld stops controlling the fans
    the firmware will make the fans go full throttle.

    Casper
    --
    Expressed in this posting are my opinions. They are in no way related
    to opinions held by my employer, Sun Microsystems.
    Statements on Sun products included here are not gospel and may
    be fiction rather than truth.

  7. Re: Sun Blade 1000/2000 power management

    Dave writes:

    >I do notice when it first switched on the fans run at full speed. They
    >then reduce speed after a while. If I switch off (either with 'poweroff'
    >or init 5), the fans speed up before the systems powers off. (I assume
    >the operating system is supposed to send some sort of 'signal' to a
    >power management system. As soon as the OS shuts down, that stops, so
    >the fans are set to run at full speed. This is my guess - I do not know
    >for sure).


    It's all related to picld; as soon as picld stops controlling the fans
    the firmware will make the fans go full throttle.

    Casper
    --
    Expressed in this posting are my opinions. They are in no way related
    to opinions held by my employer, Sun Microsystems.
    Statements on Sun products included here are not gospel and may
    be fiction rather than truth.

  8. Re: Sun Blade 1000/2000 power management

    Casper H.S. Dik wrote:
    > Dave writes:
    >
    >> I do notice when it first switched on the fans run at full speed. They
    >> then reduce speed after a while. If I switch off (either with 'poweroff'
    >> or init 5), the fans speed up before the systems powers off. (I assume
    >> the operating system is supposed to send some sort of 'signal' to a
    >> power management system. As soon as the OS shuts down, that stops, so
    >> the fans are set to run at full speed. This is my guess - I do not know
    >> for sure).

    >
    > It's all related to picld; as soon as picld stops controlling the fans
    > the firmware will make the fans go full throttle.
    >
    > Casper


    Thanks. I guessed it would be something along those lines.

    The speed control does make a significant different to the noise level.
    I've had a couple of U60's with the low-noise kit, and found it next to
    useless. On the Blade 2000, it is quite effective.

  9. Re: Sun Blade 1000/2000 power management

    Casper H.S. Dik wrote:
    > Dave writes:
    >
    >> I do notice when it first switched on the fans run at full speed. They
    >> then reduce speed after a while. If I switch off (either with 'poweroff'
    >> or init 5), the fans speed up before the systems powers off. (I assume
    >> the operating system is supposed to send some sort of 'signal' to a
    >> power management system. As soon as the OS shuts down, that stops, so
    >> the fans are set to run at full speed. This is my guess - I do not know
    >> for sure).

    >
    > It's all related to picld; as soon as picld stops controlling the fans
    > the firmware will make the fans go full throttle.
    >
    > Casper


    Thanks. I guessed it would be something along those lines.

    The speed control does make a significant different to the noise level.
    I've had a couple of U60's with the low-noise kit, and found it next to
    useless. On the Blade 2000, it is quite effective.

  10. Re: Sun Blade 1000/2000 power management

    Is there anyone with a Blade 2000 and Cu US III who could kindly
    follow the procedure above and let me know?
    HD


  11. Re: Sun Blade 1000/2000 power management

    Is there anyone with a Blade 2000 and Cu US III who could kindly
    follow the procedure above and let me know?
    HD


  12. Re: Sun Blade 1000/2000 power management

    haydude schrieb:
    > Is there anyone with a Blade 2000 and Cu US III who could kindly
    > follow the procedure above and let me know?
    > HD
    >

    I have tested my Blade 2000 (Solaris 10U4) with your power.conf settings.
    After some time of inactivity the power button light begins to flash,
    the fans will be stopped and the screen will be darkened.

    But the power consumption only reduces from approx. 270W to 260W.

    Regards
    Andreas

  13. Re: Sun Blade 1000/2000 power management

    haydude schrieb:
    > Is there anyone with a Blade 2000 and Cu US III who could kindly
    > follow the procedure above and let me know?
    > HD
    >

    I have tested my Blade 2000 (Solaris 10U4) with your power.conf settings.
    After some time of inactivity the power button light begins to flash,
    the fans will be stopped and the screen will be darkened.

    But the power consumption only reduces from approx. 270W to 260W.

    Regards
    Andreas

  14. Re: Sun Blade 1000/2000 power management

    Thank you Andreas,

    So it must be that when there is a Cu CPU the bus does not reduce the
    clock as specified in the manual, hence power consumption remains
    almost the same.

    HD
    On 10 Nov, 17:12, Andreas Wacknitz wrote:
    > haydude schrieb:> Is there anyone with a Blade 2000 and Cu US III who could kindly
    > > follow the procedure above and let me know?
    > > HD

    >
    > I have tested my Blade 2000 (Solaris 10U4) with your power.conf settings.
    > After some time of inactivity the power button light begins to flash,
    > the fans will be stopped and the screen will be darkened.
    >
    > But the power consumption only reduces from approx. 270W to 260W.
    >
    > Regards
    > Andreas




  15. Re: Sun Blade 1000/2000 power management

    Thank you Andreas,

    So it must be that when there is a Cu CPU the bus does not reduce the
    clock as specified in the manual, hence power consumption remains
    almost the same.

    HD
    On 10 Nov, 17:12, Andreas Wacknitz wrote:
    > haydude schrieb:> Is there anyone with a Blade 2000 and Cu US III who could kindly
    > > follow the procedure above and let me know?
    > > HD

    >
    > I have tested my Blade 2000 (Solaris 10U4) with your power.conf settings.
    > After some time of inactivity the power button light begins to flash,
    > the fans will be stopped and the screen will be darkened.
    >
    > But the power consumption only reduces from approx. 270W to 260W.
    >
    > Regards
    > Andreas




  16. Re: Sun Blade 1000/2000 power management

    Dave writes:

    >The speed control does make a significant different to the noise level.
    >I've had a couple of U60's with the low-noise kit, and found it next to
    >useless. On the Blade 2000, it is quite effective.


    Yep, the 2000 is nice and quiet (when the environment is OK).

    The 60 is pretty load and unfortunately so is my new 45.

    Casper
    --
    Expressed in this posting are my opinions. They are in no way related
    to opinions held by my employer, Sun Microsystems.
    Statements on Sun products included here are not gospel and may
    be fiction rather than truth.

  17. Re: Sun Blade 1000/2000 power management

    Dave writes:

    >The speed control does make a significant different to the noise level.
    >I've had a couple of U60's with the low-noise kit, and found it next to
    >useless. On the Blade 2000, it is quite effective.


    Yep, the 2000 is nice and quiet (when the environment is OK).

    The 60 is pretty load and unfortunately so is my new 45.

    Casper
    --
    Expressed in this posting are my opinions. They are in no way related
    to opinions held by my employer, Sun Microsystems.
    Statements on Sun products included here are not gospel and may
    be fiction rather than truth.

+ Reply to Thread