Constantin Gonzalez wrote:
> Hi Tom,
>
> thanks for the clarification.
>
> Just to be sure: The CPU is brand new and it is of the 0x10 family, so 
> I assumed
> that p-state scaling works. Or is the issue more subtle than that?

No. Tom thought your CPU was family 0xF. Your CPU family is supported. 
The reason that you are seeing the P-state transitions in PowerTop, but 
not in kstats is that PowerTop is averaging the P-states over a time 
period. kstats is reporting the current P-state to you as a snapshot. 
And at the moment you run the kstat command, your CPU is being run at 
its highest frequency.

And to answer one other question you had, no, event-mode cpupm (which is 
the default mode) does not use the intermediary P-states. It uses only 
the highest and lowest P-states. It also transitions between these 
P-states much faster than poll-mode, which is why you are never catching 
it at its lowest frequency when you run kstat.

Mark

>
> Thanks,
>   Constantin
>
>
> Tom de Waal wrote:
>> Constantin,
>>
>> The issue is with your CPU. AMD X2's aren't p-state invariant. 
>> Therefore p-state scaling would impact all timing on your system and 
>> therefore its   not supported. AMD X3 and newer CPU's do work (as 
>> they are p-state invariant).
>>
>> C-states are supported and do work (as you have observed).
>>
>> If you p-state scaling you need to get a newer CPU....
>>
>> Regards,
>>
>> Tom.
>>
>>
>>
>>
>> Constantin Gonzalez wrote:
>>> Hi,
>>>
>>> I'm building a new OpenSolaris home server and have poked around 
>>> some power
>>> management issues. Here are some of my observations and it would be 
>>> nice if
>>> someone could confirm if I'm corrent and maybe provide some 
>>> suggestions on how
>>> I can improve stuff.
>>>
>>> - Motherboard is an Asus M3A78-CM. It supports ACPI 2.0
>>>
>>> - CPU is an AMD Athlon II X2
>>>
>>> - OS is OpenSolaris 2009.06, latest support repository update.
>>>
>>> - OpenSolaris power management works out of the box. Thanks to the 
>>> PM team!
>>>
>>>   Powertop shows me that:
>>>   - During idle, the CPU is mostly in it's C1 state (close to 90% of 
>>> the time)
>>>   - The CPU is running at only 800 MHz when idle
>>>   - Occassionally, it bumps up the speed to it's highest frequency.
>>>   - But it never uses a P-state in between.
>>>     Why is this? Is the use of intermediary P-states not implemented in
>>>     OpenSolaris (yet)?
>>>
>>> - I've been trying to learn more about the CPU with kstat -m cpu_info:
>>>   - I can see all supported MHz settings, consistent with what 
>>> powertop shows
>>>     me.
>>>   - But current_clock_Hz never shows me the smallest setting (800 
>>> MHz) despite
>>>     powertop reporting it. It seems to only show the highest setting.
>>>   - Could it be that my use of kstat triggered the CPU to go it it's 
>>> full
>>>     frequency?
>>>   - Or am I just unlucky and kstat only shows the instances where 
>>> the CPU
>>>     runs fast by chance?
>>>   - Or something else?
>>>
>>> - I'd like to get temperature and fan statistics from the 
>>> motherboard, but
>>>   have not yet found a way to do this:
>>>
>>>   - scanning kstat -l for anything acpi related only reveals basic acpi
>>>     variables, but nothing temperature or fan related:
>>>     acpi:0:acpi:class
>>>     acpi:0:acpi:crtime
>>>     acpi:0:acpi:preferred_pm_profile
>>>     acpi:0:acpi:S3
>>>     acpi:0:acpi:snaptime
>>>     acpi_drv:0:battery warning:class
>>>     acpi_drv:0:battery warning:crtime
>>>     acpi_drv:0:battery warning:enabled
>>>     acpi_drv:0:battery warning:low capacity threshold
>>>     acpi_drv:0:battery warning:snaptime
>>>     acpi_drv:0:battery warning:warn capacity threshold
>>>     acpi_drv:0:power:class
>>>     acpi_drv:0:power:crtime
>>>
>>>   - The tzmon module is not loaded, so using tzmon.d (from one of 
>>> the pm related
>>>     threads) does not work for me.
>>>
>>>   - The tzmon man page mentions ACPI 3.0 and my motherboard only 
>>> supports ACPI
>>>     2.0. Is this the reason why OpenSolaris didn't enable the tzmon 
>>> kernel
>>>     module?
>>>     Loading the module by hand doesn't seem to do anything and 
>>> tzmon.d doesn't
>>>     do anything, so I have to assume tzmon simply doesn't work if the
>>>     motherboard is ACPI 2.0 only.
>>>
>>>   - Is there a (potentially hidden) way to still get temperature/fan 
>>> stats from
>>>     the motherboard? Do we have any ACPI tools I can play with?
>>>
>>> Again, kudos to the PM team, I'm very pleased PM is working well for 
>>> my system.
>>>
>>> I'd just like to go the next step and see P-states supported as well 
>>> as being
>>> able to read temperature and fan stats from the motherboard. Any 
>>> help in getting
>>> to this is greatly appreciated!
>>>
>>>
>>> Cheers,
>>>    Constantin
>>>
>>
>>
>


Reply via email to