Core i7 not scaling frequency down

Core i7 not scaling frequency down

[Felix von Leitner]

Hi,

I just got a Nehalem CPU and it works OK except that the fans are always
running full power and it produces heat like an oven.

  $ cat /proc/acpi/processor/P001/info
  processor id:            0
  acpi id:                 1
  bus mastering control:   yes
  power management:        yes
  throttling control:      no
  limit interface:         no
  $

Oh, no throttling and no limit support?  How can this be?  I'm using
2.6.30 but this appears to be bogus; Windows scales down the CPU
frequency OK.

Does this mean my BIOS ACPI tables are bogus?

I get this message at boot time, but it appears to be unrelated:

  ACPI Warning (tbutils-0246): Incorrect checksum in table [OEMB] - 00, should be FE [20090320]

Any ideas what I could do now to get the CPU to slow down?

Thanks,

Felix

Re: Core i7 not scaling frequency down

[Frans Pop]

Felix von Leitner wrote:
> Any ideas what I could do now to get the CPU to slow down?

What cpu frequency governor are you using?

Take a look in /sys/devices/system/cpu/cpu0/cpufreq/. I'd suggest ensuring 
you're using the "ondemand" governor as a first step in managing CPU 
speed.

Cheers,
FJP

Re: Core i7 not scaling frequency down

[Arjan van de Ven]

On Sat, 4 Jul 2009 22:37:32 +0200
Felix von Leitner <felix-linuxkernel@fefe.de> wrote:

> Hi,
> 
> I just got a Nehalem CPU and it works OK except that the fans are
> always running full power and it produces heat like an oven.
> 
>   $ cat /proc/acpi/processor/P001/info
>   processor id:            0
>   acpi id:                 1
>   bus mastering control:   yes
>   power management:        yes
>   throttling control:      no
>   limit interface:         no
>   $
> 
> Oh, no throttling and no limit support?  How can this be?  I'm using
> 2.6.30 but this appears to be bogus; Windows scales down the CPU
> frequency OK.

how do you determine linux isn't scaling down? PowerTOP?
(if not powertop.. you might want to use that ;-)

>   ACPI Warning (tbutils-0246): Incorrect checksum in table [OEMB] -
> 00, should be FE [20090320]

that's pretty bad, but might not be a table critical for C/P states...


-- 
Arjan van de Ven 	Intel Open Source Technology Centre
For development, discussion and tips for power savings, 
visit http://www.lesswatts.org