Centrino not scaling to lowest frequency

Centrino not scaling to lowest frequency

[Felix Homann]

Hi,

on my Acer Extensa 4100 (BIOS S3C11) with a 1.60GHz Centrino I'm running a 
custom Linux kernel 2.6.12.3 with acpi-dsdt-patch applied and the custom DSDT 
"ACER-Aspire_1691WLMi-S3C11-custom.asl" from acpi.sourceforge.net.

Almost everything is working fine, except that the CPU can't be scaled down to 
600 MHz as it should (should it?).

Can anybody give me a hint how to get to 600 MHz on my system?

Here's some maybe relevant information about my system:

$ cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies
1600000 1333000 1067000 800000

$ cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver
centrino

$ cat /proc/cpuinfo
processor       : 0
vendor_id       : GenuineIntel
cpu family      : 6
model           : 13
model name      : Intel(R) Pentium(R) M processor 1.60GHz
stepping        : 8
cpu MHz         : 798.157
cache size      : 2048 KB
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 2
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca 
cmov pat clflush dts acpi mmx fxsr sse sse2 ss tm pbe nx est tm2
bogomips        : 1576.96


Thanks in advance,

Felix




-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf

Re: Centrino not scaling to lowest frequency

[Dominik Brodowski]

On Mon, Aug 15, 2005 at 11:44:11PM +0200, Felix Homann wrote:
> Hi,
> 
> on my Acer Extensa 4100 (BIOS S3C11) with a 1.60GHz Centrino I'm running a 
> custom Linux kernel 2.6.12.3 with acpi-dsdt-patch applied and the custom DSDT 
> "ACER-Aspire_1691WLMi-S3C11-custom.asl" from acpi.sourceforge.net.
> 
> Almost everything is working fine, except that the CPU can't be scaled down to 
> 600 MHz as it should (should it?).
> 
> Can anybody give me a hint how to get to 600 MHz on my system?

The patched DSDT does not provide the necessary values for 600MHz --
possibly because your CPU does not support 600 MHz.

	Dominik


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf

Re: Centrino not scaling to lowest frequency

[Felix Homann]

Hi

in the meantime I found out that although my system claims (sticker) to 
contain a Pentium M 725  (a Dothan) in fact is does contain a Pentium M 730J 
(a Sonoma). 

A look into Intel's documentation 30220913.pdf then reveals (Table 1) a Lowest 
Freq. Mode at 800 MHz. Hence it seems that Dominik is right and my CPU 
doesn't support 600 MHz  (if I get it right). So everything seems to be OK.

On the other hand, in Windows I _do_ get 600 MHz without any further trickery. 
And at 600 MHz it seems to give me a lot more battery life, too. How does 
that work?

(Actually, I initially wondered why Windows reported so much more battery time 
as opposed to Linux, and then found out that Windows can run the CPU at 600 
MHz.) 

Furthermore, are there really only 4 frequencies specified for the Sonoma?

Thanks again,

Felix

On Tuesday 16 August 2005 10:06, Dominik Brodowski wrote:
> > Can anybody give me a hint how to get to 600 MHz on my system?
>
> The patched DSDT does not provide the necessary values for 600MHz --
> possibly because your CPU does not support 600 MHz.
>
> 	Dominik


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf

Re: Centrino not scaling to lowest frequency

[Ow Mun Heng]

On Thu, 2005-08-18 at 18:47 +0200, Felix Homann wrote:
> On the other hand, in Windows I _do_ get 600 MHz without any further trickery. 
> And at 600 MHz it seems to give me a lot more battery life, too. How does 
> that work?

On my Machine, 1.4ghz dothan, windows can go down to 200Mhz but it's
just lying. :-)

> (Actually, I initially wondered why Windows reported so much more battery time 
> as opposed to Linux, and then found out that Windows can run the CPU at 600 
> MHz.) 

Actually, and unfortunately, Power management in linux sucks. :-) but
people here in this list and project utopia and Gnome-power-manager
(read about the exclusive in Ed4 of MyOSS Magazine -
http://mag.my-opensource.org) is aiming to change all that. ACPI code
isn't followed 100% by most vendors and thus, you get half baked
solutions.

my 2 cents



-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf

Re: Centrino not scaling to lowest frequency

[Andrew Haninger]

On 8/18/05, Ow Mun Heng <Ow.Mun.Heng-Sjgp3cTcYWE@public.gmane.org> wrote:
> On my Machine, 1.4ghz dothan, windows can go down to 200Mhz but it's
> just lying. :-)
How do you mean? My 1.7GHz T42 sometimes shows 200MHz in Control Panel
| System at really low load. It's not really 200MHz? Is this the
throttling-on-top-of-CPU-scaling thing Dominik mentioned?

I'm hoping to be able to get to the same speed when I install Linux on my TP.

-Andy


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf

Re: Centrino not scaling to lowest frequency

[Dominik Brodowski]

Hi,

On Thu, Aug 18, 2005 at 06:47:48PM +0200, Felix Homann wrote:
> On the other hand, in Windows I _do_ get 600 MHz without any further trickery. 
> And at 600 MHz it seems to give me a lot more battery life, too. How does 
> that work?

That's because other Operating Systems do "throttling" on top of "CPU
frequency scaling", even though that doesn't save you any energy. The better
energy usage rates of Windows are a known problem; hopefully the dynamic
tick patch will improve the situation in this regard.

	Dominik


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf

Re: Centrino not scaling to lowest frequency

[Ow Mun Heng]

On Wed, 2005-08-24 at 12:50 +0200, Dominik Brodowski wrote:
> Hi,
> 
> On Thu, Aug 18, 2005 at 06:47:48PM +0200, Felix Homann wrote:
> > On the other hand, in Windows I _do_ get 600 MHz without any further trickery. 
> > And at 600 MHz it seems to give me a lot more battery life, too. How does 

> The better
> energy usage rates of Windows are a known problem; hopefully the dynamic
> tick patch will improve the situation in this regard.

Hmm.. that's a nice way to put it :-)

ANyway, where's this dynamic tick patch? Is it the one where you drop
the hertz from 1K to 100?



-- 
Ow Mun Heng
Gentoo/Linux on DELL D600 1.4Ghz 1.5GB RAM
98% Microsoft(tm) Free!! 
Neuromancer 21:57:52 up 12:50, 6 users, load average: 2.17, 1.88, 1.41 




-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf

Re: Centrino not scaling to lowest frequency

[Dominik Brodowski]

On Thu, Aug 25, 2005 at 09:57:53PM +0800, Ow Mun Heng wrote:
> On Wed, 2005-08-24 at 12:50 +0200, Dominik Brodowski wrote:
> > Hi,
> > 
> > On Thu, Aug 18, 2005 at 06:47:48PM +0200, Felix Homann wrote:
> > > On the other hand, in Windows I _do_ get 600 MHz without any further trickery. 
> > > And at 600 MHz it seems to give me a lot more battery life, too. How does 
> 
> > The better
> > energy usage rates of Windows are a known problem; hopefully the dynamic
> > tick patch will improve the situation in this regard.
> 
> Hmm.. that's a nice way to put it :-)
> 
> ANyway, where's this dynamic tick patch? Is it the one where you drop
> the hertz from 1K to 100?

Uh, no. http://lwn.net/Articles/138969/ has an explanation, don't know where
the most current patch is -- best check the LKML archives for that.

	Dominik


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf

Re: Centrino not scaling to lowest frequency

[Ow Mun Heng]

On Thu, 2005-08-25 at 16:04 +0200, Dominik Brodowski wrote:
> On Thu, Aug 25, 2005 at 09:57:53PM +0800, Ow Mun Heng wrote:
> > On Wed, 2005-08-24 at 12:50 +0200, Dominik Brodowski wrote:
> > > Hi,
> > > 
> > > On Thu, Aug 18, 2005 at 06:47:48PM +0200, Felix Homann wrote:
> > > > On the other hand, in Windows I _do_ get 600 MHz without any further trickery. 
> > > > And at 600 MHz it seems to give me a lot more battery life, too. How does 
> > 
> > > The better
> > > energy usage rates of Windows are a known problem; hopefully the dynamic
> > > tick patch will improve the situation in this regard.
> > 
> > Hmm.. that's a nice way to put it :-)
> > 
> > ANyway, where's this dynamic tick patch? Is it the one where you drop
> > the hertz from 1K to 100?
> 
> Uh, no. http://lwn.net/Articles/138969/ has an explanation, don't know where
> the most current patch is -- best check the LKML archives for that.

Ah.. Thanks for the link.

Here's the patches http://muru.com/linux/dyntick/

Anyone played with it just yet??
> 
> 	Dominik

-- 
Ow Mun Heng
Gentoo/Linux on DELL D600 1.4Ghz 1.5GB RAM
98% Microsoft(tm) Free!! 
Neuromancer 22:18:54 up 13:11, 6 users, load average: 2.20, 1.95, 1.81 




-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf

Re: Centrino not scaling to lowest frequency

[Felix Homann]

On Thursday 25 August 2005 16:20, Ow Mun Heng wrote:
>
> Here's the patches http://muru.com/linux/dyntick/
>

Hi,

I guess the latest patch is this:

 http://lkml.org/lkml/diff/2005/8/7/2/1

Regards,

Felix




-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf

Re: Centrino not scaling to lowest frequency

[Erik Slagter]

[-- Attachment #1: Type: text/plain, Size: 556 bytes --]

On Thu, 2005-08-25 at 16:04 +0200, Dominik Brodowski wrote:
> > ANyway, where's this dynamic tick patch? Is it the one where you drop
> > the hertz from 1K to 100?
> 
> Uh, no. http://lwn.net/Articles/138969/ has an explanation, don't know where
> the most current patch is -- best check the LKML archives for that.

Interesting. Over here (dell inspiron 9300, pentium M) it doesn't quite
work, in the contrary, with the mechanism enabled, it consumes more
power (default 3:15, with patch enabled 2:55 hours). I have no clue what
causes this...

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 2115 bytes --]

Re: Centrino not scaling to lowest frequency

[Felix Homann]

On Wednesday 24 August 2005 12:50, Dominik Brodowski wrote:
> That's because other Operating Systems do "throttling" on top of "CPU
> frequency scaling", even though that doesn't save you any energy. The
> better energy usage rates of Windows are a known problem; hopefully the
> dynamic tick patch will improve the situation in this regard.
>
> 	Dominik
>

Thanks for the info. The dynamic tick patch sounds really interesting! I'll 
give it a try.

BTW, Intel's CPU identifiction utility reports a system bus frequency of 400 
MHz when the CPU is reported to work at 600 MHz. Usually the system bus is 
running at 533 MHz. Is this the effect of "throttling" on top of "CPU 
frequency scaling" or is this something different?

Thanks,

Felix


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf

Re: Centrino not scaling to lowest frequency

[Dominik Brodowski]

On Thu, Aug 25, 2005 at 05:04:50PM +0200, Felix Homann wrote:
> On Wednesday 24 August 2005 12:50, Dominik Brodowski wrote:
> > That's because other Operating Systems do "throttling" on top of "CPU
> > frequency scaling", even though that doesn't save you any energy. The
> > better energy usage rates of Windows are a known problem; hopefully the
> > dynamic tick patch will improve the situation in this regard.
> >
> > 	Dominik
> >
> 
> Thanks for the info. The dynamic tick patch sounds really interesting! I'll 
> give it a try.
> 
> BTW, Intel's CPU identifiction utility reports a system bus frequency of 400 
> MHz when the CPU is reported to work at 600 MHz. Usually the system bus is 
> running at 533 MHz. Is this the effect of "throttling" on top of "CPU 
> frequency scaling" or is this something different?

Uh, this is strange. I know of no implementation on Intel CPUs to scale the
system bus frequency.

However.... Are you relying on centrino-decode of cpufrequtils? IIRC
it is broken for the 533 MHz case, so if it reports 400 MHz it actually means
600 MHz.  *clicketiclick* Yes, cpufrequtils-0.3 still contains this bug;
need to release 0.4 soon.

	Dominik


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf

Re: Centrino not scaling to lowest frequency

[Felix Homann]

On Thursday 25 August 2005 17:37, Dominik Brodowski wrote:
> > BTW, Intel's CPU identifiction utility reports a system bus frequency of
> > 400 MHz when the CPU is reported to work at 600 MHz. Usually the system
> > bus is running at 533 MHz. Is this the effect of "throttling" on top of
> > "CPU frequency scaling" or is this something different?
>
> Uh, this is strange. I know of no implementation on Intel CPUs to scale the
> system bus frequency.
>
> However.... Are you relying on centrino-decode of cpufrequtils? IIRC
> it is broken for the 533 MHz case, so if it reports 400 MHz it actually
> means 600 MHz.  *clicketiclick* Yes, cpufrequtils-0.3 still contains this
> bug; need to release 0.4 soon.
>
> 	Dominik

Sorry, I guess I haven't made myself clear enough in the previous mail. The 
Intel tool I mentioned is a Windows utility available here:

http://support.intel.com/support/processors/tools/piu/

When in Windows my Pentium M 730 is reported to run at 600 MHz the Intel tool 
reports a 400 MHz system bus. When the processor is running at  1.60 GHz it 
reports a 533 MHz system bus.

Is there a really reliable tool to get information on the respective 
frequencies?

Regards,

Felix


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf

Re: Centrino not scaling to lowest frequency

[Johan Vromans]

Felix Homann <fexpop-53koH/AXb86i2/dY4+mD5w@public.gmane.org> writes:

> $ cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies
> 1600000 1333000 1067000 800000

On my Acer Travelmate 4001WLMi the CPU frequencies in the DSDT were
wrong. Did you check you have the right DSDT and frequencies?

-- Johan

See also: http://www.squirrel.nl/people/jvromans/articles/TM4001WLMi/acpi/speedstep.html


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf

Re: Re: Centrino not scaling to lowest frequency

[Felix Homann]

Thanks Johan,

actually I'm quite sure that the CPU frequencies in the DSDT are wrong. 
Unfortunately I can't figure out where they are defined in the DSDT. 

Neither the dissassembled original DSDT nor the custom one (which adds battery 
read-out support) contain any _PSS entries. Isn't that what I should be 
looking for?

Where else might frequencies be defined in the DSDT?

Thanks again,

Felix

On Tuesday 16 August 2005 10:44, Johan Vromans wrote:
> Felix Homann <fexpop-53koH/AXb86i2/dY4+mD5w@public.gmane.org> writes:
> > $ cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies
> > 1600000 1333000 1067000 800000
>
> On my Acer Travelmate 4001WLMi the CPU frequencies in the DSDT were
> wrong. Did you check you have the right DSDT and frequencies?
>
> -- Johan
>
> See also:
> http://www.squirrel.nl/people/jvromans/articles/TM4001WLMi/acpi/speedstep.h
>tml


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf

Re: Centrino not scaling to lowest frequency

[Johan Vromans]

Felix Homann <fexpop-53koH/AXb86i2/dY4+mD5w@public.gmane.org> writes:

> Where else might frequencies be defined in the DSDT?

My DSDT has

Scope(_PR_) {
    Processor(CPU0, 0, 0x1010, 0x6) {
        Name(_PCT, Package(0x2) { ... }
        Method(_PSS) {
            If(CLNM) {
                Return(CPSS)
            }
            Return(PPSS)
        }
        Name(CPSS, Package(0x1) {
            Package(0x6) {
                0x5dc,		// 1500 Mhz
                ...
            },
        })
        Name(PPSS, Package(0x5) {
            Package(0x6) {
                0x5dc,		// 1500 Mhz
                ...
            },
            Package(0x6) {
		0x04b0,		// 1200 Mhz 
                ...

... and so on ...

-- Johan



-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf