thinkpad x60 Battery Life?(CoreDuo freq scaling problem?)

[Ari El]

Hi Paul,

   I've been having the same battery life problem in my X60s for many
months, with Dapper, and now the same with Edgy.

   I could confirm there is no difference at all in power consumption in
this machine produced by activating different governors. When using
'performance' linux (edgy) reports a constant frequency of 1.67Ghz; when
using 'powersave' linux reports a constant 1Ghz.

   With all modules unloaded (bluetooth, irda, pcmcia, usb, e1000) and the
hardware radio switch set to off, and laptop_mode set to spindown the disk
after 20sec inactivity, and display set to Level 0 (minimum), power drain
reported by acpitool -B does not change at all with the governor: 9600mW
when the machine is idle; 9850mW when I'm writing (I use abiword a lot).
These are good numbers (about 7:30 hours) but with all modules unloaded
(bluetooh, e1000, irda, pcmcia, usb) and wifi off and display in level 0
(nearly unusable). 

In WinXP i get about 8:00 hours with the subsystems on (except bluetooth and
irda), display in level 3, HD spindown set to 30secs, wifi on.  (Display in
level 3 drains about 600mW more than level 0)

The bottom line is the frequency scaling in linux for this machine does not
seem to produce any result as regards battery drain.

I asked Fabrice (the developer of the linux-hpc cpu undervolting patch)
about this problem. He wrote a small app to extract some info from the core
duo CPU and also measure actual CPU speed by different methods.

   This thread has a link to his source code: 

       
https://www.dedigentoo.org/pipermail/linux-phc-user/2006-October/000028.html

   This thread has my results when running the app: 

       
https://www.dedigentoo.org/pipermail/linux-phc-user/2006-October/000032.html

    You can see there that while the CPU internal registers show that the
speed is 1Ghz (powersave), the actual CPU speed as estimated by different
methods is 1.6 GHz which is the machine maximum speed. 

    This might be showing either a bug in the CPU firmware or in the kernel
frequency scaling code. Or maybe the code of Fabrice's utility has a
problem. 

    Given that you are part of the Ubuntu Core Dev team maybe you can
confirm Fabrice's code to measure actual cpu speed (or ask some kernel dev
to do it), and thus confirm if there is a real problem with this CPU or
Kernel or this is a false alarm. I didn't open a bug in launchpad for this.

    From what I have in other forums, this problem might be affecting other
laptop brands with coreduo as well (lots of users are complaining about
battery life when comparing with what they get with WinXP)


Thanks a lot in advance,
Ari




Paul Sladen-2 wrote:
> 
> On Thu, 20 Jul 2006, Thierry Nicola wrote:
>> My question is what is your battery life, and what can be done to
>> improve this. Til now I have no really tried to do anything(scaling CPU,
>> killing daemons, scaling graphic chip...).
> 
> Bluetooth off:
> 
>   echo disabled | sudo tee -a /proc/acpi/ibm/bluetooth
> 
> Turn the blacklight brightness right down.
> 
> Enable hard-disk spin-down, by setting:
> 
>   ENABLE_LAPTOP_MODE=true
> 
> in:
> 
>   /etc/default/acpi-support
> 
> 'thinkpad-keys' spends alot of the time polling in the kernel on recent
> Lenovopads, you could kill if you don't use it (messy though and won't
> gain
> you much and will loose the keys being passed through).
> 
> 	-Paul
> -- 
> High on a tall bridge, surrounded by noisy lorries.  Nottingham, GB
> 
> 
> -- 
> laptop-testing-team mailing list
> laptop-testing-team at lists.ubuntu.com
> https://lists.ubuntu.com/mailman/listinfo/laptop-testing-team
> 
> 

-- 
View this message in context: http://www.nabble.com/thinkpad-x60-Battery-Life--%21-tf1971716.html#a7064764
Sent from the laptop-testing-team mailing list archive at Nabble.com.