Reply
Thread Tools
Posts: 946 | Thanked: 1,650 times | Joined on Oct 2009 @ Germany
#51
Originally Posted by arbitrabbit View Post
P.S - I wish I could enable smartreflex but that makes my phone unstable. Hopefully PR1.2 would have better smartreflex drivers.
unlikely, as I already have all PR1.2 SDK patches compiled in.
 
Posts: 946 | Thanked: 1,650 times | Joined on Oct 2009 @ Germany
#52
Originally Posted by egoshin View Post
Frequency increase is for CPU and another stuff in CPU chip but not motherboard.
I would like to know some basic guidelines in this. Previously I assumed that doubling frequency == doubling CPU energy and doubling CPU voltage == quadruple CPU energy. But now it seems to be wrong.
to monitor the current for different voltages you could
Code:
modprobe bq27x00_battery
then change the voltages and average
Code:
watch -n1 /sys/class/power_supply/bq27200-0/current_now
about some period of time, preferably over ssh with the device idle and the screen locked.
 

The Following 2 Users Say Thank You to titan For This Useful Post:
Posts: 170 | Thanked: 75 times | Joined on Jun 2008 @ NYC
#53
While all the cpu manipulation is wonderful, is this really the best way to increase the battery?

With all the radios and the screen, wouldn't a single brightness click more than offset any minimal undervolt?
 
Posts: 946 | Thanked: 1,650 times | Joined on Oct 2009 @ Germany
#54
there is no best way, but the CPU+DSP is one important factor.
Ideally, the user should activate services (hard+software) only when needed.
we can optimize how the device works under load and how fast it can go back to idle.
for the same frequency, using lower voltage is definitely an improvement.

Originally Posted by Konceptz View Post
While all the cpu manipulation is wonderful, is this really the best way to increase the battery?
With all the radios and the screen, wouldn't a single brightness click more than offset any minimal undervolt?
 

The Following User Says Thank You to titan For This Useful Post:
Posts: 1,427 | Thanked: 2,077 times | Joined on Aug 2009 @ Sydney
#55
This was what master Konttori said about the battery usage a while back:

Originally Posted by konttori View Post
CPU itself has really only some effect on the total use-time. Display eats most of the power, cellmo is eating a lot with network traffic. Having higher speeds usually allows cpu to go back to idle sooner, so, indeed, it might even prolong the battery life if you overclock the CPU. Having said that, I don't especially suggest anybody to do it. I can only say that I have seen it running at 650, so within the suggested 10%.
 
Gusse's Avatar
Posts: 168 | Thanked: 206 times | Joined on Apr 2010 @ Finland
#56
Is there a way to include a script (to tune kernel setting) when display is locked automatically (i.e. after 1min, in settings/display).

Settings when device is active:
(freq range 500-850MHz)
Code:
echo 850000 > /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq
echo "125000 250000" > /sys/devices/system/cpu/cpu0/cpufreq/ondemand/avoid_frequencies
Settings when device is in-active (i.e idle &. night times):
(freq range: 125 or 500MHz)
Code:
echo 500000 > /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq
echo "250000 550000 600000 700000 750000 805000 850000" > /sys/devices/system/cpu/cpu0/cpufreq/ondemand/avoid_frequencies
Can keyboard slide ( slide GPIO 71) or lock (kb_lock GPIO 113) switch events be used as triggers for scripts? If yes, how?

Alarmd or fcron could be used to set these on/off, but I would like more dynamic way to set kernel tunings. Any ideas?
 
Posts: 1,224 | Thanked: 1,763 times | Joined on Jul 2007
#57
Originally Posted by arbitrabbit View Post
P.S - I wish I could enable smartreflex but that makes my phone unstable. Hopefully PR1.2 would have better smartreflex drivers.
You have smartreflex enabled, otherwise you could not change the voltages.

The feature you are missing is actually "auto compensation".
__________________
My repository

"N900 community support for the MeeGo-Harmattan" Is the new "Mer is Fremantle for N810".

No more Nokia devices for me.
 

The Following 2 Users Say Thank You to Matan For This Useful Post:
Posts: 1,224 | Thanked: 1,763 times | Joined on Jul 2007
#58
Originally Posted by Gusse View Post
Is there a way to include a script (to tune kernel setting) when display is locked automatically (i.e. after 1min, in settings/display).

Settings when device is active:
(freq range 500-850MHz)
Code:
echo 850000 > /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq
echo "125000 250000" > /sys/devices/system/cpu/cpu0/cpufreq/ondemand/avoid_frequencies
Settings when device is in-active (i.e idle &. night times):
(freq range: 125 or 500MHz)
Code:
echo 500000 > /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq
echo "250000 550000 600000 700000 750000 805000 850000" > /sys/devices/system/cpu/cpu0/cpufreq/ondemand/avoid_frequencies
Can keyboard slide ( slide GPIO 71) or lock (kb_lock GPIO 113) switch events be used as triggers for scripts? If yes, how?

Alarmd or fcron could be used to set these on/off, but I would like more dynamic way to set kernel tunings. Any ideas?
dbus-scripts. See here: http://talk.maemo.org/showthread.php?t=48132
__________________
My repository

"N900 community support for the MeeGo-Harmattan" Is the new "Mer is Fremantle for N810".

No more Nokia devices for me.
 

The Following 2 Users Say Thank You to Matan For This Useful Post:
Posts: 324 | Thanked: 739 times | Joined on Jun 2009 @ São Paulo, Brazil
#59
Originally Posted by egoshin View Post
BFS wouldn't have an advantage in single core CPU.
In fact, bfs has more advantages if you have less cpus. I compiled a kernel with the old CK's pathes for the N810 (diablo, maemo4) and the gain was huge, I would strongly recommend testing a kernel pathed with bfs, maybe it would increase performance without overclocking.
 
Posts: 992 | Thanked: 995 times | Joined on Dec 2009 @ California
#60
Originally Posted by traysh View Post
In fact, bfs has more advantages if you have less cpus. I compiled a kernel with the old CK's pathes for the N810 (diablo, maemo4) and the gain was huge, I would strongly recommend testing a kernel pathed with bfs, maybe it would increase performance without overclocking.
I would like to see some description of your measurements.

EDIT: I work with performance products and I have some problems with CFS on Intel platform for multi-cores. It works fine on single core.

Last edited by egoshin; 2010-04-22 at 18:47.
 

The Following User Says Thank You to egoshin For This Useful Post:
Reply


 
Forum Jump


All times are GMT. The time now is 00:22.