Active Topics

 


Reply
Thread Tools
Posts: 463 | Thanked: 103 times | Joined on Jul 2010 @ Mumbai, India
#11
Originally Posted by corduroysack View Post
already did a some weeks ago, after following karam's thread i changed the powerbias setting in my profile noticed no change still draining the same. reset powerbias and changed vfs setting and battery drain was improved a great deal.
What is ur vfs setting buddy?? From where do u change it?
 

The Following User Says Thank You to funkmunk For This Useful Post:
corduroysack's Avatar
Posts: 939 | Thanked: 366 times | Joined on Dec 2010 @ U.K.
#12
standard usage? well standard usage for me is all day everyday connected to wifi answering a few calls forwarding major amounts of texts, making notes, playing music, games occasionally.
so my standard usage is different to someone else's standard usage.
__________________
"I Reject Your Reality & Substitute My Own" Adam Savage
 
corduroysack's Avatar
Posts: 939 | Thanked: 366 times | Joined on Dec 2010 @ U.K.
#13
Originally Posted by funkmunk View Post
What is ur vfs setting buddy?? From where do u change it?
used swappolube at first mate changed vfs cache pressure to 10, but then used karam's battery script as this changes it from 10 when idle to 100 when you are using so your getting best of both worlds.
__________________
"I Reject Your Reality & Substitute My Own" Adam Savage
 

The Following User Says Thank You to corduroysack For This Useful Post:
Posts: 2,802 | Thanked: 4,491 times | Joined on Nov 2007
#14
Originally Posted by a dick View Post
Even on idle, the frequency keeps on changing, with powerbias, in my opinion, the fluctuation should be reduced improving battery life.
If reducing frequency "jitter" is your goal, you should concentrate on tuning the sampling_rate and up_threshold parameters, or even using the conservative governor. Note that this doesn't necessarily equate to longer battery life.

110 is optimum in my opinion because, in higher values, i noticed the cpu is not using the maximum frequency most of the times, if not at all.
That's the entire point of powersave_bias, you sacrifice peak performance in exchange for (hopefully) reduced energy consumption.

If someone can test the performance outcomes on idle and on usage and get talk time, usage times, idle times and power consumption per hour or more for each of those, I'd be glad
You need to define a specific workload that can be executed non-interactively (for accurate testing), and then compare the total (ie, spontaneous measurements are useless) energy expended for performing that load at various configurations.
 
Banned | Posts: 32 | Thanked: 11 times | Joined on Aug 2011
#15
Originally Posted by lma View Post
If reducing frequency "jitter" is your goal, you should concentrate on tuning the sampling_rate and up_threshold parameters, or even using the conservative governor. Note that this doesn't necessarily equate to longer battery life.



That's the entire point of powersave_bias, you sacrifice peak performance in exchange for (hopefully) reduced energy consumption.



You need to define a specific workload that can be executed non-interactively (for accurate testing), and then compare the total (ie, spontaneous measurements are useless) energy expended for performing that load at various configurations.
With hw acceleration on, peek performance in basically visual for most of us. No body likes lags even if the process is damn slow.
I'm focusing on using needed frequency and not going to up highest for almost no reason at all. At the same time, Not ignoring the highest frequency but using it only if its completely necessary.
Although conservative governor is a good choice too *i just googled it*

In terms of a phone, Its either we're using it or we're not using it.
In my opinion, even powersave governor would be useful because the processes stop the moment we close the app and phone goes back to idle. In idle, lowest possible frequency is used when powersave_bias is used which would be 500 and wont go higher too often an 250 is disabled and after installing alotta apps, it automatically goes to 500 *thats why 250 is removed from ideal kernel settings*, a constant 500mhz wont consume as much power as fluctuating btwn 500 and 600/700.

I'm looking for something that doesn't reduce the frequency to minimum and lags everything but not too high that its unnecessary and uses alot of power. Its hard to explain. I don't want extremes of either. I want it to be controllable so it saves power and doesn't get too slow. Makes Powersave_bias the best choice.


Powersave governor would cause lags. *the last time i tried it without knowing what it is LOL*
I'm about to switch to conservative gorernor to see the results.

A specific worklod for an hour.. perhaps youtube with wifi on for an hour and see the difference in power consumed. thats high end, Music on for an hour for casual. and idle for no usage. and time difference in 1 hour and difference btwn mAh recorded.

Edit: , I've noticed that on conservative, while calling, the max frew is 750 and on ondemand with powersave, its fluctuating btwn 600 and 700..
plus the effects are faster on ondemand and conservative sticks to one frequency for too long. For example, If I'm using the phone and then close all apps, It take around 10 seconds to come back to 500 and does it slowly step by step eg 750, 700, 600, 550 then 500. and same goes for when you use your phone from idle to start a process.

Last edited by a dork; 2011-08-11 at 11:03.
 
Posts: 69 | Thanked: 55 times | Joined on Nov 2009
#16
On my HTC G1 optimal values were:

POWERSAVE_BIAS=200;
SAMPLING_RATE=1000000;
CPU_UP_THRESHOLD=45;

I'm testing them now on the N900.

Last edited by farmatito; 2011-08-11 at 11:59.
 
Banned | Posts: 32 | Thanked: 11 times | Joined on Aug 2011
#17
Originally Posted by farmatito View Post
On my HTC G1 optimal values were:

POWERSAVE_BIAS=200;
SAMPLING_RATE=1000000;
CPU_UP_THRESHOLD=45;

I'm testing them now on the N900.
how did it turn out?
 
Posts: 69 | Thanked: 55 times | Joined on Nov 2009
#18
Originally Posted by a dick View Post
how did it turn out?
Not bad, lasts about 2 days with standard battery and normal usage pattern.
 

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


 
Forum Jump


All times are GMT. The time now is 16:08.