Live Wallpapers & CPU Governer - Xperia Arc General

I understand that the OnDemand CPU Governer calls for a higher frequency clock when the phone requires that extra performance. But what i don't get is why would it clock to a fixed 1GHz on my Arc device for every Live Wallpaper? This power seem unnecessary and is causing the phone to heat up and drain the battery even faster. Does not fit as titled ON DEMAND afterall.
I was thinking if Sony has implemented a new or updated CPU Governer on the Sony Xperia S regarding on this? Their Cosmic Flow WP do look pretty neat.

Related

Dynamic Overclocking of CPU?

Is this possible on Androids (mostly, this phone?) Before this I had a windows phone and I had a software where I could set a minimum and maximum cpu clock rate.. and then it will overclock dynamically as needed! If you are running a game and it needs more power, it will overclock it while the game is running.. dynamically without you having to change the clock-rate manually.
This will solve battery life problems when overclocking to 1.8ghz. There is no need for it to be draining the battery at 1.8ghz if you are only texting and probably even a 200mhz cpu can do that lol
Hope what I said makes sense.
it's easily done by installing Setcpu,provided that your phone has been rooted and kernel modded to allow overclocking.
Sent from my IDEOS X5
Actually it is already done in stock ROM without root, setCPU provides a way to change the default way of handling the CPU frequencies and governors.
The stock ROM comes right out of the factory can do "cpu stepping" itself, just like the intel "Speedstep" thing. But the factory powersaving profile (or if you wish to call the scaling) might not match our taste. Then there it comes the apps called Setcpu which enhances the cpu stepping behavior according to our preference. The enhancement considers the frequency of sampling cpu load, thresholds modifications, and other considerations.
In Setcpu, we can accord one of the following scaling setting that suits our taste. From my experience, each scaling setting behaves as follows:
ondemand - runs at Min or at Max (eg. 200 MHz or 1200 MHz)
interactive - adjusts frequencies according to the real load
conservative - behaves like "interactive" but incline to power saving
performance - always runs at Max
powersave - always runs at Min
userspace - disables Setcpu scaling and uses stock scaling
smartass - behaves like "interactive" but always runs at Min when LCD screen is off
And again, the phone must be rooted and kernel modded for Setcpu to work properly.
I've been using setcpu with conservative. I like conservative because it gives you a lot of control over the throttling of the cpu. It seems to work pretty good.
Are you getting any benefit from overclocking to 1.8GHz? I also overclocked to 1.8GHz, but I didn't see any noticeable improvement past 1.0-1.2GHz.
I like to use interactive Cox I enjoy the fast sampling rate and hence the high responsiveness.
Sent from my IDEOS X5
Personally I have not tried overclocking yet.. still need to update the kernel, but I'm afraid it might get unstable.
Didn't know that CPU could do it dynamically using different profiles.
Thanks all for the help
Tcm9669 said:
Is this possible on Androids (mostly, this phone?) Before this I had a windows phone and I had a software where I could set a minimum and maximum cpu clock rate.. and then it will overclock dynamically as needed! If you are running a game and it needs more power, it will overclock it while the game is running.. dynamically without you having to change the clock-rate manually.
This will solve battery life problems when overclocking to 1.8ghz. There is no need for it to be draining the battery at 1.8ghz if you are only texting and probably even a 200mhz cpu can do that lol
Hope what I said makes sense.
Click to expand...
Click to collapse
I have setcpu but the x5 is not listed....??

Does cpu frequency affects battery life?

The question is simple, the higer the freq. the more battery consumption? i am asking because i cannot see any difference from 1.4GHZ to 1.0GHZ, the battery consumption is the same under oxygen and francos kernels. i would be grateful if the experts would give us some advise or their opionion. I know it is subjective but i would like a second opinion.
Thanks alot guys
If you look with any cpu spy app, you cab see on what clock the processor is used. I mainly have it the lowest and sometimes higher. When you change the max, it should still automatically choose what clock is used, so battery should last as much as before, if not used in high clock.
Sent from my Huawei u8800 using XDA App
Invicta said:
The question is simple, the higer the freq. the more battery consumption? i am asking because i cannot see any difference from 1.4GHZ to 1.0GHZ, the battery consumption is the same under oxygen and francos kernels. i would be grateful if the experts would give us some advise or their opionion. I know it is subjective but i would like a second opinion.
Thanks alot guys
Click to expand...
Click to collapse
Probably not much difference, when you overclock, think the cpu voltage remains the same as it is on 800mhz... And the highest cpu freq is rarely even at use...
Invicta said:
The question is simple, the higer the freq. the more battery consumption? i am asking because i cannot see any difference from 1.4GHZ to 1.0GHZ, the battery consumption is the same under oxygen and francos kernels. i would be grateful if the experts would give us some advise or their opionion. I know it is subjective but i would like a second opinion.
Thanks alot guys
Click to expand...
Click to collapse
Well, thanks to most devoted users of U8800, we have somewhat better speed+battery life than stock also. To answer your question, yes depending on your activity the frequency have some impact, but overall shouldn't be huge in change, for example from .800 to 1.0GHz will not affect so much in daily use. However from .800 to 1.5GHz would make a somewhat huge gap difference. This doesn't mean it will drain faster if you do same activity as with .800 to 1.0GHz, for example check the watch, answer sms or few "entertainment" breaks. Only when using the phone over a longer period of time that's when you will notice the change of battery life with different frequency. Hope it clears up most hums and huhs for you. I am pretty sure some expert within this field will give a better explanation than me.
Bye~
higher freqs uses more energy, but lower uses less energy but do things slower (so energy consumption is longer). ALSO imo - if you set cpu to 1Ghz the lowest value so it always is 1ghz - it will not consume the same amount of energy if it's in idle mode - it's like your laptop - if cpu is working only in 4-7% of it's power - then the power consumption is lower no matter what freq - how do we know that? - because of heat - the more heat you get - the more energy was used. and when cpu is idle - it will not be hot.
So the answer is - if it saves then in VERY minimal amounts. But even so - i use min freq - 360mhz. it's good for me i do not get any lag so i use it.
I use the "Root System Tool Free", option CPU and I see the graphics of all clocks.
For ex. now at 245 mhz ->46%, at 368 ->10%, at 768 mhz ->18% .... and at 1612 mhz -> 1,8%, at 1804 mhz ->4,9%. Not very mutch use at 1804.
Oxygen-test-140911 + Franco.Kernel1709#1. Clock at 1804 Mhz by Menu-settings-cpu ... and smartassV2 (no profils).
The battery, I charge it all 24 hours. But I like my work... and testing things. When then will dead...I see...
ValenteL said:
I use the "Root System Tool Free", option CPU and I see the graphics of all clocks.
For ex. now at 245 mhz ->46%, at 368 ->10%, at 768 mhz ->18% .... and at 1612 mhz -> 1,8%, at 1804 mhz ->4,9%. Not very mutch use at 1804.
Oxygen-test-140911 + Franco.Kernel1709#1. Clock at 1804 Mhz by Menu-settings-cpu ... and smartassV2 (no profils).
The battery, I charge it all 24 hours. But I like my work... and testing things. When then will dead...I see...
Click to expand...
Click to collapse
cpu spy is more pretty, anyway all those apps just reads text file of cpu stat and that's it
Tommixoft said:
cpu spy is more pretty, anyway all those apps just reads text file of cpu stat and that's it
Click to expand...
Click to collapse
Thanks
But I like more the Root System Tool, because has also a Linux Console. I use it when I want make some cmd's Linux in #. I don't like the Terminal Emulator.
Well, there is a "catch" somewhere in there. The frequency does indeed affects the power consumption of the CPU and greatly at that too! But the thing is, your CPU is not the worst enemy of your battery life. Even though CPU consumes more power in higher frequencies, it still can not compete with what your screen LEDs or your GSM module or your GPS chip consumes leisurely. So, if you're looking at the overall picture -meaning if you're wondering if it will affect how long you'll be able to use your battery in your phone- the answer is, "yes but not so much". Especially if you're switching the CPU frequency based on the demand (like using smartass or on-demand governors)
Here are the thing that sucks your batteries life juice like a vampire :
Your Screen (especially background LEDs)
GSM module (talking, using GPRS/Edge/3G network communication)
GPS chip
Wireless module (this also includes Bluetooth, even though it does not consume as much as Wireless network access but everything is relative -think about playing music through A2DP headphones compared to having your wireless network active but not using it much-)
(oh yes, I love to use lots of parenthesis -and even this hyphenation thingy- )
Correct me if I'm wrong about anything by the way ..
Regards ..
I did some experiments with a msm8250 a while back and there's a graph here:
http://forum.xda-developers.com/showpost.php?p=14324649&postcount=3786
msm7x30 should be fairly similar though the graph is probably shallower since it's a smaller process size.
The CPU uses no power when it's not in use, even with the display on, the CPU is powered down completely when idle (power collapse).

Frequency and voltage settings.

Caveats: Every CPU and GPU does not come from the same bin, fabricated on the same date and possibly not manufactured in the same facility. They may each display different physical properties and a wider range of stability than others. What works for me may not work for you.
That being said, I've been stress testing my device with different settings for the past couple weeks trying to find a sweet spot of stability, speed, battery life and heat output.
I'm going to share two setups: my current one that I've stress tested for less than 24hrs but has proved stable through all conditions encountered thus far and my tried and true setup I've used for over a week with no trouble.
Tried and true setup:
Governor - ondemand
Range - 100MHz through 1.6GHz
100MHz - 800mV
200MHz - 825mV
300MHz - 850mV
400MHz - 900mV
500MHz - 900mV
600MHz - 900mV
700MHz - 925mV
800MHz - 950mV
900MHz - 1000mV
1000MHz - 1025mV
1100MHz - 1100mV
1200MHz - 1125mV
1300MHz - 1150mV
1400MHz - 1175mV
1500MHz - 1250mV
1600MHz - 1350mV
Experimental but stable battery saver:
Governor - ondemand
Range - 100MHz through 1.6GHz
100MHz - 775mV
200MHz - 775mV
300MHz - 800mV
400MHz - 800mV
500MHz - 825mV
600MHz - 850mV
700MHz - 875mV
800MHz - 900mV
900MHz - 950mV
1000MHz - 1000mV
1100MHz - 1100mV
1200MHz - 1125mV
1300MHz - 1150mV
1400MHz - 1175mV
1500MHz - 1225mV
1600MHz - 1350mV
GPU setup:
Low power state - 100MHz @ 800mV
High performance state - 400MHz @ 1050mV
Notes:
Custom governors were not stable for me AT ALL! I've found ondemand to be the best one for me and my needs, personally.
100MHz @ 750mV was so, SO close to being stable for me but my phone would routinely reboot in the screen off state. I'm assuming the stress of apps updating in the background, notifications etc was just too much.
As much as I love WidgetLocker (and I really do!), I found it to consume valuable resources, have more pronounced wake up lag and generally contribute to instability.
I use Chainfire3D to run my games etc. at x4 MSAA. As previously stated by Chainfire, the Mali can run at x4 with almost no extra overhead. I imagine that if one doesn't use x4 MSAA, one *might* be able to get away with 400MHz @ the stock 1000mV setting. That being said, I consider an extra 50mV to run at 133MHz faster to be a bargain.
Many games can be run with x16 MSAA with minimal overhead but I've found that for some resource intensive ones, especially multiplayer, they'll slow down unless the GPU is fed at 1200mV but this in turn causes a lot of heat generated so I would advise to avoid turning on x16 MSAA for those that you do find slowing down.
I use and recommend Voltage Control (donate version for extra features!) for setting up clock range and voltage for both the CPU and GPU. It also allows one to set boot settings (at setup or init.d script) and create multiple profiles. I do not recommend init.d script for untested settings as it could cause you issues.
Edit: Not everyone's kernels may support GPU OC/UV or the CPU ranges listed here. I am not responsible if you bork your device.
Here's someone else's method for testing settings:
Here's how I test UV settings.
Turn on everything. Wifi, bluetooth, max brightness, the whole works. This ensures the system is at maximum strain.
Start at maximum CPU clock
Lock the CPU clock (set the minimum and maximum allowed clock to the clock you are currently undervolting)
Lower the voltage by one step
Start a benchmark for a few minutes to see if undervolted clock is stable
If it passes, lower it again go back to step 4
When it freezes up your phone, reboot it and increase the voltage at that clock by two steps and consider it safe
Move to next frequency and go back to step 3.
You reached your lowest clock? Congrats, you should have a well undervolted CPU
Your voltages should always be lowering when your go from the highest clock to the lowest. If it happens that you have to increase the voltage at a lower clock, then also increase the higher clock frequency. I had a few hard locks because of this.
Example.
1000mAh (1GHz) > 900 mAh (900MHz) *< 950 mAh (800MHz) * > 700mAh (600mAh)
The 800MHz voltage is now higher than the 900MHz voltage. Also increase the 900MHz voltage to the same or higher voltage of the lower one.
1000mAh (1GHz) > 950 mAh (900MHz) > 950 mAh (800MHz) > 700mAh (600mAh)
Now that you have it undervolted, you may find that it could hardlock/reboot on you. When it happens do this:
Increase the voltage on all undervolted clocks by one step.
Continue using the device for a day
If the device locks up again, go back to back step 1
If its ok for a day, then every day lower the voltage back to what you had of only one clock (I suggest you go from highest to lowest)
You should be able to find which undervolt caused the reboot fairly quickly and still be able to normally use the phone and keep the rest of the "optimal" undervolts.
Click to expand...
Click to collapse
Sent from my GT-N7000 using xda premium
I don't think UV saves battery. It is display that sucks most of the juice.
You save less than 2% with extreme UV and after a single reboot caused by instability - you lose even more battery.
There's an excellent thread in Nexus S forums - "battery drain benchmarks" (please search it).
I had similar UV settings and my phone never crashed during benchmarks or stress tests.
But it always crashed while installing 100+ apps with app backup restore, restoring backups with TB or MBR, gaming.
After removing UV, it never crashed.
I haven't tested UV with ICS... would see and report if it really saves battery.
Boy124 said:
I don't think UV saves battery. It is display that sucks most of the juice.
You save less than 2% with extreme UV and after a single reboot caused by instability - you lose even more battery.
There's an excellent thread in Nexus S forums - "battery drain benchmarks" (please search it).
I had similar UV settings and my phone never crashed during benchmarks or stress tests.
But it always crashed while installing 100+ apps with app backup restore, restoring backups with TB or MBR, gaming.
After removing UV, it never crashed.
I haven't tested UV with ICS... would see and report if it really saves battery.
Click to expand...
Click to collapse
I'm not sure if you've read everything through carefully or you would have seen that I've covered several of your points.
You also would have seen the method I use for stress testing and would have noted that I aim for four things: speed/performance, stability, power management AND thermal regulation.
While I agree that the display, barring a wonky or misbehaving app, will almost always be the #1 battery drainer - power management will certainly help to conserve battery life.
You also would have seen I mention profiles. There may not be a one size fits all setting for everyone but one can most certainly set up profiles for different scenarios.. Such as TiB backups/restores.
Sent from my GT-N7000 using xda premium
Did you do some benchmarks at the highest speed several times to make sure you are getting extra performance? With this phone I noticed that while the phone wont crash.. .some times performance will drop when running at settings now fully correct.
Sent from my GT-N7000 using XDA
You covered a lot of points but UV is total waste of time.
You get nothing out of it.
http://forum.xda-developers.com/showthread.php?t=1478406
Could you please post your data, how much battery do you save after UV?
Disagree boy, cause with wakelock screen is off, there is significant battery drain, I went to 10 hours life on single charge, due to wakelock.
Normally with deepsleep about 2 days. That's a reduction of 87.5% with screen off. Cpu running @200mhz.
Do the same with undervolting will dramatically increase battery life in that situation. So overal it will be a fraction compared to using the device with screen on, but still significant.
Edit: guess I was wrong here
Sent from my GT-N7000 using Tapatalk 2
baz77 said:
Disagree boy, cause with wakelock screen is off, there is significant battery drain, I went to 10 hours life on single charge, due to wakelock.
Normally with deepsleep about 2 days. That's a reduction of 87.5% with screen off. Cpu running @200mhz.
Do the same with undervolting will dramatically increase battery life in that situation. So overal it will be a fraction compared to using the device with screen on, but still significant.
Sent from my GT-N7000 using Tapatalk 2
Click to expand...
Click to collapse
I actually did the test on Gingerbread.
I set min and max to 200 MHz, activated flight mode and had stock music player running for 3 hours - with undervolt and without undervolt.
To my surprise battery consumption was the same.
May be experts who know about our processor architecture can shed some light here.
Boy124 said:
You covered a lot of points but UV is total waste of time.
You get nothing out of it.
http://forum.xda-developers.com/showthread.php?t=1478406
Could you please post your data, how much battery do you save after UV?
Click to expand...
Click to collapse
I understand where you're coming from, boy.
I don't have data at the moment though I wish I did. But to be honest, it'd be scrambled anyway since whenever I'm not working or mission critical when I need proven stability, I'm testing out all different sorts of settings leading to lots and LOTS of reboots and such!
That being said, anecdotally, I have seen improved battery life for myself but maybe it's a placebo and I could be wrong about it - I have been before in the past. I do feel though that under my normal usage scenarios, I am experiencing less battery drain. It's difficult to quantify though exactly what this is due to since I experiment with kernels, voltages and frequencies.
But if all I'm getting is a 2% boost, man - I'll take it! Like any modder, whether it's min/maxing in a game, working on a car or whatever else, every little bit of a parameter squeezed out is something.
I also feel that you're too caught up on a single aspect, the battery life thing, to the detriment of my overarching holistic goal - efficiency.
Originally I started undervolting and experimenting with frequencies because of thermal output. I had wanted to experiment with x16 MSAA settings, which led to my GPU needing 400MHz and 1200mV which led to lots of heating up which led to me experimenting with everything I could.
Efficiency is what I want. The best performance at the best speeds at the best battery life at the best thermal regulation I can manage.
Now I'm looking at energy efficiency. I'm seeing suggestions that 100MHz may not be as efficient as 200MHz on our Exynos because the tradeoff in frequency power usage isn't worth the longer time spent completing tasks. I'm also seeing that in some situations, a performance best governor targeting max freq may be efficient because less time is spent completing a task and a quicker return to sleep.
I'm just sharing what I'm doing and hopefully others can benefit.
http://forum.xda-developers.com/showthread.php?t=1369817
Sent from my GT-N7000 using xda premium
Wow, thats illogical makes me wonder the math behind it.
Sent from my GT-N7000 using Tapatalk 2
While I appreciate the effort thrown into this, I humbly acknowledge the conclusion is incorrect.
When you lower Voltage slightly, without affecting stability, you pretty much put a toll on the processor for extra "wear and tear" and reduce its lifespan. However, this comes at the reward of reduced current.
So, it should be saving you battery. Underclocking it (safely) is also going to save you battery. And the same thing with different governors, like interactivX compared to regular ondemand, by finishing off processes quicker and reducing the frequency and voltage quicker, and going into Deep Sleep quicker.
I don't have the means to run a Scientific Experiment to prove these claims, nor the time to conduct them. But the majority of "hackers" synonymously agree it saves a noticeable power. These include themers, kernel developers and the casual user. I don't think an educated MAJORITY can be incorrect to the scale of this test's claims.

The best CPU governor for optimal battery saving

Guys, i need your opinions -
How do u think - what is the best CPU governor for maximum battery life ?
The possibilities are in the poll.
Keep in mind, that different cpu governors have different modes, so it will be advisable not only to vote for your favourite, but also to mention in the topic your prefered profile for the respective governor...
The mods for the poll are taken from arter97kernel...
that's how I use when I need that the battery lasts longer and gives me results.
sometimes even put the processor speed to 1000MHz .
It gets a little slow but it can handle
I forgot screen cpu but I have at 1400MHz and undervolt light
Hey ^^
well i prefer zzmoove with the zzmoove battery yank profile on Boeffla kernel 2.8 for Slimkat (check my Profile for more Infos: Link). It runs smooth, gives me more then enough Performance and also good Battery Saving. It also features a frequency lock when turning off the Screen and limits the max Freq to 600 mhz which is perfect if you want to run downloads or something similar without problems in the background and your screen is off, but also gives you good powersaving. Unfortunately if you hear Music it can stutter in zzmoove because of the Hotplugging functionality even when the screen is off. To solve this problem you have to turn off the Hotplugging Functionality at Screen off (i read that somewhere, but forgot where).
greetings kaz
gsms said:
Guys, i need your opinions -
How do u think - what is the best CPU governor for maximum battery life ?
The possibilities are in the poll.
Keep in mind, that different cpu governors have different modes, so it will be advisable not only to vote for your favourite, but also to mention in the topic your prefered profile for the respective governor...
The mods for the poll are taken from arter97kernel...
Click to expand...
Click to collapse
I use S5 Sensation in my I9300 and the battery life is the better I have in my phone. The governor used by default is pegasusq.

Is it safe to change my CPU governor to "performance" to force max clock speed?

Is it safe to change my CPU governor to "performance" to force max clock speed?
I don't want to overclock my Note 4, don't worry, not talking about that. But I am talking about MAX clocking it - forcing it to run at maximum rated speed. I've already tested out SetCPU and used it to change my governor to performance which forces the clock to max, and it nearly doubled my framerate in many games, especially the ones that struggled to play on this device like Xcom:EW.
But I quickly turned it off because I wasn't sure it was safe to do in the new era of smartphones, what with their DVFS and all. I'm worried that I'm going to overheat the CPU, and it's not going to be able to downclock because of temperature anymore. I'm only modifying the governor, but what if I actually used SetCPU to just change the CPU clock to max, without even touching the governor?
Can I hurt my phone by doing this? Can I safely start forcing my CPU to run faster while playing games, knowing that the only thing I risk is my battery draining faster, or am I actually risking damaging components by doing this?
Hello and thank you for using Q/A,
your CPU will not be damaged, but the battery life time will be shorted.
Regards
Trafalgar Square
RC
I personally have used Performance governor on Moto X 2013 for almost the whole 8 months I had it, 24x7 I mean. Never had a problem, yeah maybe battery life was little less than normal but I never did really care about it. Then I ran the same governor for a good period of time on my Note 3 too, same, no problem at all. Like you said, in games the frame rate difference is massive, but I don't play much games, I simply used that governor because it gets rid off all those micro lags and jitters which are Android's trademark, I simply can't them, with default Interactive the micro lags are very apparent.
However with Note 4 I am pretty happy with the BluActive governor, it makes most of the micro lags to go away, so sticking with it.
In any case unless you plan to use a mobile phone for maybe 5 years or so, I don't see any problem at all with it, other than a slightly increased heat, and maybe a little less battery backup, but you will find so many comments which might scare you, that chip burns off if you run it and all that, but those mainly are BS.

Categories

Resources