Major Power Regression in Linux 2.6.38 kernels - EVO 4G General

Because of this post and this article I did more digging and playing around with earlier kernels vs. 2.6.38
What I found out is that my phone's battery life almost doubles each time I flash older kernels, based on say, 2.6.37 such as Tiamat 3.3.7.
I've tried most of the most popular 2.6.38 based kernels out there, such as Tiamat, Savaged-Zen, TigerBlood, GoDmOdE - my battery lasts significantly less with all of them comparing to pre 2.6.38 kernels. I should mention I have not changed any settings, in between kernel trials, not a damn thing and I tested it without doing any overclocking.
Devs and Android users beware.

http://forum.xda-developers.com/showthread.php?t=695878

DomSim said:
http://forum.xda-developers.com/showthread.php?t=695878
Click to expand...
Click to collapse
this is not Q&A. This is important info for kernel and ROM devs to take into account, when they decide on direction of future developement endeavors

This is not development, you are not introducing a Rom or kernel, you are commenting on results from personal testing, essentially asking " why does this thing do what it do?"

NewZJ said:
This is not development, you are not introducing a Rom or kernel, you are commenting on results from personal testing, in essence asking " why does this thing do what it do?)
Click to expand...
Click to collapse
If the mods think it doesn't belong here they'll move it. Personally, I think this needs to have high visibility and needs to be sticky topic.

Moved to General, ill gladly move back when it has some factual data .(as in numerical, comparisons ect)

I'd have to agree with the OP. Im at work from 9-5 mon-fri and I do alot of kernel testing. The 3.3.7 tiamat and earlier savage zen kernels give way better battery life than the newer 2.6.38 based kernels. I also feel like my battery and performance is better using .37 kernels and its funny you brought this up!

For what it's worth, pretty much each revision of Tiamat (and possibly other kernels) are less and less undervolted, which may affect things to an extent. I'm not saying that's the sole source of decreased battery life, but it's a factor nonetheless.

that is true of sz. Blaim all the complaining people... and very picky evos.

jug6ernaut said:
Moved to General, ill gladly move back when it has some factual data .(as in numerical, comparisons ect)
Click to expand...
Click to collapse
Numeric comparisons between different kernels might upset the kernel devs, whose battery "benchmarks" in my testing are lower. I feel it wouldn't be fair thing to do
Also, it's pretty hard to provide objective numeric data, because battery drain heavily depends on the type of usage, and even though my day-to-day usage is similar, its not quite the same, so those numbers would not be accurate.
If it helps - I get ~17 hours of moderate usage with 2.6.37 or earlier kernels and I'm lucky if I can squeeze 9 hours with the same usage out of 2.6.38. And that info is more or less accurate for ALL 2.6.38 kernels across the board.
Consider this: if toastcfh, the rest of TeamWin gang, Cayniarb, Savaged-Zen crew and other devs miss this info because the thread was moved - it'll be our loss...

Aren't these kernels meant for 2.3.4?
midget tossing is habit...2010 midwest regional champion... hw 001

Apo11on said:
Numeric comparisons between different kernels might upset the kernel devs, whose battery "benchmarks" in my testing are lower. I feel it wouldn't be fair thing to do
Also, it's pretty hard to provide objective numeric data, because battery drain heavily depends on the type of usage, and even though my day-to-day usage is similar, its not quite the same, so those numbers would not be accurate.
If it helps - I get ~17 hours of moderate usage with 2.6.37 or earlier kernels and I'm lucky if I can squeeze 9 hours with the same usage out of 2.6.38. And that info is more or less accurate for ALL 2.6.38 kernels across the board.
Consider this: if toastcfh, the rest of TeamWin gang, Cayniarb, Savaged-Zen crew and other devs miss this info because the thread was moved - it'll be our loss...
Click to expand...
Click to collapse
so you know more about these kernels than these developers who rip them apart and rebuild them from scratch huh? and your gonna teach em all something right? sheesh...

Great thread, this is the type of stuff I like to know about. I am going to run a few short tests on Tiamat 4.0.2 and Tiamat 3.3.8 and see how noticeable the power consumption is between the two.

I call BS. To test properly.. you would need to wipe battery stats and calibrate between kernels. Then give it a couple days to settle. Also... there is no way to ever have the usage be the same if your actually using the phone. You would need to turn the radios off so that the service you are getting is equal. A whole bunch of other things too.. so basically you would need to not use it as a phone to test the kernel right.
Oh yeah, and also use the same under volt.
Sent from my PC36100 using XDA Premium App

I ran a 5 minute idle test with Tiamat 3.3.8 and 4.0.2 with all radios turn on and governor set to performance. The power consumption was logged by PowerTutor.
I parsed each log to pull the total-power measurements and then averaged the middle 25% of the measurements to ensure they were from the phone's true idle state.
Tiamat 4.0.2 - 52mW
Tiamat 3.3.8 - 53mW
So essentially they are almost identical. Of course the 4.0.2 does have lower voltages in HAVS and the radios were on, so there could be room for improvement on these measurements.
Bottom line....even if 2.6.38 uses more power, you shouldn't even notice a difference between the two, due to HAVS voltage differences between these two kernels.

Redid the test with all radios turned off. Essentially these are CPU only readings. The power difference still seems rather moot. It could be a different story under intense CPU utilization though if the percentage difference scaled. I'll leave that to someone else though. Also, remember that these numbers are in no way entirely accurate...there is going to be a +/- deviation in any power consumption measurements unless it is done in a truly controlled environment.
Tiamat 4.0.2 - 5mW
Tiamat 3.3.8 - 2mW

aimbdd said:
I call BS. To test properly.. you would need to wipe battery stats and calibrate between kernels. Then give it a couple days to settle. Also... there is no way to ever have the usage be the same if your actually using the phone. You would need to turn the radios off so that the service you are getting is equal. A whole bunch of other things too.. so basically you would need to not use it as a phone to test the kernel right.
Oh yeah, and also use the same under volt.
Sent from my PC36100 using XDA Premium App
Click to expand...
Click to collapse
I did recalibrate battery for each new kernel...I put a couple of weeks of hard work into this and I started well before I saw that phoronix article because I knew something is up. I'm sorry if this is not good enough "proper testing" for you. If you chose to be ignorant and pretend the issue doesn't exist - its your decision of course, but many others would want to know.
As per your advice to turn off the radios - what if the power hog regression in 2.6.38 is related to one of them?

The power regression was solved by phoronix last week.
http://phoronix.com/scan.php?page=article&item=linux_2638_aspm&num=1
Basically, active state power management isn't being used properly. There needs to be modifications based around the parameter "pcie_aspm=force". An example in a different kernel git is below:
http://android-x86.tarot.com.tw/?p=...ff;h=d6d385743463f38a0da899cd4607e526ad9a049f
With this fix on my laptop (applied to GRUB), I saw an increase in battery life of 27%.

LifeInTheGrey said:
The power regression was solved by phoronix last week.
http://phoronix.com/scan.php?page=article&item=linux_2638_aspm&num=1
Basically, active state power management isn't being used properly. There needs to be modifications based around the parameter "pcie_aspm=force". An example in a different kernel git is below:
http://android-x86.tarot.com.tw/?p=...ff;h=d6d385743463f38a0da899cd4607e526ad9a049f
With this fix on my laptop (applied to GRUB), I saw an increase in battery life of 27%.
Click to expand...
Click to collapse
So in other words, the OP was right even though people doubted him?
Not surprised.
Hopefully the devs are aware of this and can incorporate it into their kernels.

mattykinsx said:
So in other words, the OP was right even though people doubted him?
Not surprised.
Hopefully the devs are aware of this and can incorporate it into their kernels.
Click to expand...
Click to collapse
+1
Bump
10char

Related

[Q] [Poll] Post your favorite ROM/Kernel combination for the Epic 4G!

I have to admit, I have selfish reasons for posting this poll. I never know what is the most stable/feature rich ROM available without reading through countless posts on several forums so I thought it would be good to have a central place for people to get an idea on what everyone else is currently using.
Right now I'm using Epic Experience with Phoenix kernel and my screen goes black with no way to bring it back without a battery pull. So I need to know what ROM to try next. My goals are simple:
Fast, stable, and most importantly... good battery life.
I just finally convinced my wife to let me root her epic in order to fix battery life and its not working very well.
I am using Baked Snack v1.3 and was on the Boss Rom before with Mixup Kernel. On BakedSnackShack.com you can download it for free. Herver was kicked off for asking for donations ($5min) before the public release, but let me tell you I have not run into 1 problem in ANY program in the last month...I don't plan on ever going to another developer because of his great success and features. He should be allowed back on but every time he does more drama happens. A rom is different for everyone and thats why there are so many developers to suite your needs. Good luck and if nothing else just try it out....I assure you will be impressed!
you need to realize what works for one person will not work for another. different combanations work and suite others needs then their own. i suggest trying everything out and find what works the best for you
I have to say, I agree with mysteryemotionz. What one person may prefer may be what another dislikes. What one person may notice as being a problem, another may overlook. For instance, on the EVO forums a lot of people like cyanogenmod. I personally do not due to the lack of certain functions like 4G and other things (havent look at their forum in awhile so IDk if its still not working now).
Im currently using Epic Experience and mixup kernel. I loved AOSP Magic, but Epic Experience seems to have better support. I prefer the look of AOSP magic tho. I havent tried any other kernel since Im currently satisfied with how I have mixup running.
Yup...baked snack is the best so far I'd say. I've tried em all. Still haven't gotten the battery life that I've read about in some posts. But the battery life is by far better than stock. I'm using epic experience 1.2.0.4 with the latest mixup kernel also tried the latest phoenix kernel with no significant difference. So I just keep trying new things till something better comes along.
Sent from my spot on the earth
Seems clear I've missed a few ROM's like "baked snack". I just used the wiki post for my poll info.
jeremyglass said:
I am using Baked Snack v1.3 and was on the Boss Rom before with Mixup Kernel. On BakedSnackShack.com you can download it for free. Herver was kicked off for asking for donations ($5min) before the public release, but let me tell you I have not run into 1 problem in ANY program in the last month...I don't plan on ever going to another developer because of his great success and features. He should be allowed back on but every time he does more drama happens. A rom is different for everyone and thats why there are so many developers to suite your needs. Good luck and if nothing else just try it out....I assure you will be impressed!
Click to expand...
Click to collapse
At first you say that you're using Baked Snack v1.3 but then later on you mention Herver and in the same sentence say "...I have not run in to 1 problem in ANY program in the last month".
So which is it? What ROM are you using that isn't giving you any grief right now?
Baked Snack v1.3 or Herver?
Also, while I'm on the subject, how do I add more options to the poll?
I think Herver is the developer of bakedsnack
Sent with my 4G interwebs.
This post is genius.
decalex said:
I think Herver is the developer of bakedsnack
Sent with my 4G interwebs.
Click to expand...
Click to collapse
Yep hero is the developer of baked snack...I am using the wisdom rom...evo themed with baked snack kernel oc to 1.3 ghz and it runs very smooth..no problems whatsoever
Sent from my SPH-D700 using Tapatalk
mrkrabs said:
Yep hero is the developer of baked snack...I am using the wisdom rom...evo themed with baked snack kernel oc to 1.3 ghz and it runs very smooth..no problems whatsoever
Sent from my SPH-D700 using Tapatalk
Click to expand...
Click to collapse
Win.
Sent from my SPH-D700 using XDA App
I don't think there's a "best" ROM / kernel combo for everyone.
It seems that each individual's phones and experiences can and are different.
For example, I was running stock rooted ROM with every version of MixUp since it's inception. On the most recent MixUp I started having battery life issues with no other changes so I decided to try a few things. First thing I did was change to the Epic Experience ROM which did improve my battery life a bit with the latest MixUp back to where it had been with the stock ROM and the previous MixUp. After reading a lot that the Phoenix kernel worked really good with Epic Experience I changed to the most recent version of it. Battery life went off the charts in the good direction with no other change. Phoenix is suppose to run at interactive governor by default but I was getting some weird behavior from it, so I swapped it to conservative and it's working perfect for me. When using any of the kernels I always kept my minimum set to 200Mhz to avoid any sleep/lockup/wakeup issues.
I don't think Phoenix is a better kernel than MixUp, nor do I think MixUp is, just for right now, Phoenix is working better for MY phone. I'll probably continue to bounce back and forth between these two as newer versions of each are released.
The below screen shot was after some light usage, a few phone calls, little bit of texting, maybe about 1/3 to 1/2 of the time on Wi-fi, maybe 30min to 1hr of games, maybe an hour of web surfing.
Baked Snack 1.3 with Phoenix 1.49.
I got about 13.5 hours of battery yesterday.
I used the baked snack kernel for a few days with noobl's andromeda (de odex) rom.. With many apps I didn't need, removed. I had many lockups (especially while charging) where a battery pull was the only way to get out. The only consistent crash I could find was if I was plugged into ac, with screen off, and received a notification, e.g. gmail message.. Which meant for me, no wake up alarm in the morning. I liked the speed, but the instability, on my setup at least, had me switch back to mixed up kernel. I still don't have the best battery life, but I'm getting to the bottom of that in the mixed up thread. Fwiw.
Sent with my 4G interwebs.
by far my favorite and my current set up: apophis 1.1 (rom) and baked snack 1.3 (kernel), overclocked to 1.3 ghz.
i get linpak scores of 10.5 + consistantly and quadrant scores of 1010 + consistantly. this is the fastest combination ive ever used. when i first unplug it, the battery usually drops down to 95% in a couple minutes but steadies out after that. by the end of the day, once i get home from school, i usally have at least 20% left. and that's with 4g always on and moderate to heavy usage. once plugged in, it charges at a steady and fast rate.
i have my phone stripped down to bare minimum and only a few apps that i need. ive had zero black screens and few to none force closes upon rebooting but none to freak out about or anything that proves a problem.
props to the developers. thank you for providing the community with your hard work.
edit: battery info
Sent through the Matrix.
wtf OP? did you have to make a poll?
I had the same problems as the poster, but i think it was setcpu, i kept messing with the settings. I ended up leaving it on conservative, and it has not turned off on me since.
Im running phonix kernal / epic experience
I get like 20 maybe more hours of usage, fast charges, and only drains very little battery overnight..
i like this setup
I personally use Epic Experience 1.2.0.4 and Phoenix 1.49. I switched the governor to conservative because I had issues with the default interactive and have had zero issues since I switched. I got amazing battery life just 1ghz but I underclocked to 800mhz to improve my battery life even more without compromising on performance. If I'm doing something processor intensive I do bring it back up to 1ghz or even to 1.2ghz haha. But this combo works for me and I got approximately 35hrs with what I would call light usage. With moderate to heavy usage I average about 18hrs.
It definitely seems that Epic Experience / Phoenix is the most popular combo, at least with those voting.
Glad to see others have reported similar experiences to mine with this combo as well.
Could just be mental, but I also seemed to notice that the phone charges a lot quicker than before with this Rom/Kernel as well as not getting as hot while charging. Not sure how/if it could effect this, but it sure "seems" that way.
wjsmaggle said:
I personally use Epic Experience 1.2.0.4 and Phoenix 1.49. I switched the governor to conservative because I had issues with the default interactive and have had zero issues since I switched. I got amazing battery life just 1ghz but I underclocked to 800mhz to improve my battery life even more without compromising on performance. If I'm doing something processor intensive I do bring it back up to 1ghz or even to 1.2ghz haha. But this combo works for me and I got approximately 35hrs with what I would call light usage. With moderate to heavy usage I average about 18hrs.
Click to expand...
Click to collapse
Maybe that was my problem. I'm now running Baked Snack 1.3 and the kernel says Here_Over. The volume won't work and there are other issues. My wife is gunna kill me if I don't get her a working phone soon. She's going to ask me to take her back to stock. (Ugh)
When following the instructions for flashing Epic Experience I never saw anything about setcpu or a governor. I just left everything at defaults.

So what kernel do you tend to have better battery life with?

I've always had problem with the CM kernel, so I switched to wildmonk's and I have to say I've had some improvement, even though I believe its more voltage.
How is everyone else's battery on which kernel? Just curious on which ones I can squeeze the most out of.
oh geez...
I'm using the kernel that comes with CM, but I've got SetCPU installed and set to lower the CPU speed when the screen is off. Seems to make a world of difference.
If you don't already have SetCPU, I highly recommend purchasing it.
The stock kernel of CM 6.1 is pretty good itself. The Wildmonks ones are excellent though. However, you will need to use applications for root users like:
SetCPU - To configure CPU speed. Set the profiles properly. Otherwise, it's no good
Spare Parts - To determine which application(s) use more power, CPU and Data
DroidWall - To configure which apps are allowed to access the internet and when
Apart from these you may use various rendering effects found in CyanogenMod Settings for reducing the brightness of the screen
Please keep in mind there are a lot of other factors that affect battery life rather than just the kernel. Usage and configuration are two HUGE factors that a lot of people seem to forget about. Battery usage will also vary between ROM builds, software installs, radio firmware versions, network availability (weak vs strong signals), and even minute hardware differences.
Just because something works for one person doesn't mean it will work for everyone. Your mileage WILL vary.
Personally, I've gotten good results with WMs kernels, Kangorama ROMs, the 5.08 radio, and a couple of software tweaks (SetCPU, JuiceDefender + JuicePlotter to view usage trends, and Tasker with Brightness/Wifi/Location profiles).
codesplice said:
Please keep in mind there are a lot of other factors that affect battery life rather than just the kernel. Usage and configuration are two HUGE factors that a lot of people seem to forget about. Battery usage will also vary between ROM builds, software installs, radio firmware versions, network availability (weak vs strong signals), and even minute hardware differences.
Just because something works for one person doesn't mean it will work for everyone. Your mileage WILL vary.
Click to expand...
Click to collapse
That may be true, but that doesn't mean that the kernel can't have a huge impact on the battery life, especially if the kernel lacks optimization. There are only so many things that the kernel itself can do to improve battery life, but there are tons of things it can do to worsen battery life. Personally I have been pretty prudent on managing my phone's battery through various apps and tweaks, and was really disappointed with the stock CM 6.1.1 kernel, as it pretty much seemed like it was ignoring or overpowering all my tweaks.
After only about two hours of sitting idle with the screen off, I could lose as much as 30% of my battery, whereas with older ROMs I could have my phone unplugged for about 10 hours with minimal use, and only lose about 15% battery. I was running the exact same apps and settings as in CM 6.0, and I had been sitting at home in the same room the whole time, without any noticeable changes in network signals. Switching to Wildmonk's kernel immediately brought my battery life back to normal, and I doubt that it was merely coincidence, especially after flashing to a nandroid backup of (mostly) stock CM 6.1.1 made the battery problems come back.
Based on my recent experience, I wouldn't rule out the possibility that the kernel can make noticeable improvements. If you can rule out network problems, display settings, and most of your software tweaks, then the kernel is the only thing left to check.
earlyberd said:
Based on my recent experience, I wouldn't rule out the possibility that the kernel can make noticeable improvements. If you can rule out network problems, display settings, and most of your software tweaks, then the kernel is the only thing left to check.
Click to expand...
Click to collapse
Absolutely, but the kernel is not the ONLY variable that affects battery drain. A lot of people are quick to blame the kernel, and then are shocked to find that they have crappy battery life on all kernels because they haven't changed their ROM/radio/settings/usage/configuration.
wildmonk kernel has been great for me. i tried a bunch of them..but i settled on
2.6.35.4-925mv-cfs-svs-aix-720p-monks-082710
the recent ones tend to drain the battery when wifi is enabled. the android system process seems to be in use. normally it doesnt hit above 10%, but with the latest, it goes to 40%

Battery Life Testing - Kernel + Rom Combo's

Iv always considered battery life a very important factor in my purchasing of a phone and was one of the big reasons i moved from my Desire HD to the Nexus S (amongst many others). However having been on the android scene for a while im used to tinkering with kernels and roms etc and due to the new battery graph in gingerbread im now able to make direct comparisons rom by rom and kernel by kernel to see how they all differ. This is all shown + explained below.
First id like to explain my method;
I install the rom and kernel with the same apps I always use and charge the phone fully until i am about to go to sleep. I then unplug the phone and leave it in the exactly the same place on my desk with exactly the same settings (sync intervals etc) and i kill all apps before i leave it just to make sure the test is fair.
In the morning i then take a screen shot of the battery life over the period of usually 8 hours and will upload this for you to see. I then look at the over-night battery drain as a comparison on how good battery life is on specific roms and kernels.
Justification - Many of you may be thinking who cares about standby battery life? I care about the battery when im using it!?
Well although that's valid there are many reasons i have taken this approach;
1. It is completely fair i.e. no individual differences such as 'im a heavy user' but we all have different definitions of a 'heavy user'. This is eliminated by the fact the phone is not used and thus the same every time (no individual differences)
2. If a phone has severe battery drain during sleep general knowledge should dictate that battery drain should increase when the screen is on, cpu is working and your syncing your facebook wall, thus a direct correlation with sleep battery drain and usage drain should be apparent.
Therefore i have decided over the next few days/weeks etc i will upload on a regular basis a screenshot of the battery drain over night, along with the battery use details and screen on time etc etc. In addition to the rom and kernel info and i hope that over time this proves to be a very useful thread.
I have been doing this over the last few days on a less strict system and there have been some VERY interesting results which i will upload over the next few days.
(Just a final note iv never really done anything like this on xda before so if its a stupid idea then please tell me now so i dont waste my time doing it, however if people are really interested please let me know and il make sure i do my best!!)
Thanks for reading my extremely long post!
Bratfink
SCREENSHOT 1: This was using modacos custom rom r10 and the kernel that comes with it, i had quite a short sleep (about 6 hours) but the curve pretty much shows what the point im trying to make is. The battery life is great!!
SCREENSHOT 2: This is CM7 Nightly build 27# with Netarchy's latest 1.2.1 kernel and although i didnt use the screen once of charge for a few minutes as you can see once the screen was off battery life barely decreased and was over a long period of time.
If anyone wants me to test specific combos please let me know.
I'm actually looking forward to reading your results.
Sent from my Nexus S using XDA App
Yep, this would actually be pretty useful. Good luck and good work!
i'm a sucker for these types of tests, always love doing them and seeing the results. post soon!
I like others will be keeping an interested eye on your findings.
Uploading pictures of modaco r10 with it's own original kernel today. Its an interesting result!
Sent from my Nexus S using XDA App
Uploaded to OP
I'm interested too.
Can u try modaco r10 with Netarchy 1.2.1 kernel.
bratfink said:
Iv always considered battery life a very important factor in my purchasing of a phone and was one of the big reasons i moved from my Desire HD to the Nexus S (amongst many others). However having been on the android scene for a while im used to tinkering with kernels and roms etc and due to the new battery graph in gingerbread im now able to make direct comparisons rom by rom and kernel by kernel to see how they all differ. This is all shown + explained below.
First id like to explain my method;
I install the rom and kernel with the same apps I always use and charge the phone fully until i am about to go to sleep. I then unplug the phone and leave it in the exactly the same place on my desk with exactly the same settings (sync intervals etc) and i kill all apps before i leave it just to make sure the test is fair.
In the morning i then take a screen shot of the battery life over the period of usually 8 hours and will upload this for you to see. I then look at the over-night battery drain as a comparison on how good battery life is on specific roms and kernels.
Justification - Many of you may be thinking who cares about standby battery life? I care about the battery when im using it!?
Well although that's valid there are many reasons i have taken this approach;
1. It is completely fair i.e. no individual differences such as 'im a heavy user' but we all have different definitions of a 'heavy user'. This is eliminated by the fact the phone is not used and thus the same every time (no individual differences)
2. If a phone has severe battery drain during sleep general knowledge should dictate that battery drain should increase when the screen is on, cpu is working and your syncing your facebook wall, thus a direct correlation with sleep battery drain and usage drain should be apparent.
Therefore i have decided over the next few days/weeks etc i will upload on a regular basis a screenshot of the battery drain over night, along with the battery use details and screen on time etc etc. In addition to the rom and kernel info and i hope that over time this proves to be a very useful thread.
I have been doing this over the last few days on a less strict system and there have been some VERY interesting results which i will upload over the next few days.
(Just a final note iv never really done anything like this on xda before so if its a stupid idea then please tell me now so i dont waste my time doing it, however if people are really interested please let me know and il make sure i do my best!!)
Thanks for reading my extremely long post!
Bratfink
SCREENSHOT 1: This was using modacos custom rom r10 and the kernel that comes with it, i had quite a short sleep (about 6 hours) but the curve pretty much shows what the point im trying to make is. The battery life is great!!
If anyone wants me to test specific combos please let me know.
Click to expand...
Click to collapse
Sent from my Dell Streak using XDA App
I'd be interested in seeing CM7, just whatever nightly build you manage to catch.
I don't think that result will give you precise result.
I found that statistic is wrong sometimes.
It is better to measure using "play video loop".
So, find a movie, set to the same brightness and play in loop?
gogol said:
I don't think that result will give you precise result.
I found that statistic is wrong sometimes.
It is better to measure using "play video loop".
So, find a movie, set to the same brightness and play in loop?
Click to expand...
Click to collapse
Unfortunately that will not give you how well the rom does while it is idle.
I'm really looking forward to your results. My only issue with this phone is the battery life is terrible. I ordered an extended battery off Ebay 3 weeks ago and have yet to receive it. Never ordering anything from Hong Kong again!
Fuse8499 said:
I'm really looking forward to your results. My only issue with this phone is the battery life is terrible. I ordered an extended battery off Ebay 3 weeks ago and have yet to receive it. Never ordering anything from Hong Kong again!
Click to expand...
Click to collapse
What were you using before that you consider this phone to have bad battery life? By far this phone has the best battery life of any Android device I've owned. If you want bad battery life, get an Evo. That is horrific battery life.
bratfink said:
Uploading pictures of modaco r10 with it's own original kernel today. Its an interesting result!
Sent from my Nexus S using XDA App
Uploaded to OP
Click to expand...
Click to collapse
I am looking forward to the rest of these, however just a quick question. What was your battery use like in that situation with the stock ROM? Just for reference, because I'm thinking I might switch to modaco soon. Thanks!
More soon
Iv got a few more screenies but it seems i may have to end this thread prematurely. Its the SNS or 2.3 but idle battery usage for me is pretty much identical no matter what the kernel or rom (il upload a CM7 + Netarchy 1.2.1 shot top op). This was unexpected on my behalf because with my DHD kernels and roms had a big impact on idle battery consumption. I have noticed however that without a full wipe when changing some kernels such as with M r10 idle battery life can increase drastically. I will continue for the next few days and if i dont see any big changes i may move method to a video loop or something of the sort.
Interesting find
I found out something today that may interest may of you. I was messing around with the sleep profiles on SetCpu with netarchys kernel and i found that its better to leave the range from 100-800 than 100-400 or less. This i believe is because when the phone is asleep and carries out a process it will be forced to clock to the maximum of 400mhz and then stay at that speed for 10 seconds, instead of say going up to 800mhz for 1-2 seconds to complete the same process. So a tip to you would be set your profile to 100-800mhz conservative and see what you get!
bratfink said:
I found out something today that may interest may of you. I was messing around with the sleep profiles on SetCpu with netarchys kernel and i found that its better to leave the range from 100-800 than 100-400 or less. This i believe is because when the phone is asleep and carries out a process it will be forced to clock to the maximum of 400mhz and then stay at that speed for 10 seconds, instead of say going up to 800mhz for 1-2 seconds to complete the same process. So a tip to you would be set your profile to 100-800mhz conservative and see what you get!
Click to expand...
Click to collapse
This kinda enforces the point that SetCPU is useless for the NS unless you really want to overclock...
Any battery saving tactics used buy lowering screen off profiles are counteracted by the system process;( So i think in the context of battery life you are probably better to not use SetCPU
JD
bratfink said:
Iv got a few more screenies but it seems i may have to end this thread prematurely. Its the SNS or 2.3 but idle battery usage for me is pretty much identical no matter what the kernel or rom (il upload a CM7 + Netarchy 1.2.1 shot top op). This was unexpected on my behalf because with my DHD kernels and roms had a big impact on idle battery consumption. I have noticed however that without a full wipe when changing some kernels such as with M r10 idle battery life can increase drastically. I will continue for the next few days and if i dont see any big changes i may move method to a video loop or something of the sort.
Click to expand...
Click to collapse
I would definitely move to a video loop or some other 'heavy use' task because when you think about it, not too much can really be changed when almost no power is being used in the first place.
You could also just mention your general feeling for the battery life of each rom, because quite often that is what I go by as my usage varies depending on a range of things (school day or weekend etc) .
Sent from my Nexus S using XDA App
Could you please test modaco r10 with Netarchy's latest 1.2.1 kernel ?
I would love to see the difference between modaco's default kernel vs netarchy's kernel.
On CM nightly 13, I just went 11 hours and I'm at 90%. That's 1/2% an hour. Not bad. That is with stock kernel.
Sent from my Nexus S using Tapatalk

Using CM7? You must read this.

Please delete this thread, wrong account was used to make the post.
raberry said:
Ok guys. I have to admit, I was about to give up on CM7 due to battery issues and what seemed to be memory leaks progressively making my phone lag more and more throughout the day. I have found something some of you may not be aware of and I wanted to share this amazingly helpful kernel with you.
Using x99, Scary, and stock kernels regardless of governor I seemed to maybe get 8 hours of battery life with moderate to heavy use. I am proud to say, using bcnice20's kernel I have had my phone on for 9 hours with 60% battery life today. This is incredible, even with lighter use on the other kernels I had NOWHERE near this amount of battery left.
Most of you may not have seen this kernel because it isn't posted. (as far as im aware) Let me help you out. First, go into your Market and download Kernel Manager. There is a lite version in there that is free. If you have an extra buck or two donate to the developer for making a neat app where you can download kernels in a convenient place by downloading the pro version.
What you will be downloading is called bcblend SIO tinyRCU kernel. I have not true VR tinyRCU kernel yet. Not sure what the differences are, if someone could elaborate that'd be appreciated.
Regardless, I find the sweet spot for this to be at 245MHz min and 1GHz max on Interactive governor to work really well. The phone does not seem to get laggier throughout the day and consistently performs well, snappy, and just way better than the other kernels. (No offense to x99 and Scary, I appreciate all you've done!)
Give it a try and let me know if you all have similar results! I have heard that AOSP gets less battery than Sense ROMs from people on this very forum, and this may be the kernel to put that little debate to rest.
Oh, and be sure you have the latest CM7. I am on 7.0.2.1 trying this out.
Click to expand...
Click to collapse
I have been using the x99 kernal and i have about 1 d and 12 hrs with about 45 % left.. I also have the calibrator apk from the market so it wipes the battery stat when im at 100%. and various other apps that deal with the speed of the sd card and what nots. So, me coming from the Hero is very happy with this phone...
Thank you for sharing your findings... THIS IS SO AWESOME !!!!!
That's amazing battery life. I can't seem to get that great of battery life on CM7 as compaired to stock sense roms. What calibration app are you talking about exactly? There are a few on the market.
Sent from my PG06100 using XDA App

Oreo 8.1 Custom Kernel Testing for Oneplus 3t, Battery Friendly Tweaks and more

Anyone else curious? I am doing tests.
I'm on NitrogenOS, and use my own tweaks.
This test is not to make any kernel look bad.
These are the kernels I will test. (See Screenshot)
Plan to post results at end of entire testing, not per kernel.
Note: I have more than one Mady kernel listed, but I will only be testing latest. All test will be done using latest version of that kernel compatible with Latest NitrogenOS (4/13 version, until I complete my testing).
Edit : Usage will be mostly consistent. Plan on playing Injustice bout 30mins /day, otherwise just checking email, checking updates, playstore etc. Will tell if I do something different that day that I don't usually do.
Thanks for any interests. I am primarily doing these tests for fun & my benefit. Just thought others might like to see MY results.
You can all these kernels and A LOT more from this link: Thanks to @sidd007
https://forum.xda-developers.com/oneplus-3t/how-to/info-list-oreo-roms-kernels-available-3-t3688139
Thanks to @sidd007 for helping me figure out how to get this thread up and running and for letting me take up space on his incredible thread
Thanks to @mady51 for his kernel. I vow to make these tests unbiased, but if it was not for his tremendous kernel, I would have to charge my phone a lot more often
Thanks to @flar2 because, well, because he's flar2
@osmosis To me, he is one of the greatest gifts to Android
Tweaks I will be using :
GFknQz TWeaKz
Depending on kernel, a lot of these settings may be set by default, with no need to adjust.
Set all on boot:
Underclock cpuz
Little Cluster - 1132
Big Cluster - 1056
Both min - 307
Both governors - Interactive (alternative alucard)
Disable Cpu TouchBoost
Disable msm_thermal
Enable Core Control
Do not allow work queues
Gpu:
Max frequency - 315
Min frequency - 133
Initial - 214
Powersave governor
Backlight Dimmer Enabled
High Brightness Disabled
Adaptive Memory Killer Disabled
ZRAM Disabled
Dirty Ratio 20
Dirty Background Ratio 5
Min Free kbytes 147456
VFS Cache Pressure 100
Overcommit Ratio 50
I/O - cfq (alternative maple)
Readahead 128
TCP Congestion - cubic (alternative westwood)
Block Wakelocks:
Wlan_wow_wl
Wlan_pno_wl
Wlan_ipa
Wlan_extscan_wl
Wlan
Sensor_ind
Qcom_rx_wakelock
Netmgr_wl
IPA_WS
[timerfd]
NETLINK
Disable gentle fair sleepers
I have added my kernel settings above. They may be adjusted, modified etc varying by kernel and/or rom I am using at the time, but generally they stay the same. My settings are ALWAYS the tweaks I use to get the best battery for my usage, BUT I will not tolerate lag or poor performance, and if I experience such things, I will update my posts and experiences. These settings work great for me, your usage surely differs, and therefore more than likely, your battery life and performance may differ as well. If my settings cause issues either don't use them, adjust the proper settings according to your issues, or post about it here, and I will TRY to find an answer or solution.
Thanks everyone
Anyone else curious? I'm offering to do a test. Semi-scientific. If I get enough likes here, I'll do it. If not, I'm not gonna waste my time.
I'm on NitrogenOS, and use my own tweaks.
My test would be these custom kernels. My tweaks are based on battery life more than performance. That being said, performance must be fluid, not noticeably lagging. Plan to charge phone to 100%, flash kernel, reboot, and let run for 24hrs discharging. I just want to see results using exkm battery monitor, and device battery use(phone idle, android system, etc). Edit : May include some bbs or accubatty stats
This test is not to make any kernel look bad. My tweaks, as mentioned, are to test kernels based on my battery saving settings. That does not mean that a kernel tested cannot test relatively poorly on my test, and not be the best possible kernel for performance, stability, game play, etc. You can do your own test if that's your preference.
These are the kernels I will test.
First test will be NitrogenOS stock kernel.
Plan to post results at end of entire testing, not per kernel.
Note: I have more than one Mady kernel listed, but I will only be testing latest. All test will be done using latest version of that kernel compatible with Latest NitrogenOS (4/13 version, until I complete my testing).
Edit : Usage will be mostly consistent. Plan on playing Injustice bout 30mins /day, otherwise just checking email, checking updates, playstore etc. Will tell if I do something different that day that I don't usually do.
Thanks for any interests. I am primarily doing these tests for fun & my benefit. Just thought others might like to see MY results.
Sent from my OnePlus3T using XDA Labs
Well to make it a truly scientific test, I would need about, let's think, about 40 different roms, a dozen different kernels, stock oxygen os with stock kernel, that many roms on custom kernels no tweaks, that many roms on my settings, plus each rom on each custom kernel, so that's somewhere in the neighborhood of 16 billion oneplus-3 or 3t phones. My math may be off. More confused ??
I'm just testing NitrogenOS, the rom I use, on half a dozen different custom kernels. This experiment may not help anybody but me. But I can tell you without any doubt that kernels and settings do make a difference. I'm struggling to make 24 hrs on Nitrogen stock kernels with my tweaks. I started yesterday at 4pm, where I live, and just woke up at 7am and only have 49% battery, after minimal use yesterday. And this is usually the time I'm busiest on my 3t. So to think I'll make it 9 more hours of moderate usage is pretty hopeful. Maybe my setup merely conflicts with stock Nos kernel. That's a possibility. And that's a part of what I'm trying to determine. After I test the kernels on my tweaks, I'm going to test best 3 or 4 on devs stock settings. Maybe I am testing in reverse order of what I should. Regardless, I'm spending over a week to run the test. All kinds of factors go into usage, and all usage is different. If anybody else wants to test their way with their usage, I encourage it. We've all probably tried custom kernels on our roms at least a little. And found one that works for us. Slightly or vastly better than another. Im just doing it my way, and making my findings public here. People are welcome to ignore my results or take them with a grain of salt. That's entirely up to readers.
Sent from my OnePlus3T using XDA Labs
I've revamped the way I'm doing test. For 2 reasons. My test started planning to use nos 4/13. But 4/15 dev made tons of changes to stock kernel again. I've been on Nitrogen long enough, and read every changelog, to know that @xyyx generally only updates kernel when it's important. So, in fairness, I'm using Nitrogen 4/15 now. Secondly, I decided, also in fairness, to use stock kernel settings per kernel. Meaning flash & forget. No tweaks on my end. To get more accurate results for the kernels themselves. Then I will take the best 3 or 4 and apply my tweaks and test. Once again, I am doing this voluntarily, and mostly for myself, but I will gladly share results. Each kernel, I will test 4pm(where I live) to 3pm the next day. I realized I need an hour to recharge . I'm about 5hrs away from my first actual test completion for Boeffla's. I'll post result here.
Sent from my OnePlus3T using XDA Labs
Tweaks I will be using :
GFknQz TWeaKz
Depending on kernel, a lot of these settings may be set by default, with no need to adjust.
Set all on boot:
Underclock cpuz
Little Cluster - 1132
Big Cluster - 1056
Both min - 307
Both governors - Interactive (alternative alucard)
Disable Cpu TouchBoost
Disable msm_thermal
Enable Core Control
Do not allow work queues
Gpu:
Max frequency - 315
Min frequency - 133
Initial - 214
Powersave governor
Backlight Dimmer Enabled
High Brightness Disabled
Adaptive Memory Killer Disabled
ZRAM Disabled
Dirty Ratio 20
Dirty Background Ratio 5
Min Free kbytes 147456
VFS Cache Pressure 100
Overcommit Ratio 50
I/O - cfq (alternative maple)
Readahead 128
TCP Congestion - cubic (alternative westwood)
Block Wakelocks:
Wlan_wow_wl
Wlan_pno_wl
Wlan_ipa
Wlan_extscan_wl
Wlan
Sensor_ind
Qcom_rx_wakelock
Netmgr_wl
IPA_WS
[timerfd]
NETLINK
Disable gentle fair sleepers
I have added my kernel settings above. They may be adjusted, modified etc varying by kernel and/or rom I am using at the time, but generally they stay the same. My settings are ALWAYS the tweaks I use to get the best battery for my usage, BUT I will not tolerate lag or poor performance, and if I experience such things, I will update my posts and experiences. These settings work great for me, your usage surely differs, and therefore more than likely, your battery life and performance may differ as well. If my settings cause issues either don't use them, adjust the proper settings according to your issues, or post about it here, and I will TRY to find an answer or solution.
Thanks everyone
I'm charging my phone now. Test is complete for Boeffla kernel. And I have to say results are impressive. I'll post them after charge is complete
Boeffla's Latest for custom roms. Much better than my last Boeffla test on 8.1 custom rom.
Overall really good results. I was worried for a while. I played Injustice for almost an hour and my screen on percentage was over 18%.it is extremely cpu & gpu intensive tho, as I've said before. It evened out over time tho, to under 14%.if I was to rate this kernel, STOCK BOEFFLA SETTINGS, from MY EXPERIENCE, it would land a solid 4.1 out of 5. Or good to very good. Or above average. I have gotten better stats in the past with similar usage. But not much better. Had better battery stats and accu battery results for you guys, but they weren't in my screen shots when I tried to add them. But they were really good. Good dozing times, percentages etc. Idle time was very good.
Sent from my OnePlus3T using XDA Labs
I like what you're doing. If you manage to keep your usage pattern consistent, that will be great.
One variable that I know has a big impact is the ratio of wifi to mobile data. I get get better usage when on wifi. And mobile data is also sensitive to signal strength. So, are you doing the tests in the same location or at least with the same profile of locations? And are you using wifi or mobile data?
One suggestion, though I know you've started your tests, is to charge the phone just before you go to sleep and start the test from that point. This will show how well the phone sleeps. The battery stats when you wake up are useful info to record.
You gave Boeffla a score of 4.1. Is this a subjective score or are you using an objective basis for the scoring?
One last comment: on my phone (stock OOS 5.0.1, rooted) I nearly always get excellent sleep results (battery drain of <0.3%/hr overnight). But occasionally I will get ~1.2%/hr due to the system not sleeping (100% awake with the screen off). A reboot fixes this.
So just make sure you aren't dealing with this outlier in your tests.
Looking forward to the rest of your tests.
Sent from my OnePlus3T using XDA Labs
BillGoss said:
I like what you're doing. If you manage to keep your usage pattern consistent, that will be great.
One variable that I know has a big impact is the ratio of wifi to mobile data. I get get better usage when on wifi. And mobile data is also sensitive to signal strength. So, are you doing the tests in the same location or at least with the same profile of locations? And are you using wifi or mobile data?
One suggestion, though I know you've started your tests, is to charge the phone just before you go to sleep and start the test from that point. This will show how well the phone sleeps. The battery stats when you wake up are useful info to record.
You gave Boeffla a score of 4.1. Is this a subjective score or are you using an objective basis for the scoring?
One last comment: on my phone (stock OOS 5.0.1, rooted) I nearly always get excellent sleep results (battery drain of <0.3%/hr overnight). But occasionally I will get ~1.2%/hr due to the system not sleeping (100% awake with the screen off). A reboot fixes this.
So just make sure you aren't dealing with this outlier in your tests.
Looking forward to the rest of your tests.
Click to expand...
Click to collapse
Thanks so much for your valid input. I meant to add information to op, but this is my first thread, and I'm still figuring some things out.
My tests are almost 100% on my home wifi, very little mobile usage. A couple of phone calls and several texts. I may be on mobile signal 10-15 mins a day, going to store and back for beer and/or cigarettes. I am disabled (foot) and work from home which is good for these tests. My wifi signal is pretty much 3 to 4 out of 4 bars constantly, so not much deviation there.
I was going to do my test before bed, but I may go to sleep 9 one night and midnight the next. Same with waking up. May be 6am one morning, and 10 the next. That's why I am doing it afternoonish.
I'm finding the hardest part of the test IS time consistency. Yesterday I completed the test for Boeffla kernel at 3pm, but then it took to 4:20 to fully charge for ceasium test. Then I updated a magisk module, and accidentally reboot killing my battery stats. So ceasium test started at 9pm. What I will make sure of is that it gets the same 23hr test as Boeffla s kernel, with same usage. By creating my own thread, I can share more stats throughout the day.
One last piece of info. I use greenify and naptime for better idle stats. Everything else right now is stock kernel settings the dev of that kernel cooks in. I do not necessarily consider naptime or greenify kernel tweaks, considering either one of those can be used on all kernels with or without root. Those settings will remain unchanged throughout testing, to maintain consistency.
Thanks again
Sent from my OnePlus3T using XDA Labs
So far, about halfway through Caesium testing
Sent from my OnePlus3T using XDA Labs
BillGoss said:
You gave Boeffla a score of 4.1. Is this a subjective score or are you using an objective basis for the scoring?
Looking forward to the rest of your tests.
Click to expand...
Click to collapse
I'm just using my past experiences. I honestly do not expect any test to result in extremely poor results. But ir is paramount to me to be consistent. There is no official grading system being used. Pretty much rating kernels in relation to each others performance. Once again, to be clear. I do not want to criticize any kernel. I respect all of these developers work. I certainly do not have the skills (yet) to build my own kernel. I admire that ability very much
Sent from my OnePlus3T using XDA Labs
There is a poll here!
https://forum.xda-developers.com/oneplus-3t/how-to/oreo-8-1-custom-kernel-test-t3779020
Sent from my OnePlus3T using XDA Labs
gFknQ said:
So far, about halfway through Caesium testing
Click to expand...
Click to collapse
Is that with stock settings ?
amrtalaat said:
Is that with stock settings ?
Click to expand...
Click to collapse
Yes. All I did is flash ceasium kernel and reboot.
Not using any kernel manager to apply any tweaks
Sent from my OnePlus3T using XDA Labs
gFknQ said:
Yes. All I did is flash ceasium kernel and reboot.
Not using any kernel manager to apply any tweaks
Click to expand...
Click to collapse
Im in omni rom . If i flashed this kernel will i get ota weekly updates as regular ?
amrtalaat said:
Im in omni rom . If i flashed this kernel will i get ota weekly updates as regular ?
Click to expand...
Click to collapse
Wrong thread. Ask that in the Omni ROM thread.
Sent from my OnePlus3T using XDA Labs
amrtalaat said:
Im in omni rom . If i flashed this kernel will i get ota weekly updates as regular ?
Click to expand...
Click to collapse
That is good question, unfortunately I do not have an answer. I'd ask omni or ceasium thread. I'm on Nitrogen, so can't say. But I don't see why kernel would interfere with ota at all. It's easy to flash and find out. All you have to do if it doesn't work for some reason is reflash omni rom as you usually would
Sent from my OnePlus3T using XDA Labs
Ok thanks dude :good:
I have concluded this round of testing for Caesium DastardlyDurian.
I'm charging my phone for next test. I'm giving Boeffla a slight edge over Caesium. I'm giving it a 4 out of 5. I'll provide stats, screenshots, reasoning etc after my phone charges a little. I need to be at 100% to start next test.
Sent from my OnePlus3T using XDA Labs
Here is my results for Caesium DastardlyDurian kernel. I was debating between 3.9 and 4, but will gladly give it a 4 and here's why.
I got very slim margin of difference between Boeffla and Caesium kernels. Here's why I was considering a 3.9
1) I've been using NitrogenOS with stock or Mady's kernels for almost 6 months. Startup time using either of these kernels is 99% between 42secs & 56secs. Caesium was also within this range. With Boeffla, it was consistently 32-42secs. This is not a significant factor to me. If I restart my phone 3times a day that's a total difference of about 30 seconds. Not a big deal to me. But it is a difference, and may matter to some. I have experienced this advantage of speedy boot up times with Boeffla kernels in the past. But still it makes little difference to me, but it is a difference.
2)Everything I did while using Boeffla's kernel was extremely snappy. We're talking fractions of a second of difference. But still a difference. I would not say Caesium lagged at all. But I noticed, so. And I had one forced close playing Injustice on Caesium. That is not extremely uncommon to me playing injustice on my tweaks, which are extremely battery saving centered. But I have no tweaks currently set up, just kernel stock. So...
Now here's the thing, and the ultimate reason I would still most likely favor Caesium kernel.
I used Caesium kernels for the majority of the time Oreo first came to OnePlus3T. I mean at the very beginning, it was there. The developer is GREAT. He seems no nonsense, but still allows off topic on his thread. What other OP does that!? I will, I'm fine with it if I can help somebody as long as they don't get upset if I can't. Also, he updates super frequently, and is open to a lot of questions, criticisms, suggestions, etc other devs and op are not.
Boeffla kernel is still considered beta(by him), for custom Oreo on 3t. But Boeffla loves charging as much $ as he can to use his kernel. I know, I've paid. And that's his choice, he can do that if he wants. He's extremely brilliant when it comes to getting the most out of a device and has always been and had one of the very best kernels for any device he's built for. Having said all those things, Jarvis has never charged for Caesium, and remains consistent with development etc. Technically Boeffla doesn't charge to use his kernel. Just for his kernel manager license to tweak his kernel. Currently Boeffla Oreo kernels are free to grab and tweak. Hopefully they will continue to be so after going stable. But based on MY EXPERIENCES, the guy is kind of a jerk. Extremely arrogant. If you don't pay for his license, he will not help you, and possibly not if you do. Not at all open to questions, or criticism etc. That is my opinion and other readers may not agree with my opinions or experiences with either of these developers or kernels.
But here is the results for Caesium. Almost identical usage and stats to my previous test. Great kernel, great developer. To me at this point, between these two kernels, it's a toss up, for the reasons listed above
Sent from my OnePlus3T using XDA Labs
I'm starting next test in 12 mins for ElementalX
Sent from my OnePlus3T using XDA Labs

Categories

Resources