I just got my Thunderbolt yesterday, and so far, I love it. But I've been seeing the battery drain far too fast. According to the stats, it's been awake for a little over 2 hours, not doing anything significant, and the battery is at 60%.
I think this might have something to do with the CPU usage.
If you look at this top screen on the device:
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
The CPU is mostly idle, but the load is >6. I'm not sure what would case this, but I suspect it's related to the power drain.
Anyone have any ideas?
quick q. do you use a lockscreen with pattern lock or PIN or something?
your best bet would be to use system panel, the paid version unfortunately, to analyze a period of a few hours and it will show you a break down of each and every process that hits the CPU, and how much it used, etc all in a very nice graphical view. its similar to the top command, but much more helpful.
ddarvish said:
quick q. do you use a lockscreen with pattern lock or PIN or something?
Click to expand...
Click to collapse
I do (enforced by company policy). Do you think that's related?
Ahhh, I've seen stuff like this on Linux servers before.
In one case, all the system processes were idle, but the load spiked to 25.00 during intensive disk writes. As it turns out, all of this processing was packed invisibly into the kernel; I was using software RAID at the time, and the RAID5 subsystem has to calculate parity bits for written data. So that's what was hogging CPU -- that, and the fact that my stripe width was 1 MB, probably way too big.
In another case, there was a kernel bug with the development kernel I was using (I think it was an -rc of 2.6.32) where something in the kernel core was constantly awake -- an infinite loop or livelock or something.
There's certainly the possibility that HTC and Verizon might have both missed something as glaring as that, but then their testing would have revealed markedly low battery life -- could they ship the device in good conscience in that state? WOULD they?
Basically, you can get high "load" in top even when it doesn't correspond to a userspace process or a labeled kernel thread. Things that contribute to load include:
1. Context switches (between user and kernel)
2. I/O (i.e., network, disk)
3. Kernel activity, both within the main kernel binary and in modules
4. CPU activity in userspace processes
Unfortunately, the process list in top doesn't always jive with the numbers in the load, because not all of the things that contribute to load can be attributed to a single process.
For example, if there's a kernel bug, and a single syscall from a userspace process foo.bin costs an unreasonable number of context switches, the CPU will spend as much time as it can trying to context switch to satisfy the syscall. I suppose if there were a "kernel" task in top, you could attribute the CPU time to that, but it wouldn't be fair to blame it on "foo.bin" because all the work (the context switches themselves) is being done in the kernel. That's just one possible example.
Oh, I forgot to ask: are you root? If not, you may not see processes reported in top that are owned by other users. So you might not even be getting the full picture of the userspace processes running on the device, let alone kernel tasks and other less tangible effects.
Before I rooted, I had System Panel installed and everytime I checked my CPU was at 99% and full 1024 on the clock.
After root, my CPU drops into the 1 digit %'s all the time, CPU speed runs between 600 - 1000mhz.
edit: this is on the stock RUU and stock kernel
Adam B. said:
I do (enforced by company policy). Do you think that's related?
Click to expand...
Click to collapse
i dunno.. it may be. read the thread i started here on why..http://forum.xda-developers.com/showthread.php?p=12301209#post12301209
i am trying to run a couple experiments in the next day or two to figure it out. its crazy that battery life varies SO much for people.
allquixotic said:
Ahhh, I've seen stuff like this on Linux servers before.
In one case, all the system processes were idle, but the load spiked to 25.00 during intensive disk writes. As it turns out, all of this processing was packed invisibly into the kernel; I was using software RAID at the time, and the RAID5 subsystem has to calculate parity bits for written data. So that's what was hogging CPU -- that, and the fact that my stripe width was 1 MB, probably way too big.
In another case, there was a kernel bug with the development kernel I was using (I think it was an -rc of 2.6.32) where something in the kernel core was constantly awake -- an infinite loop or livelock or something.
There's certainly the possibility that HTC and Verizon might have both missed something as glaring as that, but then their testing would have revealed markedly low battery life -- could they ship the device in good conscience in that state? WOULD they?
Basically, you can get high "load" in top even when it doesn't correspond to a userspace process or a labeled kernel thread. Things that contribute to load include:
1. Context switches (between user and kernel)
2. I/O (i.e., network, disk)
3. Kernel activity, both within the main kernel binary and in modules
4. CPU activity in userspace processes
Unfortunately, the process list in top doesn't always jive with the numbers in the load, because not all of the things that contribute to load can be attributed to a single process.
For example, if there's a kernel bug, and a single syscall from a userspace process foo.bin costs an unreasonable number of context switches, the CPU will spend as much time as it can trying to context switch to satisfy the syscall. I suppose if there were a "kernel" task in top, you could attribute the CPU time to that, but it wouldn't be fair to blame it on "foo.bin" because all the work (the context switches themselves) is being done in the kernel. That's just one possible example.
Oh, I forgot to ask: are you root? If not, you may not see processes reported in top that are owned by other users. So you might not even be getting the full picture of the userspace processes running on the device, let alone kernel tasks and other less tangible effects.
Click to expand...
Click to collapse
Interesting info. I'm also a Linux admin, but I haven't really run into situations like this.
It is very odd that I see nothing in the user space that would cause this. http://pastebin.com/J4PwuMeZ Every process is sleeping, and nothing seems to be waiting on IO.
I suppose it could be an issue with the kernel. We'll have to wait and see what happens when the source is released and people start to release new updates.
I am root, btw.
Related
Geek 4 Me is proud to release
Fresh Toast v1.1 by toastcfh and flipz
First off let me remind everyone that this is a dev rom. Meaning it is a proof of concept type rom and is NOT going to be fully functional. This is built on the .27 kernel which means no camera and no light sensor (causing bottom lights to not come on). With that said...!
New features in 1.1:
Full scaling overclocking. You can now use setcpu to set the speed. Anywhere from 245mhz to 767mhz - toast. Thanks to phh and maejrep for their help. Download setcpu from the market for $1 or free here. Note: When you set it up and asks for device, scroll all the way to the bottom and pick custom!
BFS (Brain **** Scheduler) version 316 has been added and customized. This dramatically reduces latency on the phone. While it doesn't increase benchmarks it will really increase the usability and smoothness of the phone. Go here to learn more. -- toast. Thanks to maejrep for help.
Fixed apps2sd and ext4 not working together
Removed AutoKiller. It will likely still show up on peoples phones because it's on your /data/ partition but I decided to remove it after I realized it was reporting location and who knows what else. I already created a modified version that leaves the ads for the developers sake but strips the Flurry reporting however it's not playing nicely with installing. I'll get an update.zip posted when I figure out it's problem. -- flipz
Found and fixed a bug in HTC/Sprints code that was causing 99% cpu usage, draining battery, and causing overall slowness. This bug doesn't seem to affect everyone however it was affecting me big time. It exists in every Sprint 2.1 RUU based build. --flipz
Features:
Custom kernel -- toast
OVERCLOCKED to 767mhz -- toast, with help from phh and madcoder
Compcache and ramzswap enabled -- toast and flipz (? and ? if you don't know what they are. **read warning below)
Ext4 support -- toast (has many benefits over ext3. ? for more info)
Fixed vanilla lock screen (including airplane mode/emergency call glitch) -- flipz
Some modified artwork (toolbar and selector, because toast loves it) -- flipz
Newest autokiller (more info on this app: http://geekfor.me/faq/autokiller)
Disabled home screen menu from popping up when using menu button to unlock --flipz ( patch to disable menu to unlock if you prefer: http://geekfor.me/fixes/bugfixes-requests-lock-screens)
Uses Fresh Rom 2.0d as it's base so all other fresh rom features are in here -- flipz
Not working:
Camera
Light sensor (causing bottom lights to not come on)
Please don't ask when these will be fixed. The short answer is they require the new kernel to work on Android 2.1, however we can't enable all these cool features on the new kernel until we get the source from HTC... again. But as always, toast is working on fixes. So buy him some coffee.
Compcache/ramzswap questions:
How do I make it work? Your sdcard must be partitioned with a swap partition. Most anyone running apps2sd will have this already. Other than that it's automated.
How do I know if it's working? Open an adb shell and type in "free". If you have "swap" listed with any numbers next to it, it's working.
What is required to run it? A class 6 sdcard is highly recommended for speed and stability
How do I disable it? Just delete or rename the script at /system/etc/init.d/compcache.sh to anything else and reboot
**Will this wear out my sdcard? Just like apps2sd, ramzswap puts extra strain on your sdcard. If you are using a decent sdcard then it shouldn't be a problem. Many professional grade HD video cameras use sdcards as their medium, and they don't burn out. If you are using a low quality card then this may cause it to prematurely fail. Do not blame us. You have been warned!
Overclocking questions:
How do I know if it's working? This is hard to prove, other than to say you can run benchmarks and clearly see it. Any CPU applications (including setcpu and OC widget) are going to continue to report the stock speed of 528mhz. Full scaling IS coming soon which will allow you to select the correct max speed.
Will this hurt my phone? Toast and I have been running this for days with no ill effects, even when stress tested for hours. There were no crashes, no errors, and no significant heat gains. However, like always, you are doing this at your own risk and we can't be held responsible.
Do I need to use setcpu/oc widget? They will provide you 0 benefit. The phone has HTC's power manager enabled which means it will throttle up the CPU when it's needed, and throttle it down when it's not. As mentioned above, full scaling is coming soon which will allow you to adjust the speed through those apps if you choose.
Make sure you are using a temperature monitor if you are overclocking and watch your temps.
How do I flash this rom?
It is very likely that you can flash this rom on top of Fresh 2.0d without a wipe and have it work without any issues, however the following procedure would be recommended (assuming you have a class 6 sdcard).
Boot in to RA and do data wipe
Format your sdcard with 32mb swap, 512-1gb ext2, and the rest fat32
Convert ext2 to ext3, and then ext3 to ext4
Flash the rom and reboot
Download: http://geekfor.me/new-release/fresh-toast-v1-1/
Info:
...................................
Yess been waiting on this bout to flash
You all are amazing.. Im downloading now.
by fixing the lockscreen vanila you mean you can change wallpapers? is it asop?
magicalan said:
by fixing the lockscreen vanila you mean you can change wallpapers? is it asop?
Click to expand...
Click to collapse
No, sorry about the confusion. Just meaning it has the working vanilla lock screen including the fix where the emergency call button would show up after being in airplane mode. And no it's not AOSP, it's based off of fresh rom. It says that.
Goodnight all!
It does seem faster especially in sense
Love it and you all, in a very manly way!
You guys are bad ass, flashing now. Thank you!
BENCHMARKS
Best scores received so far Will be updating if I get better- Screen shots might be different
Linpack: 5.022 mflops
BenchmarkPi: 6298 ms
Neocore: coming soon
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
How's battery life?
Yeah the question above me ^^^^^^^
Try it and see?
Technically battery life should be about the same as whatever you're getting now. The speed increase is 'on demand' so unless you spend all day doing cpu intensive tasks (and no, phone calls aren't cpu intensive ) you should be fine.
I'm going to wait for the .29 kernel source to be released before I play with this. Unfortunately I need my camera.
Jus10o said:
BENCHMARKS
Best scores received so far Will be updating if I get better- Screen shots might be different
Linpack: 5.022 mflops
BenchmarkPi: 6298 ms
Neocore: coming soon
Click to expand...
Click to collapse
Nice! Would you mind sharing your settings/setup? You are currently kicking my phone's trash.
Trying it just to give it a shot but not keeping it till I get a class 6 card. Noone in my area sells them. Max is 4 atm. Hate living in an area where technology can only be acquired online.
smw6180 said:
Try it and see?
Technically battery life should be about the same as whatever you're getting now. The speed increase is 'on demand' so unless you spend all day doing cpu intensive tasks (and no, phone calls aren't cpu intensive ) you should be fine.
I'm going to wait for the .29 kernel source to be released before I play with this. Unfortunately I need my camera.
Click to expand...
Click to collapse
I spend all day using Google Nav
Costa276 said:
Trying it just to give it a shot but not keeping it till I get a class 6 card. Noone in my area sells them. Max is 4 atm. Hate living in an area where technology can only be acquired online.
Click to expand...
Click to collapse
I am surrounded by retail stores and I only buy technology online. Amazon and Google Shopping are your friends.
I am enabling JIT now to see if I can get some more competitive benchmark scores.
danknee said:
I am enabling .jit now to see if I can get some more competitive benchmark scores.
Click to expand...
Click to collapse
Me too! Just waiting for it to d/l. 5MFLOPS just with the clock is amazing enough, I can't wait to see what it hits with JIT.
Mad props guys!! Nice work.. hope you guys can get it on .29 kernel
abcdfv said:
Me too! Just waiting for it to d/l. 5MFLOPS just with the clock is amazing enough, I can't wait to see what it hits with JIT.
Click to expand...
Click to collapse
I'm positive that Jus10o's scores are with JIT enabled. Without it, I can't even get close to those scores.
Plans for next revision:
*Add in Decad3nce's temperature control
*Bug nullghost till he gets init scripts running so we can have this startup at boot
This is really a tool for power users
If you don't consider yourself 'comfortable' with messing with your phone, then don't.
This is mostly the work of collin_ph, all the ideas are mostly his and I have only contributed in part to getting some things functioning correctly and some optimizations for the Epic.
What does it do?
This tweak creates a service that periodically checks the phone's power source.
If the power source changes, it reconfigures the device for maximum performace, battery and safety.
Tweaks memory settings for current power source
Re-clocks CPU based upon battery remaining or current power source
Reclocks for maximum performance if on AC power
Reclocks for maximum performance and charging safety if on USB
Reclocks for performance and battery if on Battery
As Battery life is diminished, Maximum CPU frequency is lowered
Tweaks scaling charactaristics depending on power sourcee
CPU Scales up less frequently on battery power
CPU Scales up more frequently on USB/AC power
The CPU is allowed to scale dynamically in all charging states based upon CPU usage
Includes a batt-cfg utility to configure the system
You may load 1 of 3 pre built profiles
Manually configure CPU Frequencies in all charging states
Determine the percentage of underclocking as the battery drains
Configure how often to poll the power state / battery statys
Disable batt.sh service from running at startup
Includes a batt-rm.sh uninstall utility
Includes a batt-diag diagnostic utility
Includes optional Disk performance tweaks (enabled by default)
The Tools
The entire 'application' is written in shell code and was first written just to give a performance and battery boost, but since has evolved more and more.
All the tools mentioned will be utilized either through the terminal, or in adb shell.
batt.sh
This is the core of the entire script. Don't mess with it.
The one thing you should do after installing the script:
Code:
Code:
pidof batt.sh
If theres any number showing up after it, then it's working fine.
batt-cfg
This is the 'main menu' of sorts, as it gives the user all the options and allows them to toggle tweaks/mods on the fly.
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
batt-diag
It's a diagnostics utility, allowing you or someone knowledgeable to figure out any issues that you may be having with your device and be able to pinpoint exactly where the problem is.
batt-diag -v gives a more thorough output as you can see here; http://pastie.org/1162034
batt-rm.sh
It's a battery tweak removal script, follow the onscreen prompts and it should remove just about any trace of the battery tweak script.
The Presets
The presets are just default configurations that I have set up that can make your phone more efficient without you, the user, having to manually set everything.
The preset of 800 literally means that 800000 is the scaling_max_freq, or the max frequency, that your phone can achieve after you enable it.
Just typing 1, and then enter, you'll be ready and set to go. Same with 2 or 3.
Customize Settings
This is only for the experienced user, and it allows you to manipulate most/ifnotall the configurables on your device.
Note, you do not need to configure anything to enable this battery mod. Default settings will be assigned at 800MHz.
All the settings explain what each option does, so read them carefully.
Disk Boost
Remounts via noatime and nodiratime. Look up what that means on your own, and why it benefits performance the way it does.
Prerequisites needed to run this
Phoenix kernel 1.48 with Epic Experience should run this just fine
If you're running any other kind of rom, feel free to try it but i cannot guarantee that it will work.
This script will WORK even if you don't have the import call at boot, but it won't survive on the next reboot.
You can check via:
Code:
Code:
pidof batt.sh
NOTE: This script will also prompt you to try to run itself as root if you use it in terminal.
Contributors:
Collin_ph
Decad3nce
myn
onicrom
shabbypenguin
nullghost (aka tehdarkknight)
*make sure you are running these as root*
*disable setcpu while running this*
*been reported as rom manager fc's after batt tweak, simple fix reinstall rom manager* - thanks for the tip Raiderep
Finally got that working
This is perfect. As a former hero user, I was thinking to myself a couple days ago "I wonder if anyone is going to port the collin_ph battery tweak," and here it is.
Thanks!
I'm about to head into work and I got this downloaded. I'll put it to the test for the day. Thanx for the port or whatever you call it, I had this on hero and loved being able to get rid of setcpu(which I never really liked) but still have something governing battery use
Is this compatible or even needed with the mix up kernel?
compatible? maybe as i said this is known working on phoenix kernel as for others i cannot say. needed? well that depends are you happy with your current battery life?
Using Phoenix kernel 1.48 with Epic Experience bogged down my phone after installing, im only at 76% battery life, i cant imagine how slow it would be at 20%
edit:
its fine now, i guess it had to load initially?
Whats the best settings you think we should run..for the best battery life
800mhz or you can use configure to keep that even lower... obviously the lower the cpu freq the less voltage it uses = better battery life
I just set this up to the 800mhz defaults. I will come back with results
How does this govern cpu clocking and usage? I'm assuming there's no live governer such as ondemand or interactive at play here?
phoenix 1.48 has a working interactive governor
shabbypenguin said:
compatible? maybe as i said this is known working on phoenix kernel as for others i cannot say. needed? well that depends are you happy with your current battery life?
Click to expand...
Click to collapse
I don't think anybody is That being said, great work shabby, can't wait til nullghost gets his init scripts working.
shabbypenguin said:
phoenix 1.48 has a working interactive governor
Click to expand...
Click to collapse
So how does this affect setcpu? I'm runnin baked 1.3 ghz kernel..does this mean after I install I won't be at 1.3?
Sent from my SPH-D700 using Tapatalk
this is set for 800/1ghz/1.2ghz i included the configure option so people can make teh .conf however they see fit.. if batt.sh runs on it then it will scale from 1.3ghz after you configure it. witht he beta versions i dont believe anyone got it working on baked snack but if you try please let me know if it works what doesnt etc etc
when i try to run batt.sh i get this message...Mount: cant find /mnt/sdcard in /proc/mounts..what does this mean
its an issue im trying to figure out, batt.sh works fine without it, thats part of the disk speedups, in batt-cfg you can disable it
new feature will update tomorrow
not sure why but shootme is screwy for me recently as you can see. i have 0% tws, 37% voice calls at 30 mins 29s display is on for over an hour. screen is at auto, 3g, drm services RUNNING. oh yea im at 1.2ghz
Still have problems with this tweak...this is wat iam getting when I do batt-cfg I hit number one and I get this....
Loading 800mhz default
[: Bad number
Save settings...
Settings saved
Starting batt.sh
Note: when you exit ADB or terminal, you may have to hit
CTRL-C
Then my phone freezes and I have to pull the battery
Sent from my SPH-D700 using Tapatalk
I'll start by saying this isn't about posting highest bm results. I'm just trying to find best results in statistical form for my daily driver. But i'm not sure if this is a good practice or not. I'm running Basic with a twist, kg4 modem, and Antonx barebone. I go into voltage control, change a setting(io scheduler, and overclock) then run a bm test, take a screenshot. I did this 8 times, and picked the setting with the best results(which I found to be noop, smartassv2, 200/1300, -25} Other than battery life, would you say this is a good way of doing it? Or am I completely off base? Phone feels great, and battery lasts all day BTW, but just looking for some insight. Thanks!
Sent from my SGH-T959V using xda app-developers app
I think you're on the right track.
There are sort of two things to consider, as I see it, past "the phone is responsive enough for my needs"
Deep sleep time when not "in use"
Drain when in use and screen on
The first of those is probably best examined by using Better Battery Stats religiously. Probably half the nights of the week I charge my phone fully before going to bed, pull the charger, and turn off the screen. Then just leave it there. No checking your stock prices at 4 A.M. When I wake up, I immediately open Better Battery Stats and save a report (make sure you have configured it ahead of time to include everything on the list). If you're seeing more than 5-10% of time time that your phone isn't in deep sleep on CM9 or AOKP, your drain "problems" are more likely your apps or settings than the ROM.
Given that I am seeing on the order of 1% per hour average battery drain (~20 mA average over an hour) and that it comes in spurts (my email wakes the phone for 5 seconds, it goes back to sleep, my Twitter app wakes the phone for 5 seconds,...) I think you would be hard pressed to measure the current with a meter that most had access to.
Wake time, in my opinion, is dominated by screen current (hence why I wracked my brain on the auto-brightness issue and agonize over adding another step in the brightness curve that would consume more current, even if it is easier to read) and, for some users, CPU/GPU. The Android screen that shows battery usage by app is probably pretty good for this to a first order, but as far as app by app, I don't believe it (there is nothing in the device that actually measures current draw other than total from the charger cord), and certainly not for app-by-app decisions.
I think the idea of a "benchmark" of your everyday apps the way you use your phone is a good one. When I was tuning Hefe kernel I tried to use my phone the way I generally did everyday, then collected data over several days at each setting change. You're on the right track about taking several data points to try to smooth out "normal" variation.
There have been other attempts to measure power consumption, but I have some serious doubts about the methodology. For example dBm is a power measurement and requires a known impedance, which the phone certainly is not. The mentions of "logarithmic scale" support that the wrong measurement was made (mA is a linear scale on an analog meter). As a result, I find the results "interesting," but I wouldn't bank on any of them to guide decisions.
Edit -- For example, the post referenced in the previous paragraph suggests that undervolting can only have a 2% effect. However, power required to charge a capacitor goes like the square of voltage (IC "gates" look a lot like capacitors when they are working), and this quadratic behavior can be easily confirmed elsewhere. Undervolting by 5% (60 mV on 1200 mV) would then be roughly a 10% gain; undervolting by 10%, if your phone could take it, roughly a 20% gain.
Talking with Raver on irc a while back he said overclocking past 1.2 will end up hurting battery life. Im not sure if that applies to gb roms or just the ics flavors tho. But it is something to consider when using oc/uv settings. I never go past 1.2 and my voltages are -50 from top to bottom. I average 12-14 hours depending on how many phone calls I have to make.
eollie said:
Talking with Raver on irc a while back he said overclocking past 1.2 will end up hurting battery life. Im not sure if that applies to gb roms or just the ics flavors tho. But it is something to consider when using oc/uv settings. I never go past 1.2 and my voltages are -50 from top to bottom. I average 12-14 hours depending on how many phone calls I have to make.
Click to expand...
Click to collapse
This is specific to the DVFS based cpufreq.
The aries kernel and the blastoff v3 (to be released... soon) use standard (non-DVFS) cpufreq.
We will see different OC/UV results with these kernels.
I'd also add that no matter how it is done, overclocking any amount will pretty likely increase CPU power consumption.
(make sure you have configured it ahead of time to include everything on the list).
Ok,
This is extremely helpful and very interesting. Thanks Jeffsf! (and Eollie and Bhundven)
I just downloaded BBS off of Play store. I'm attempting to set up and include everything on the list, but it's only letting me select one at a time ( Other, kernel wakelock, partial wakelock,alarms etc. etc.) Is this what you're referring to? I was reading through BBS devs O.P., maybe I'm missing the first time setup instructions. But I see the "getting started", and "how to" but nothing on how to select all. However "Other=Group of different indicators about what is consuming the battery. This statistic should always be checked first as it gives a good idea about the draining profile". <This quote makes me think each has to ran seperately.... True?
Thanks
They all "run" even if you don't have them selected. I looked and the current settings for what goes into the dump file are from the ICS-style settings icon near the top right, then scroll down to Dumpfile and click it. Select everything on the next screen.
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
One of the great things about BBS is that it primarily uses stats the kernel already collects, so it doesn't interfere much, much, if any, with what the phone would otherwise do.
Posted from my SGS4G, thanks to Team Acid development
I started my first step last night. The results are not terrible, but there is certainly some room for improvement. I was surprised to find the top culprit under wakelock is *backup* Android system= 6min. 44s ( second entry is only 11s., huge jump). Seems the consensus online is go to Settings>Privacy>Backup data, and uncheck. I started to do this, but when I unchecked the backup, I got a message saying something like "are you sure you want to do this? You're gonna F up your phone" So I am holding off for a minute. 2nd highest wakelock is Alarm manager (which I'm stil looking into) at 11s., and 3rd is GTALK_ASYNC_Conn at 7sec. Which I believe I may have fixed by Super Manager>APK startup>Google framework g-talk>disable startup..... we'll see at next run.
The how to guide on BBS says if Power Manager is not the top entry under Kernel Wakelocks, then check Network stats. I actually have 4 entries above that. Starting with SVnet at 5min 58s. I'm thinking this may be email check frequency, but there is nothing listed under Network on my dump file.
Radio interface 3min 54s
mmc_delayed_work 2min 32s
Anyways, I'm still looking into all of this, and going to attack little by little each day. But I figured I'd share, and maybe get some insight.
Thanks all!
Sounds like you're on your way to at least understanding your phone a lot better than most. It's well worth the time, at least in my personal experience. I went from not being able to make it through the day on GB, to getting a day's run time without a sweat in ICS.
I believe you're right that svnet is the radio (the "phone" one, as compared to the "WiFi" one).
mmc_delayed_work, from staring at kernel code, appears to be part of the "normal" in the process of writing to disk. I haven't gone much further than that, but I am guessing it is a periodic "flush" of queued write requests.
I used search and looked through several pages of Q&A threads and didn't see an answer to this. If I missed the thread, please point me to it and accept my apology.
I updated to check for the recall and I'll be getting a new Shield Tablet eventually but my battery life nosedived immediately after the update. I read that Lollypop has had battery issues but the only fix I seem to read about elsewhere is flashing CM which does not interest me particularly. My battery life instantly dropped from 4-6 hours to barely 2 since the update.
When I went to battery stats prior to the update, I saw all the apps that were running and after the update I only see things like "screen," "android os," "android system," and "chrome." I'll have lost 50% of the battery and battery stats will tell me that the screen has used 15%, os & system have used 5% but there aren't enough percentages to add up to either the 50% lost or 100% of used battery life. I'm not even sure how to read the stats anymore.
Is there a kernel I can flash or something to regain the battery life I had prior to the update?
Any assistance is most appreciated.
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
Hi
I've actually noticed the same thing, thought I had lost my mind or something. Luckily I'm getting a RMA ( faulty battery recall)
But I would really like to find a solution to this in the mean time...
Tried a factory reset?
Its all about using hardware ....
Narsil007 said:
I used search and looked through several pages of Q&A threads and didn't see an answer to this. If I missed the thread, please point me to it and accept my apology.
I updated to check for the recall and I'll be getting a new Shield Tablet eventually but my battery life nosedived immediately after the update. I read that Lollypop has had battery issues but the only fix I seem to read about elsewhere is flashing CM which does not interest me particularly. My battery life instantly dropped from 4-6 hours to barely 2 since the update.
When I went to battery stats prior to the update, I saw all the apps that were running and after the update I only see things like "screen," "android os," "android system," and "chrome." I'll have lost 50% of the battery and battery stats will tell me that the screen has used 15%, os & system have used 5% but there aren't enough percentages to add up to either the 50% lost or 100% of used battery life. I'm not even sure how to read the stats anymore.
Is there a kernel I can flash or something to regain the battery life I had prior to the update?
Any assistance is most appreciated.
Click to expand...
Click to collapse
Okay so In your tablet you got 4 core
4 core means twice as much battery use because you can adjust it y dividing them to half
which is with update they made the cpus using full capacity all 4 of them
so when your not gaming or with your gaming which is it didnt show any difference to me while Im playing heartstone
settings >shield power control>my power mode enable again and customize my power mode >max cpu cores 2
then try with this and youll be allright also
another poing I would love to make and share my experience with you
So far if any of your google apps or location apps is on they draw approx 30 percent more battery
Its disgracefull I know so for that
Settings>location>off
then another thing is I come across a lot
settings > wifi >advanced options ( top of right corner 3 dots)>Scanning always available is off
:fingers-crossed:
Limiting the number of active cups is not the option. But three are other settings related to CPU which not te reduce their number but reducing their frequency first. Your room possibly is too much conservative about power usage because Android is made to used the "on demand" governor for CPU settings and assumes that there's a scheduler that will update the frequency every few seconds instead of every minute where all cups will email active for at least some time, so that it is never really idle. The exists some tools like CPU tuner that allows defining the CPU parameters and the threshold of time usage for which the usage rates can be updated down or up.
Envoyé de mon Desire 816 en utilisant Tapatalk
Has anyone noticed that under the shield power control --> apps , one can no longer set apps to "let Nvidia optimise it" . Only two option available, disabled or custom
Running the latest update
Anyone else?
verdy_p said:
Limiting the number of active cups is not the option. But three are other settings related to CPU which not te reduce their number but reducing their frequency first. Your room possibly is too much conservative about power usage because Android is made to used the "on demand" governor for CPU settings and assumes that there's a scheduler that will update the frequency every few seconds instead of every minute where all cups will email active for at least some time, so that it is never really idle. The exists some tools like CPU tuner that allows defining the CPU parameters and the threshold of time usage for which the usage rates can be updated down or up.
Envoyé de mon Desire 816 en utilisant Tapatalk
Click to expand...
Click to collapse
Well I did it and its perfect.
I even got blissrom now
And I did customised again cooler then ever so
Your wrong by saying that "not the option" matey .
Sent from my SHIELD Tablet using Tapatalk
So, after months of using Pixel 7 family, surely we all know that Tensor 2 is not exactly the best flagship SOC in term of raw power or energy usage, probably due to it's "Exynos Genetics Trait"
that being said, doesnt mean it's down bad. And with some optimisation, we can always try to make it run better. One of the example is by seeing how much better it run on custom kernel. But what about other stock kernel user ?
Well, fortunately, as long as we have root access, the stock kernel still allow us to play with some parameter. And here are some tunables parameters that i use to balance the energy usage and responsiveness of the device. (my aim is 10% battery for 1 Hour of SoT and ±0-1% decrease during 6-10hrs idle)
If you want to know more about each parameters, you can google a lot of linux kernel documentation (like here for example : https://access.redhat.com/solutions/177953)
but for now i'll just write down the change and my configurations. (I used EX Kernel Manager to change the parameters) :
CPU Section :
Switch governors for all CPU clusters from 'sched_pixel' to 'shedutil'
Open "Governor Options" and change the 'rate_limit_us' for all clusters from '10000' to '0'
Open "CPU scheduler options" and use bellow value for respected parameters :
sched_wakeup_granularity_ns : 10000000
sched_tunable_scaling : 1
sched_min_granulatity_ns : 2000000
sched_latency_ns : 12000000
Memory Section :
Disable "ZRAM"
That's it for now. Try and feedback if you find any improvement in your energy usage.
I've been using it for weeks, and pretty happy especially with idle drain. As for SoT, normally i end up with 5-6 Hrs screen time when i down to bed with ±30% battery remaining.
otonieru said:
So, after months of using Pixel 7 family, surely we all know that Tensor 2 is not exactly the best flagship SOC in term of raw power or energy usage, probably due to it's "Exynos Genetics Trait"
that being said, doesnt mean it's down bad. And with some optimisation, we can always try to make it run better. One of the example is by seeing how much better it run on custom kernel. But what about other stock kernel user ?
Well, fortunately, as long as we have root access, the stock kernel still allow us to play with some parameter. And here are some tunables parameters that i use to balance the energy usage and responsiveness of the device. (my aim is 10% battery for 1 Hour of SoT and ±0-1% decrease during 6-10hrs idle)
If you want to know more about each parameters, you can google a lot of linux kernel documentation (like here for example : https://access.redhat.com/solutions/177953)
but for now i'll just write down the change and my configurations. (I used EX Kernel Manager to change the parameters) :
CPU Section :
Switch governors for all CPU clusters from 'sched_pixel' to 'shedutil'
Open "Governor Options" and change the 'rate_limit_us' for all clusters from '10000' to '0'
Open "CPU scheduler options" and use bellow value for respected parameters :
sched_wakeup_granularity_ns : 10000000
sched_tunable_scaling : 1
sched_min_granulatity_ns : 2000000
sched_latency_ns : 12000000
Memory Section :
Disable "ZRAM"
That's it for now. Try and feedback if you find any improvement in your energy usage.
I've been using it for weeks, and pretty happy especially with idle drain. As for SoT, normally i end up with 5-6 Hrs screen time when i down to bed with ±30% battery remaining.
Click to expand...
Click to collapse
Someone said that it reverts back once you reboot?
elong7681 said:
Someone said that it reverts back once you reboot?
Click to expand...
Click to collapse
In the Franco Kernel Manager, there's an option to "keep" the new settings on a reboot. (Not sure about the EX Kernel Manager.)
elong7681 said:
Someone said that it reverts back once you reboot?
Click to expand...
Click to collapse
Right, it reverts back for CPU Governor (Cluster 3) after a reboot. (with EXKM)
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
Then I tried with FKM and problem is the same.
I will see after the night how is my battery drain.
But i can't complain because i loose only 2 or 3% during night (8h)
(Bluetooth and 5G on) No wifi because 5G signal is good.
No custom rom, rooted, january update.
Plathoon said:
Right, it reverts back for CPU Governor (Cluster 3) after a reboot. (with EXKM)
View attachment 5817379
Then I tried with FKM and problem is the same.
I will see after the night how is my battery drain.
But i can't complain because i loose only 2 or 3% during night (8h)
(Bluetooth and 5G on) No wifi because 5G signal is good.
No custom rom, rooted, january update.
Click to expand...
Click to collapse
It's expected that it reverts back to default because it's controlled by multiple things. People shouldn't change the governor. There is a write-up on this in Freak's kernel thread.
ekin_strops said:
It's expected that it reverts back to default because it's controlled by multiple things. People shouldn't change the governor. There is a write-up on this in Freak's kernel thread.
Click to expand...
Click to collapse
Can you provide a link to the write-up please?
Pixel86 said:
Can you provide a link to the write-up please?
Click to expand...
Click to collapse
Post
Mrcactuseater what's your deal always laughing at genuine posts?
ekin_strops said:
It's expected that it reverts back to default because it's controlled by multiple things. People shouldn't change the governor. There is a write-up on this in Freak's kernel thread.
Click to expand...
Click to collapse
Actually you can change the CPU governor from the default one to one of the additional existing alternatives.
The issue is just that the frequency scaling driver for the affected CPU cluster(s) will break if the PowerHAL detects that the Sched_Pixel governor isn't "there" anymore for doing the job that it is supposed to do.
Plathoon said:
But i can't complain because i loose only 2 or 3% during night (8h)
(Bluetooth and 5G on) No wifi because 5G signal is good.
No custom rom, rooted, january update.
Click to expand...
Click to collapse
Damn that's less than half of mine...
Not_Purgatory said:
Actually you can change the CPU governor from the default one to one of the additional existing alternatives.
The issue is just that the frequency scaling driver for the affected CPU cluster(s) will break if the PowerHAL detects that the Sched_Pixel governor isn't "there" anymore for doing the job that it is supposed to do.
Click to expand...
Click to collapse
I never wrote you can't change it, I wrote people SHOULDN'T change it to avoid the issues you mentioned.
You say to try to deactivate zram. Is there any real benefit? Zram is useful?
Techboyz97 said:
You say to try to deactivate zram. Is there any real benefit? Zram is useful?
Click to expand...
Click to collapse
You save CPU cycles at the expense of additional vRAM.
If you use smartpack kernel manager there's an option to apply all changes with a script on boot. That's the way I've always worked with kernels that won't allow changes to stick on boot
Pixel86 said:
You save CPU cycles at the expense of additional vRAM.
Click to expand...
Click to collapse
WUT? lol