[BENCHMARKS]Kernel Features, common misconceptions, myths busted - Nexus S Android Development

Hello, here are some benchmark i made to test if some features being used in kernel development are usefull, useless, bull**** or make things worse.
HOW I DID THE TESTS
DEVICE= i9023.
ENVIRONMENT= fixed 25° Celsius.
OS= ANDROID 4.4.1 JRO03E Factory Image by Google.
SOFTWARE USED= 0xBenchmark 1.1.5 - AnTuTu Benchmark 2.9 - Screen Timeout Toggle.
OTHER TOOLS= A/C Charger / Standard Chronometer.
KERNEL= Kernels are built from source using the standard herring defconfig.
Additional notes:
The system is booted up once, every tutorial is closed, 0xbenchmark, Antutu and Screen Timeout Toggle are installed, airplane mode is toggled, system is rebooted in recovery, battery stats are deleted, cache and dalvik are cleared, system is nand backupped.
Every test starts after 30 min of phone off to let him cool, restoring the nand backup and waiting 5 min after system is booted up. Phone is connected to A/C charger.
Kernel are swapped after the nand restore.
Tests are done 5 times and then the average is calculated && till results are almost the same every run.
TEST N. 1
".. i use teh latest toolchain, mah kernel is imba ima pro !!111!1one!!eleven"
Google toolchain 4.4.3 vs Google toolchain 4.6
This test is inspired by an Ezekeel work that demonstrate how every different toolchain from the google base 4.4.3 used to compile our NS kernel resulted in 0 increased performance. Same goes for "optimized" compiler flags. You can see some bench here.
What i'm going to do is to test latest google prebuilt toolchain and see if it differs from above test.
- 0xBenchmark reds results are better.
Code:
Toolchain 4.4.3 Toolchain 4.6
Linpack [COLOR="Red"]18,81[/COLOR] 18,31
C [COLOR="Red"]21,65[/COLOR] 21,15
FFT [COLOR="Red"]13,92[/COLOR] 13,59
JSOr [COLOR="Red"]39,79[/COLOR] 39,14
MCi [COLOR="Red"]7,20[/COLOR] 6,60
Smm [COLOR="Red"]17,80[/COLOR] 17,45
dLUmf [COLOR="Red"]29,61[/COLOR] 29,17
- AnTuTu reds results are better.
Code:
Toolchain 4.4.3 Toolchain 4.6
RAM [COLOR="Red"]260[/COLOR] 257
CPU Integer 416 416
CPU Float-Point [COLOR="Red"]106[/COLOR] 105
GFX 2D [COLOR="Red"]278[/COLOR] 277
GFX 3D [COLOR="Red"]1115[/COLOR] 1111
TL;DR
USING LATEST GOOGLE TOOLCHAIN DOES IMPROVE KERNEL PERFORMANCE? NO
TEST N.2
"..undervolting teh lcd display MUST save battery!!"
LCD @ 3.0 V vs LCD undervolted to 2.4V
Same environment as before. Since % battery are not always accurate i made 3 tests:
2.1: let phone fully discharge, charge it up for 30 min. Boot it up, put max brightness and count how much time passes till it poweroff by himself.
2.2: let phone fully charge, boot it up, put max brightness and count how much time passes till it loose 10 points %.
2.3: let phone fully charge, boot it up, put max brightness and count how much time passes till it goes from 60% to 50%.
RESULTS
After days of tests, can pretty sure say that at the cost of 20% undervolt (from 3.0 to 2.4) there isn't any noticeable battery saving. What i came up with is something like 5%, that means something like 10 more screen time with standard use, even less, and considering this small margin, can also be unrelated at all to the undervolt.
Remember these tests were made on a slcd panel not amoled.
Did i say these tests were made on an i9023?
Tests made on slcd i9023.
TL;DR
THERE IS ANY NOTICEABLE BATTERY SAVING UNDERVOLTING THE LCD?NO
TEST 3
".. ye ye but removing lot of crap makes mah kernel faster!"
Stock Kernel vs Config tweaked (debug and crap removed) Kernel
Removed all possible debuggers, governors, tv tuners radio and all unused crap.
Let's see if it's really better.
- 0xBenchmark reds results are better.
Code:
Stock Kernel Cleaned Kernel
Linpack [COLOR="Red"]18,81[/COLOR] 18,72
C [COLOR="Red"]21,65[/COLOR] 21,54
FFT [COLOR="Red"]13,92[/COLOR] 13,65
JSOr 39,79 [COLOR="Red"]39,85[/COLOR]
MCi [COLOR="Red"]7,20[/COLOR] 6,77
Smm 17,80 [COLOR="Red"]17,82[/COLOR]
dLUmf [COLOR="Red"]29,61[/COLOR] 29,54
- AnTuTu reds results are better.
Code:
Stock Kernel Cleaned Kernel
RAM 260 [COLOR="Red"]261[/COLOR]
CPU Integer 416 416
CPU Float-Point [COLOR="Red"]106[/COLOR] 104
GFX 2D 278 278
GFX 3D 1115 [COLOR="Red"]1116[/COLOR]
TL;DR
REMOVING CRAP FROM KERNEL LIKE UNUSED DRIVERS, USELESS GOVERNORS, ALL DEBUGS, DO INCREASE KERNEL PERFORMANCE? NO
TEST 4
".. ye ye u fools seeking big numberz! Merging kernel with latest linux mainline makes teh battery drop fastah"
Kernel merged with linux 3.0.39 / 3.0.40 / 3.0.41
Same standard condition, let phone full drain with all 4 kernels in idle / airplane mode and standard usage.
RESULTS
After comparing 4 different kernel the battery stats were all the same, no weird wakelocks no battery drains. Also with standard usage, with data always on and few wifi and standard usage managed to reach 36 hours on same charge.
{
"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"
}

awesome work!

Good work.

I respect your work.

Hasn't there been devs around here that specifically said kernels don't actually affect battery drain? I'm not too familiar with all the technical stuff, so if anyone can explain exactly what the kernel is, that might help explain things even further. I do know, and everyone else (hopefully), more aggressive scaling can have an effect on battery life . Nice to see another test showing undervolting is pretty much not needed and isn't worth the instability it may cause. But hey, whatever floats your boat.

I find it rather strange that you claimed to use Google factory images on the 9023, but your battery screen shots show on screen buttons.

albundy2010 said:
I find it rather strange that you claimed to use Google factory images on the 9023, but your battery screen shots show on screen buttons.
Click to expand...
Click to collapse
"It depends on what the meaning of the word 'is' is."
"... I didn't inhale ..."

what CPU speed(high/low) are you benchmarking, I don't see it posted.

I'll just come out and say it: seems fishy.
I've seen you obsessed before at some kernel "myths" like thalamus claim that latest mainline updates were hurting performance/battery and to be honest I don't see in these studies a sufficient amount of rigor, objectivity and data to withdraw any conclusions except your clear agenda against some things that are said.
For example, some flaws:
1. 0xBenchmark and AnTuTu only measure one kind of performance, you may think you are gauging something when you're not.
2. A more recent toolchain supposedly provides improvements in other areas which weren't taken into account.
3. Such benchmarks have fluctuations, they are not particularly accurate.
4. The LCD undervolt test lacks data results (we are to believe your word?) and the methods chosen aren't good - too many parasite variables.
5. Again, vague information (you don't specify which debugging was removed). Not to mention some debug are proven to hurt performance like Frame Pointer. If you're going against theory, one more reason to be concise.
6. And once more, removing debug/crap should improve other things which were completely ignored (mm, pm, etc).
7. The last test just doesn't make sense, there are too many things involved to be that linear.
8. Why do the screenshots have battery % and the galaxy nexus keys if you were on OTA JRO0E?
Long story short, I can't really bring myself to take this too seriously as it lacks data and there's just too much hate undermining the credibility of the post. I should also mention that I don't have a position regarding each of those claims; I believe we should experiment, analyse, collect feedback and withdraw conclusions for everything but this just didn't convince me, especially when it comes to your neutrality. Thanks though.

chronophase1 said:
Hasn't there been devs around here that specifically said kernels don't actually affect battery drain?
Click to expand...
Click to collapse
I guess I know whom you are talking about, but may be you have not read his latest thread. lol...
When this dev released his .39 kernel, I asked how does it impact the battery. He shouted back at me in a rude voice saying Kernels doesn't impact battery and it is only the ROM. Fare enough.
But today he claims around that merging in to mainline from .31 to .39, .40 etc drains more battery and he is going back to .31 and says he has data etc.
I am glad this test has proved it actually doesn't matter.

anshumandash said:
When this dev released his .39 kernel, I asked how does it impact the battery. He shouted back at me in a rude voice saying Kernels doesn't impact battery and it is only the ROM. Fare enough.
But today he claims around that merging in to mainline from .31 to .39, .40 etc drains more battery and he is going back to .31 and says he has data etc.
Click to expand...
Click to collapse
I said generally, which is correct. The vast majority of the time the kernel has nothing to do with battery drain.
And yes, merging mainline does make a difference.
If you actually bother to read my blog post, you will see I don't actually mention battery drain at all as my reasons for ditching mainline updates.
Personally, I don't use the Nexus S enough to notice increases / decreases in drain, it's my development phone. However, quite a lot of users *have* told me that they have noticed improvements since I rolled back. Perhaps they are all wrong too?
It's not wakelocks, it's not obvious drain, it's subtle increases in drain which are impossible to track down.
However, In the case of the GNex when I merged to .40 I got 6% an hour drain, but when I went back to .31, I got less than 1% an hour drain. Merging back up again gives me the 6% an hour drain again with *nothing* else changed. It doesn't take a genius to figure out that it must be the kernel merging which has caused it, or is someone going to argue with that too? Lol.
I'm not entirely sure what myths have been busted here. It seems like a non kernel developer wasted their time to prove utterly nothing, which amuses me slightly. Do you honesty think I apply any modification, tweak or anything without testing the impact? If it makes no difference, it doesn't go in.
Removing unused stuff is simply to make the compile slightly quicker and the resulting zImage smaller. I don't believe there are any performance improvements to gain by doing that, but what is the point having junk built in which isn't needed?
As for removing all debugging, it's not a good idea, because how are you going to get a stacktrace if you panic? Again, that is something I won't do, and I know it makes little difference.
Anyway, if you want to test more accurate real world usage, use the 2D tests on 0xBench. They are CPU bound and they are greatly affected by small changes. Here are some I did a few weeks ago to test the best toolchain for the Nexus 7.
As you can see, there clearly is a difference between the speed of code that they produce. Raw speed is one thing, but graphics benchmarks more accurately represent real usage.
tl;dr: Ignore the agenda driven opinions and dubious results in the first post, they are meaningless.

simms22 said:
what CPU speed(high/low) are you benchmarking, I don't see it posted.
Click to expand...
Click to collapse
Performance, 1000.
albundy2010 said:
I find it rather strange that you claimed to use Google factory images on the 9023, but your battery screen shots show on screen buttons.
Click to expand...
Click to collapse
knzo said:
8. Why do the screenshots have battery % and the galaxy nexus keys if you were on OTA JRO0E?
Click to expand...
Click to collapse
If you read carefully i tested the battery with daily use aswell.
Quoting myself : " Also with standard usage, with data always on and few wifi and standard usage managed to reach 36 hours on same charge..
knzo said:
A more recent toolchain supposedly provides improvements in other areas which weren't taken into account.
Click to expand...
Click to collapse
If you read carefully the test is inspired by the Ezekeel one, that's why i used the same tools/approach. I do thrust his work more.
After tons of test around the web can pretty much assure you that toolchains may give something good compiling the OS ITSELF not the kernel.
knzo said:
The LCD undervolt test lacks data results (we are to believe your word?) and the methods chosen aren't good - too many parasite variables.
Click to expand...
Click to collapse
This is a rude attitude. Why don't you explain why the methods are wrong?
If after days of testing with almost just the lcd on, seeing wich charge lasted longer isn't good feel free to explain why.
Maybe i should've used a tester? I simply want to see if my phone last longer with lcd undervolt, simply.
About data results: the results is around 5%, would it better if i wrote how much every % lasted and then making simple math operations? No thanks.
i don't like to edit OP posts so i'll write it here. Quoting myself:
What i came up with is something like 5%, that means something like 10 more screen time with standard use, even less, and considering this small margin, can also be unrelated at all to the undervolt.
i meant 10 minutes more screen time
knzo said:
5. Again, vague information (you don't specify which debugging was removed). Not to mention some debug are proven to hurt performance like Frame Pointer. If you're going against theory, one more reason to be concise.
Click to expand...
Click to collapse
I disabled them one by one aswell and never noticed an increase performance. So your statement is wrong. I don't have the config anymore but for sure kernel, slub dm and cgroup subsys were disabled.
I develop my own rom and kernel just for myself. I make these tests for myself not to prove anything, i'm just sharing.
These tests took me one week to be made. Do you really think i would ruin them posting wrong informations or ruining my reputation?
Actually i was surprised by some of them.
You're welcomed to made them again or better since you didn't like the methods.

atl4ntis said:
If you read carefully the test is inspired by the Ezekeel one, that's why i used the same tools/approach. I do thrust his work more.
Click to expand...
Click to collapse
That single sentence sums up this entire thread.
Basically, you started these tests with an agenda which was to validate ezekeels tests and you 'proved' what you wanted to prove to fit your agenda.
Anyone can do that. It doesn't actually prove anything though, it just generates FUD. Congrats.

No just one test and just becouse similar test were made.
Go troll somewhere else.

atl4ntis said:
No just one test and just becouse similar test were made.
Go troll somewhere else.
Click to expand...
Click to collapse
I'm not trolling. I'm rightly questioning your extremely dubious results and the fact you have an agenda, which you clearly do. If you don't want people to question them, perhaps go and post them on Rootzwiki instead where they will be blindly accepted as gospel.
The definition of trolling is here.
Your post is simply confusing users and generating FUD, but perhaps *that* is your true agenda, is it not?

If you think these tests are wrong just provide some proof instead of offending people or talking about agenda or even worse reporting them as wrong because "other people said so" instead of testing them by yourself or saying some issues are not trackable.lol.
Some moderator should get rid of this thanks.

I think the OP has good intentions and had shown aptitude in collecting data, which deserves praise.
Just because something isn't perfect in the first attempt doesn't mean it deserves to be torn down.
Efforts like this need to be carefully nurtured because they go towards dispelling the prevalent aura of general guff that is spouted here in the development section on a daily basis.
To improve the study, make sure you are clear about the test conditions, and run the same test repeatedly until the mean and median converge to within some acceptable tolerance, e.g. 1%
You can then use standard deviation to make accurate statements about the data including its variability.
If you run multiple benchmarks you can later do regression testing to eliminate the tests that aren't correlated to the end result. You can combine the results of multiple benchmarks using the geometric mean.
If that's confusing, then I'll happily explain it in more detail.

Any more shenanigans, this thread gets locked. Either discuss the post like an adult or don't post.

You quoted my post about battery screens + on screen buttons but have not answered what is the deal with it.
What are those on screen buttons doing there on a official ROM for the 9023?

Already answered but maybe i wasn't clear enough.
I tested those kernel with daily usage aswell, that mean with mods apps and every crap i use daily. Those ss refers to the daily usage with my rom.

Related

[POLL][BOUNTY] DT2W battery drain fix like LG G2 (bump if you vote)

I'd like to gauge interest in setting a bounty for reverse engineering the touch screen controller chip sleep mode to reduce the screen off battery drain from 2%/hour to <1% like the LG G2 enjoys.
Correct me if I'm wrong, the stock LG G2 ROM accomplished this by sleeping [important chip that watches the capacitive touch screen input] setting it to interrupt mode or something. We use a different chip, with different I2C or SPI bus commands, and don't know which ones would get similar functionality, if any exists on the Synaptics S3350B and associated hardware.
Me, I'd put in like $5 or 10 if a solution with supporting documentation/explanation was found to implement in all ROMs.
Logistics:
Gauge interest with this poll
Gauge/solicit developer interest
Mod or trusted 3rd party like me receives donations to escrow paypal account
If have enough follow through, ...
Someone burns the night oil and finds a fix, publishes results, community of developers verify, submitter receives lump sum
If no fix found within (3 months), paypal donation returned.
thoughts?
If we don't have this so called chip how are we supposed to be able to developed something like it? Not to sound negative but I'm not sure I understand your thinking
Sent from my Nexus 5
Pretty sure it's a limitation on the digitizer.
psycho693 said:
If we don't have this so called chip how are we supposed to be able to developed something like it? Not to sound negative but I'm not sure I understand your thinking
Sent from my Nexus 5
Click to expand...
Click to collapse
it uses another one. We have to keep it awake to listen to the capacitive touch because we don't know how to send it to interrupt mode. Vs. the LG G2, you write a byte to it and it goes to interrupt mode. In my discussions with some of the DT2W firmware developers, that's the key difference.
If we had the full spec sheet on the chip this would be easy. No one has tracked that down.
Alternatively, someone with a little bit more embedded experience than I (ideally someone that has worked on S2W/DT2W) applies intuition and pokes at registers on the SPI address for this chip to see what happens. That's tough, hence the bounty.
well, flar2 apparently has lowered the drain on d2w on his elementalX kernel, but is still testing it. you may want to check out the thread and leave a question there.
Sent from my Nexus 5 using Tapatalk
my battery drains less then 1% an hour as it is, without anything like this..
{
"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"
}
simms22 said:
my battery drains less then 1% an hour as it is, without anything like this..
Click to expand...
Click to collapse
Nah, not with S2W or DT2W enabled.
rancur3p1c said:
Nah, not with S2W or DT2W enabled.
Click to expand...
Click to collapse
ah.. i missed that part :angel:
simms22 said:
ah.. i missed that part :angel:
Click to expand...
Click to collapse
lol!
simms22 said:
my battery drains less then 1% an hour as it is, without anything like this..
Click to expand...
Click to collapse
rom?:cyclops:
ludorexmundi said:
rom?:cyclops:
Click to expand...
Click to collapse
lichtis rastakat. but rom nor kernel dont give you extra battery life. battery life is all dependant on your personal usage, your personal setup, your apps installed, and the quality of your phone/data signal.
On cyanogenmod I get less 1% drain per hour, more like 0.3%. I get an average of 1h screen time for every 20% battery lost with wifi on location set to power save, mobile data on H+ always on. From 7 in the morning till 0.30 at night I get to about 35% battery left with is fine and about 4h+ SOT, however I do not use my phone for gaming, mostly texting and whatsapp, 2h of YouTube videos and one hour of Web browsing, and like 40min of voice calls.
Sent from my Nexus 5 using XDA Premium 4 mobile app
Pezmet said:
On cyanogenmod I get less 1% drain per hour, more like 0.3%. I get an average of 1h screen time for every 20% battery lost with wifi on location set to power save, mobile data on H+ always on. From 7 in the morning till 0.30 at night I get to about 35% battery left with is fine and about 4h+ SOT, however I do not use my phone for gaming, mostly texting and whatsapp, 2h of YouTube videos and one hour of Web browsing, and like 40min of voice calls.
Sent from my Nexus 5 using XDA Premium 4 mobile app
Click to expand...
Click to collapse
Guys this is about double tap to wake/sweep to wake
simms22 said:
kernel dont give you extra battery life.
Click to expand...
Click to collapse
yeah that's not true. For one custom kernels have higher levels of compiler optimzation that result in the processor executing fewer commands; to governor use and limiting max screen off frequency to 600mhz, locking cores offline all the time; there are plenty of reasons why custom kernels have better battery life that everyone sees.
rancur3p1c said:
yeah that's not true. For one custom kernels have higher levels of compiler optimzation that result in the processor executing fewer commands; to governor use and limiting max screen off frequency to 600mhz, locking cores offline all the time; there are plenty of reasons why custom kernels have better battery life that everyone sees.
Click to expand...
Click to collapse
see thats where you are mistaken. as many people that say that they get great battery life on a specific kernel, there are just as many, if not more, that say the opposite. you can get great battery life on ANY kernel. on every single custom kernel here on xda there will be people that claim that it gives great battery life, and many more that dont claim it. battery life is mostly about your personal use, your setup, and the quality signal you get. everything else makes very little difference. and this is from experience. as a hobby, i test kernels. not just test like most people on xda, but test them thoroughly. i work for trinity kernel, and have for years. i test all the kernels, to see where trinity has to go. i can get wonderful battery life, on any kernel. its all because of my setup and great quality signal reception, not because of the kernels.
now pick a kernel, any kernel, no matter how over developed it is(yes, many kernels are overly developed here and dont actually gain anything considerable), and ill beat it in battery life. what the issue actually is though is that you believe that a magic battery life pill exists, while im more of a realist an know that no such thing exists.
ludorexmundi said:
rom?:cyclops:
Click to expand...
Click to collapse
Stock can achieve that without tweaks
Mawtin said:
Stock can achieve that without tweaks
Click to expand...
Click to collapse
right, there arent any tweaks needed for that at all. just a proper setup will do
simms22 said:
ah.. i missed that part :angel:
Click to expand...
Click to collapse
I was about to chime in with same comment....thanks for being first! . Lol
rancur3p1c said:
it uses another one. We have to keep it awake to listen to the capacitive touch because we don't know how to send it to interrupt mode. Vs. the LG G2, you write a byte to it and it goes to interrupt mode. In my discussions with some of the DT2W firmware developers, that's the key difference.
If we had the full spec sheet on the chip this would be easy. No one has tracked that down.
Alternatively, someone with a little bit more embedded experience than I (ideally someone that has worked on S2W/DT2W) applies intuition and pokes at registers on the SPI address for this chip to see what happens. That's tough, hence the bounty.
Click to expand...
Click to collapse
It may be a hardware limitation. The LCD panel and the touch screen share a power source. It's no problem to set IRQ wake for the touchscreen, the problem is the interrupts are not triggered unless the touchscreen is powered on. I've poked around quite a bit trying to figure out a way to reduce the drain, but no luck.
You are correct that the g2 has a special bit that can be set for dt2w, but even if this exists on the n5, it would b be like finding a needle in a haystack without the spec sheet. Besides, I've added doubletap to several devices and it was sufficient to use IRQ wake. No special bit required.
Note the red line going from PMIC_1 to the LCD panel. The Touch panel shares power with the LCD panel, and it's routed through the panel on its way to the touch sensor. The only other connection to Touch is the i2c bus. Without power, it can't sense touches and trigger the interrupt.
The way I got it to detect touches during suspend was to modify the touch panel driver so it doesn't send the command to power off the panel when sweep2wake or doubletap2wake are enabled: https://github.com/flar2/ElementalX-N5/commit/03411f9859369bf83324716730bccb77014b31a7
Since then, I've messed around with the regulators, off cmds and clocks trying to reduce the power consumption, but drain remains at about 2.5% per hour with the wake features enabled.

SetCPU

Anyone using SetCPU to throttle the device and get tolerable battery life?
I wish. Not on Verizon!
Bump!!
I just started using trickster mod and set max frequency to about 2.0GHz. The battery life has increased with no lag. On Lollipop.
obtained said:
I just started using trickster mod and set max frequency to about 2.0GHz. The battery life has increased with no lag. On Lollipop.
Click to expand...
Click to collapse
What did you have to do to get trickster working? Just download it? It works straight from the market? H3lp!
zgroten said:
What did you have to do to get trickster working? Just download it? It works straight from the market? H3lp!
Click to expand...
Click to collapse
Yeah just download it. You will have to download a busybox installer if you don't have busybox installed as well.
Awesome thanks. What mods and settings have you found useful?
Anyone has had success tweaking new Moto X with SetCPU?
martinezma99 said:
Anyone has had success tweaking new Moto X with SetCPU?
Click to expand...
Click to collapse
SetCPU caused major lag for me personally.. I switched to trickstermod and all seems fine so far
I'm using No Frills, working great across 422MHz - 1GHz ( hardcore testing )
I have never noticed any battery improvement from underclocking, and correct me if I'm wrong but I thought the general consensus (among those actually informed) was that it doesn't actually help because it causes the processor to take longer to complete tasks, canceling out any actual gains.
_MetalHead_ said:
I have never noticed any battery improvement from underclocking, and correct me if I'm wrong but I thought the general consensus (among those actually informed) was that it doesn't actually help because it causes the processor to take longer to complete tasks, canceling out any actual gains.
Click to expand...
Click to collapse
After testing for a day almost, I'll say that's crap.. I'm running this guy comfortably at 0.8GHz with ultra small lags, see the battery for yourself. I used to get 4-4.5 hours of max screen on running it untampered and now this damn battery won't die!
ManiacShri said:
After testing for a day almost, I'll say that's crap.. I'm running this guy comfortably at 0.8GHz with ultra small lags, see the battery for yourself. I used to get 4-4.5 hours of max screen on running it untampered and now this damn battery won't die!
Click to expand...
Click to collapse
Yeah I dunno man, I'm not underclocked and I can probably get 8+hrs SOT as well if all I'm doing is web browsing and reading. I mean, I average around 6-7hrs SOT on mine (still on 4.4.4) depending on what I'm doing. Everybody's usage patterns are different, which is why I don't trust SOT screen shots. Plus, there is nothing in your second screenshot that shows any actual usage. You have screen, cell standby, wifi, android system... I'm not seeing any apps in there which leads me to believe that isn't a heavy usage scenario for you.
And again, I have underclocked before with many, many devices and haven't seen any actual gains. Also, I have yet to see any actual empirical evidence that shows underclocking helps. Couple that with the fact that many people who are well versed in this stuff (including the CyanogenMod team) have come forward and said that any gains from underclocking are placebo due to the reason I stated before, and I am HIGHLY skeptical.
_MetalHead_ said:
Yeah I dunno man, I'm not underclocked and I can probably get 8+hrs SOT as well if all I'm doing is web browsing and reading. I mean, I average around 6-7hrs SOT on mine (still on 4.4.4) depending on what I'm doing. Everybody's usage patterns are different, which is why I don't trust SOT screen shots. Plus, there is nothing in your second screenshot that shows any actual usage. You have screen, cell standby, wifi, android system... I'm not seeing any apps in there which leads me to believe that isn't a heavy usage scenario for you.
And again, I have underclocked before with many, many devices and haven't seen any actual gains. Also, I have yet to see any actual empirical evidence that shows underclocking helps. Couple that with the fact that many people who are well versed in this stuff (including the CyanogenMod team) have come forward and said that any gains from underclocking are placebo due to the reason I stated before, and I am HIGHLY skeptical.
Click to expand...
Click to collapse
The reason there's no specific apps you see is because I did what you do in a typical battery test, gaming a bit, reading ,browsing and all the others... So the individual. consumption is well b below 5%.. But yeah even without underclocking i have gotten even 5.5hours which is not bad at all
_MetalHead_ said:
Yeah I dunno man, I'm not underclocked and I can probably get 8+hrs SOT as well if all I'm doing is web browsing and reading. I mean, I average around 6-7hrs SOT on mine (still on 4.4.4) depending on what I'm doing. Everybody's usage patterns are different, which is why I don't trust SOT screen shots. Plus, there is nothing in your second screenshot that shows any actual usage. You have screen, cell standby, wifi, android system... I'm not seeing any apps in there which leads me to believe that isn't a heavy usage scenario for you.
And again, I have underclocked before with many, many devices and haven't seen any actual gains. Also, I have yet to see any actual empirical evidence that shows underclocking helps. Couple that with the fact that many people who are well versed in this stuff (including the CyanogenMod team) have come forward and said that any gains from underclocking are placebo due to the reason I stated before, and I am HIGHLY skeptical.
Click to expand...
Click to collapse
You can be sceptical because you should be, the cpu throttles already and only uses max speed when needed. With light usage like me (WhatsApp, telegram, 20m phone calls and some mails a day) I get 4 days on 4g and wifi. (I turn my phone on flight mode during the night) I've also disabled moto application (voice is useless in Dutch/not working) The biggest energy consumer is the screen so change your wallpaper to a dark one and close unused background apps (moto voice, google now, virus scanner and all other useless apps. If you use all the social media out there on your phone, no battery capacity is sufficient. I could limit my max cpu speed because the cpu governor probably throttles to max even when launching undemanding apps, but I would probably won't notice any savings.

[Q] Recommended ROM, Kernel, governor, etc

Hey all. Which ROM, kernel, and governor do you recommend I use on my N6? There are so many options that I figured I'd ask. Currently running Benzo ROM and the kernel it came with. Thanks!
"Best" ROM.
There is no such thing as a best ROM.* The question itself is ambiguous.* "Best" is obviously a subjective term.
What I want from a ROM may well differ from what you want from a ROM, ergo - what is best for me could be worst for you.
If you are asking what the most popular ROMs are, or which ROMs people are using, you can see which threads stay around on the first few pages (and have the most posts) in the Android Development or Original Android Development forums. You can also see what other people are running by reading the What are YOU running on your Nexus 6??? thread.
If you are asking which is the most stable, being a Nexus device - they're all pretty stable.
If you are asking which is best on Battery, ROMs only affect battery if they have a feature that is badly coded.* You will likely be able to read about this in the ROM threads.* ROMs do not impact battery life.* The only impact to battery life are your apps, your settings, how you use the phone and mostly, environmental issues such as Phone Signal.
For tips about improving battery life, please read [Battery Life Help] Troubleshoot battery issues here!
"Best" Kernel
There is no such thing as the "Best" kernel.* What we all want from a kernel is different. Again, many people have the misconception that Kernels affect battery life.* Let's get this cleared up.* Although Kernel devs will build in optimisations and efficiencies that will improve battery life, these are very, VERY tiny...and if 1 kernel has these optimisations, they likely all have.
People will often say "Kernel x is better than kernel y for battery life".* This is actually wrong.* Kernels respond to user settings. Setting up the governor to favour either battery life or performance is simple enough to do, you just have to do some learning.* The reason people think Kernel x is better than y is because developers set their kernels up with their preferred governor settings.* This is what we refer to as out-of-the-box settings.* The out-of-the-box settings for kernel x may well produce better battery results than the out-of-the-box settings for kernel y, which favour performance.* The fact is, you as the user have the ability to tune kernel x or y to perform the same, be that battery or performance - so start learning how to do this yourselves - that way, you can choose the kernel based on the FEATURES you want, and not the fictional performance benefits of one kernel over another.
Hope this helps
rootSU said:
"Best" ROM.
There is no such thing as a best ROM.* The question itself is ambiguous.* "Best" is obviously a subjective term.
What I want from a ROM may well differ from what you want from a ROM, ergo - what is best for me could be worst for you.
If you are asking what the most popular ROMs are, or which ROMs people are using, you can see which threads stay around on the first few pages (and have the most posts) in the Android Development or Original Android Development forums. You can also see what other people are running by reading the What are YOU running on your Nexus 6??? thread.
If you are asking which is the most stable, being a Nexus device - they're all pretty stable.
If you are asking which is best on Battery, ROMs only affect battery if they have a feature that is badly coded.* You will likely be able to read about this in the ROM threads.* ROMs do not impact battery life.* The only impact to battery life are your apps, your settings, how you use the phone and mostly, environmental issues such as Phone Signal.
For tips about improving battery life, please read [Battery Life Help] Troubleshoot battery issues here!
"Best" Kernel
There is no such thing as the "Best" kernel.* What we all want from a kernel is different. Again, many people have the misconception that Kernels affect battery life.* Let's get this cleared up.* Although Kernel devs will build in optimisations and efficiencies that will improve battery life, these are very, VERY tiny...and if 1 kernel has these optimisations, they likely all have.
People will often say "Kernel x is better than kernel y for battery life".* This is actually wrong.* Kernels respond to user settings. Setting up the governor to favour either battery life or performance is simple enough to do, you just have to do some learning.* The reason people think Kernel x is better than y is because developers set their kernels up with their preferred governor settings.* This is what we refer to as out-of-the-box settings.* The out-of-the-box settings for kernel x may well produce better battery results than the out-of-the-box settings for kernel y, which favour performance.* The fact is, you as the user have the ability to tune kernel x or y to perform the same, be that battery or performance - so start learning how to do this yourselves - that way, you can choose the kernel based on the FEATURES you want, and not the fictional performance benefits of one kernel over another.
Hope this helps
Click to expand...
Click to collapse
this. and its very well said at that.
Thanks for the reply root.... However, nowhere did I ask what the "best" ROM is, I totally disagree that the stability will be equivalent because there are AOSP ROMs out there that are all in alpha for 5.1, as well as stock ROMs that may have certain rather experimental things, with nightlies being released. I guess the goal of my query was to find out what people have set some of their variables to (ROM, settings, kernel, kernel settings) to get better battery life with minimal performance drop.
YevOmega said:
nowhere did I ask what the "best" ROM is
Click to expand...
Click to collapse
Its my generic answer. Copy and paste. But you asked for recommendations and that gets the same answer because other than semantics, there's no difference in the question.
YevOmega said:
I totally disagree that the stability will be equivalent because there are AOSP ROMs out there that are all in alpha for 5.1, as well as stock ROMs that may have certain rather experimental things, with nightlies being released
Click to expand...
Click to collapse
That's you're prerogative but I'm not really talking about feature sets. That aspect of the answer is more generic, usually for people coming from HTC or Samsung where "camera sux" because the binaries were never released.
YevOmega said:
I guess the goal of my query was to find out what people have set some of their variables to (ROM, settings, kernel, kernel settings) to get better battery life with minimal performance drop.
Click to expand...
Click to collapse
Performance uses power, which uses battery. There is no way to get more battery life back than the amount of performance you're willing to sacrifice. Its a direct correlation.
I can tell you my settings, but they're not magic. Everything I gain in battery life, I lose in performance.
Edit. I think a good thing to do would be to pick a kernel based on features, then speak to users of that kernel to see how they set it up
@YevOmega, please check post #5 of the stickied Q&A thread for details on this question.
Sent from my Nexus 6 using Tapatalk
YevOmega said:
Hey all. Which ROM, kernel, and governor do you recommend I use on my N6? There are so many options that I figured I'd ask. Currently running Benzo ROM and the kernel it came with. Thanks!
Click to expand...
Click to collapse
The answer depends on how many ROM/kernel/governor combinations there, because that is how many answers you are likely to get. Stock based are usually the most stable, and everything is likely to work the way its's supposed to while CM/AOSP ROMs will have more cusomization and tweaks but are often troubled by instability or things that are broken. Me, I'm on Stock rooted 5.1 for now, until I flash something else. What is it that's important to you when it comes to the ROM?
I want it to be crisp most importantly, but it's nice to have things like PIE. Minimal bloatware, because yes there are still some system apps that I don't want but can't uninstall without jumping through hoops.
rootSU said:
Its my generic answer. Copy and paste. But you asked for recommendations and that gets the same answer because other than semantics, there's no difference in the question.
That's you're prerogative but I'm not really talking about feature sets. That aspect of the answer is more generic, usually for people coming from HTC or Samsung where "camera sux" because the binaries were never released.
Performance uses power, which uses battery. There is no way to get more battery life back than the amount of performance you're willing to sacrifice. Its a direct correlation.
I can tell you my settings, but they're not magic. Everything I gain in battery life, I lose in performance.
Edit. I think a good thing to do would be to pick a kernel based on features, then speak to users of that kernel to see how they set it up
Click to expand...
Click to collapse
Again, please read my statement. I understand that I can't just get better battery life out of thin air. I said minimal performance loss, not none. There is a variety of optimizations and my hope is that there are combinations that will improve my battery life without compromising performance too much.
YevOmega said:
Again, please read my statement. I understand that I can't just get better battery life out of thin air. I said minimal performance loss, not none. There is a variety of optimizations and my hope is that there are combinations that will improve my battery life without compromising performance too much.
Click to expand...
Click to collapse
You've misinterpreted what I've said. There is no minimal. Its loss or its not. Its an direct correlation. If you want to save x battery you must sacrifice y performance. There's no magic, smaller z performance that still gets you x battery.
I didn't say you wanted to get battery out of thin air (perhaps you need to read my statement). I'm saying you want to get battery out of medium air, but I'm telling you that it only comes out of thick air.
Let me try something else.
To increase battery by 6, you must decrease performance by 6. There is no magic setting that will allow you to get 6 battery out of 3 performance. If you only want to sacrifice 3 performance, you'll only get 3 battery.
Break it down. What uses battery most?
-Screen
-Radio
-CPU
So forgetting screen and radio which are out of the scope of this thread, let's look at CPU.
CPU voltage is controlled by the kernel. The kernel has a table that has a predetermined amount of voltage for every clock cycle step. As you can see here.
{
"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"
}
So how you save power on CPU is to either prevent certain CPU frequency being used, limit the time it is being used for or try to not need that frequency. This will save 6 battery, but will lose 6 performance because in each case that battery is being saved, the CPU frequency is not being used. Its a relative battery to performance ratio.
So I can list all the kernel settings that will save you battery, but they all have an equal performance hit - which is what I'm trying to explain.
Alright thanks. Go for it. I have vindicator kernel, as a reminder, so if you don't mind listing some settings, that would be nice ?. Thanks!
YevOmega said:
Alright thanks. Go for it. I have vindicator kernel, as a reminder, so if you don't mind listing some settings, that would be nice ?. Thanks!
Click to expand...
Click to collapse
I use elementalx, but here's some things I like to do.
1 set up threshold higher (99) CPU ramps up at 99% load
2 hot plugging set to 2 cores offline when not in use
3 turn off all touch boosts
rootSU said:
I use elementalx, but here's some things I like to do.
1 set up threshold higher (99) CPU ramps up at 99% load
2 hot plugging set to 2 cores offline when not in use
3 turn off all touch boosts
Click to expand...
Click to collapse
How much SOT do you get, and what carrier are you on?
YevOmega said:
How much SOT do you get, and what carrier are you on?
Click to expand...
Click to collapse
It depends. At home, I can get 6 or maybe 7. On a work day, perhaps 4 or rarely 5.
I live in the UK so not sure the Carrier matters.
YevOmega said:
How much SOT do you get, and what carrier are you on?
Click to expand...
Click to collapse
rootSU said:
It depends. At home, I can get 6 or maybe 7. On a work day, perhaps 4 or rarely 5.
I live in the UK so not sure the Carrier matters.
Click to expand...
Click to collapse
Carrier only matters if you are comparing signal strength in regards to the same bands. Kernels can also allow you to undervolt if you forgot, there are other things we can do such as drop voltage on certain parts of the device, though this can cause instability.
Optimizations on ROM and kernel side can reduce the overhead on the CPU and we can also increase throughput on several aspects of the device such as memory and BUS. Other things that "may" increase battery life could be removing code for rotational storage and using flash based alternatives/optimizing it for non-rotational storage.
Thanks. And yeah, I know why carrier matters

Benchmark is bad on RR

I am running RR and am getting under 40k with antutu benchmark
i am wondering why it is so low and if someone can help me.
i am running xposed with the following modules
amplify - not blocking anything though yet
app settings - used for a couple apps that need 640 dpi to operate since i am running at 480 dpi
blacklist
boot manager
google offline voice
greenify
iFont
unbelovedhosts
xinstaller
xprivacy - not blocking anything with it yet
youtube adaway
is there any apps that would cause such a low performance. i tried installing the kernel by @BayButcher and modifying it to his battery suggestions and i am about to try his performance setup for his kernel to see if that does anything. but i was getting near 50k with stock so i figure i should be better with CM12.1 or RR.
johnbravado said:
I am running RR and am getting under 40k with antutu benchmark
i am wondering why it is so low and if someone can help me.
i was getting near 50k with stock so i figure i should be better with CM12.1 or RR.
Click to expand...
Click to collapse
In thermal settings, you will have to set the temperature higher. Intellithermal aggressively throttles the CPU. I set it for 65 degrees for just "normal" operations, because it seems to start throttling 10 degrees sooner than what you set.
So for a benchmark that makes you feel better, set the degree limit even higher. How high is however high you are willing to risk... Also, before you take the benchmark tests, stick the phone in the freezer with a ziplock bag for a few minutes to let it really cool down. Then start the test, put it back in the freezer while the test is running. The phone will run faster longer while cool, and won't throttle the CPU.
I mean, benchmarks are for bragging rights, yes?
When you finish the benchmarks test, then put the degree limit back down to like 65 degrees for everyday use.
______
50,000 -- here's Antutu results at 65 degrees throttling and freezer trick. I could get higher if I set my thermal setting even higher -- which would be safe temporarily as long as I chilled the phone first IN the freezer and kept it there during testing.
{
"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"
}
ChazzMatt said:
In thermal settings, you will have to set the temperature higher. Intellithermal aggressively throttles the CPU. I set it for 65 degrees for just "normal" operations, because it seems to start throttling 10 degrees sooner than what you set.
So for a benchmark that makes you feel better, set the degree limit even higher. How high is however high you are willing to risk... Also, before you take the benchmark tests, stick the phone in the freezer with a ziplock bag for a few minutes to let it really cool down. Then start the test, put it back in the freezer while the test is running. The phone will run faster longer while cool, and won't throttle the CPU.
I mean, benchmarks are for bragging rights, yes?
When you finish the benchmarks test, then put the degree limit back down to like 65 degrees for everyday use.
Click to expand...
Click to collapse
I agree benchmarks are for bragging, but they are still a benchmark. if stock out of the box i can get a 48-49k benchmark and i do some mods and can get a 55k benchmark that is for bragging. but if it drops to 40k something is broke. . I am going to try the increased temperature and run another test. the performance setup only got me to 41k.
i am really confused now. i reinstalled computerfreeks ROM which before bootloader unlock gave me 50k and now it gave me 35k. i am going back to RR because I like it more but need to find out what caused performance issues.
johnbravado said:
I agree benchmarks are for bragging, but they are still a benchmark. if stock out of the box i can get a 48-49k benchmark and i do some mods and can get a 55k benchmark that is for bragging. but if it drops to 40k something is broke. . I am going to try the increased temperature and run another test. the performance setup only got me to 41k.
i am really confused now. i reinstalled computerfreeks ROM which before bootloader unlock gave me 50k and now it gave me 35k. i am going back to RR because I like it more but need to find out what caused performance issues.
Click to expand...
Click to collapse
My main point is while the custom ROM may give you increased performance and options, the default kernel thermal throttling points (for frequency as well as for core -- two separate settings) may be sabotaging benchmark tests.
I increased frequency point to 65 degrees, because that REALLY means 55 degrees, for some reason.
I didn't even touch the "core" thermal setting...
And my 50,000 result was on stock 640 dpi.
Cm12 is the same way. Low benchmark. I'm saying with exposed and stock
Sent from my XT1254 using Tapatalk
tecsironman said:
Cm12 is the same way. Low benchmark. I'm saying with exposed and stock
Sent from my XT1254 using Tapatalk
Click to expand...
Click to collapse
I'm on CM, but with @baybutcher27 kernel. See my benchmark results above.
ChazzMatt said:
I'm on CM, but with @baybutcher27 kernel. See my benchmark results above.
Click to expand...
Click to collapse
Yeah, but you obtained those results by tweaking the phone and subjecting it to non-real world conditions. I can get 52K on the stock rom by doing absolutely nothing and having a bunch of xposed modules active, and I think that's the issue that the OP is concerned about. The relationship between benchmark scores and real world performance is debatable, but I think he has a valid point.
johnbravado said:
I agree benchmarks are for bragging, but they are still a benchmark. if stock out of the box i can get a 48-49k benchmark and i do some mods and can get a 55k benchmark that is for bragging. but if it drops to 40k something is broke. . I am going to try the increased temperature and run another test. the performance setup only got me to 41k.
i am really confused now. i reinstalled computerfreeks ROM which before bootloader unlock gave me 50k and now it gave me 35k. i am going back to RR because I like it more but need to find out what caused performance issues.
Click to expand...
Click to collapse
The caused of the performance issues is simple heat vs battery level.
Has the device get hot it get low score because more throttle down, has the battery gets low same ting there is a setting to lock the clock under a frequency if the battery get too low, it starts only at 20%.
very simple table:
Code:
thresholds 80 85 90
thresholds_clr 79 84 89
actions cpu cpu cpu
action_info 2265600 1958400 1728000
this is editable this is in the file
system/etc/thermal-engine-quark.conf
I never try to edit it but you can.
But in my test I find that there is more then this file make the performance get low, and for me is just the battery as it get low the performance get too, but the battery is only responsible for some very little % of this the heat is 90 to 95% responsible.
So be aware that you problem is more heat you heat your device installing ROM make and restore backup don't give time to the device cools down and test.
No matter the ROM you try that will be the result, unless you can set the temp threshold higher like in a custom Kernel.
baybutcher27 said:
The caused of the performance issues is simple heat vs battery level.
Has the device get hot it get low score because more throttle down, has the battery gets low same ting there is a setting to lock the clock under a frequency if the battery get too low, it starts only at 20%.
very simple table:
Code:
thresholds 80 85 90
thresholds_clr 79 84 89
actions cpu cpu cpu
action_info 2265600 1958400 1728000
this is editable this is in the file
system/etc/thermal-engine-quark.conf
I never try to edit it but you can.
But in my test I find that there is more then this file make the performance get low, and for me is just the battery as it get low the performance get too, but the battery is only responsible for some very little % of this the heat is 90 to 95% responsible.
So be aware that you problem is more heat you heat your device installing ROM make and restore backup don't give time to the device cools down and test.
No matter the ROM you try that will be the result, unless you can set the temp threshold higher like in a custom Kernel.
Click to expand...
Click to collapse
I am willing to buy into the heat issue since i am changing between ROMS and such so i got a few things against me
1) chargin phone
2) restoring different ROM
3) not giving adequate time to cool
i am going to let eveything settle a little with RR and retest. i am using your kernel and used your performance recomendations. i am not a gamer i just do not like clicking a button and waiting. or wa5tching me type faster than the phone can output to a field
TheSt33v said:
Yeah, but you obtained those results by tweaking the phone and subjecting it to non-real world conditions. I can get 52K on the stock rom by doing absolutely nothing and having a bunch of xposed modules active, and I think that's the issue that the OP is concerned about. The relationship between benchmark scores and real world performance is debatable, but I think he has a valid point.
Click to expand...
Click to collapse
Yes exactly. i did nothing between test a and test b except unlock bootloader and install RR, CM12.1. and yet i got a huge performance drop. but same issue was with the computerfreek version which is like stock. so i am thinking it has to do with something else not playing friendly.
i would assume if with stock out the box i could pull 50k then with RR and a kernel boost i should be able to pull 52k.
TheSt33v said:
Yeah, but you obtained those results by tweaking the phone and subjecting it to non-real world conditions. I can get 52K on the stock rom by doing absolutely nothing and having a bunch of xposed modules active, and I think that's the issue that the OP is concerned about. The relationship between benchmark scores and real world performance is debatable, but I think he has a valid point.
Click to expand...
Click to collapse
Tweaking the phone? Mostly all I did was change the CUSTOM kernel throttling conditions. When you root your phone, from that point on you are TWEAKING your phone. Give me a break.
Yeah, I also cooled the phone, so that is non-real world conditions, but changing the quite arbitrary throttling point is quite logical when it's been PROVEN that throttling point affects performance and actually starts throttling even sooner!
Read this.
xxspookyxx said:
I think change gorvenor to conservative dont solve the problem.
To me dont make any difference, the better Gorvernor works well with my Moto maxx are Interactive_pro.
To getter better performance of my phone, i need change Thermal Frequency Throttle to 80º C, most custom roms use Intellithermal method, and they are very sensivity to temperature.
If you choose 80º C, they will reach in max 70º to begin the throttle the processor.
Stock Custom Roms are 60ºC in Thermal Frequency Throttle, i change to 80ºC and my phone are very fast now.
Click to expand...
Click to collapse
So, on your custom ROM kernel, change your governor, change your throttle points.
OK i am satisfied here. i unplugged the phone and benchmarked at 50.7k. apparently charging the phone cripples its productivity. i did not think it would cripple it that bad. and if this is a known thing i apologize for the posting.
i can now go back to enjoying the awesome RR and tweaking it.
ChazzMatt said:
I'm on CM, but with @baybutcher27 kernel. See my benchmark results above.
Click to expand...
Click to collapse
Yes sir I saw that.. Maybe I'll give cm12 another shot with butchers kernel. I'm at work ATM with just my otg and little time.
Sent from my XT1254 using Tapatalk
My Turbo that I haven't bootloader unlocked. All stock with some stuff disabled in settings.
ChazzMatt said:
Tweaking the phone? Mostly all I did was change the CUSTOM kernel throttling conditions. When you root your phone, from that point on you are TWEAKING your phone. Give me a break.
Click to expand...
Click to collapse
Yeah, okay, so all you did was change some throttling conditions. But you know what I had to do in order to get a higher benchmark than the one you got? Nothing. At room temperature. That's a problem. One of the main purposes of unlocking one's phone is increasing performance, not reducing it.
If the problem is simply an incorrectly set throttle point, all you should have to do to get 52K at room temperature is change the throttle point high enough, right? No cooling required. So let's see that. If you can get 52K without melting your phone in a normal usage environment, then good. No worries. Mine didn't even get hot during the benchmark, so you should be able to easily exceed my 52K score. If not, there is an issue that I think should be addressed.
---------- Post added at 09:33 PM ---------- Previous post was at 09:24 PM ----------
johnbravado said:
OK i am satisfied here. i unplugged the phone and benchmarked at 50.7k. apparently charging the phone cripples its productivity. i did not think it would cripple it that bad. and if this is a known thing i apologize for the posting.
i can now go back to enjoying the awesome RR and tweaking it.
Click to expand...
Click to collapse
Charging = heat generation = throttling, so yes, that would do it. I'm still curious to see if CM and this custom kernel can break 52K though. I'm waiting for CM13 before I make the switch, but I might reconsider if I have to sacrifice some performance. The technical power of this phone was half the reason I bought it.
Stock cm12.1 benched 48,000 screen on.
Screen off it was 43,000.
I'm happy with it. I also think the battery life is mildly better.
Sent from my DROID Turbo using XDA Free mobile app
ChazzMatt said:
Tweaking the phone? Mostly all I did was change the CUSTOM kernel throttling conditions. When you root your phone, from that point on you are TWEAKING your phone. Give me a break.
Yeah, I also cooled the phone, so that is non-real world conditions, but changing the quite arbitrary throttling point is quite logical when it's been PROVEN that throttling point affects performance and actually starts throttling even sooner!
Read this.
So, on your custom ROM kernel, change your governor, change your throttle points.
Click to expand...
Click to collapse
Can I just flash butchers kernel and not mess with it? Is it good to go after flash or will I have to play around with settings?
Sent from my XT1254 using Tapatalk
After you flash it there is nothing saying you need to change anything.
I would imagine he has the presets at what he thinks is best.
Sent from my DROID Turbo using XDA Free mobile app
Just FYI, I'm running computerfreek's latest unlocked ROM with xposed and about 10 modules. No changes to thermal setting (don't know the default for this ROM either), and just have it sitting on the desk. I'm also, getting unbelievable battery life with servicely, amplify.
tecsironman said:
Can I just flash butchers kernel and not mess with it? Is it good to go after flash or will I have to play around with settings?
Sent from my XT1254 using Tapatalk
Click to expand...
Click to collapse
http://forum.xda-developers.com/moto-maxx/development/kernel-bhb27-kernel-t3207526
the writer of the kernel has said it needs some adjustments since he does not turn them on by default. but he gives 2 setups, 1 for battery and 1 for performance on post #2. i just set it up as per post #2 until i understand more what is going on with the kernel.

[EXPERIMENTAL] Throttling and it's caveats

While being owner of my Redmi Note 7, which definitely has some issues, I started experimenting with it.
Throttling is a common term amongst electronics and such. It's a way to safe our precious silicon from eventually cooking under high temperatures. There're a lot of issues involved here. Let me explain:
Xiaomi's belief is maximizing battery life while forgetting about optimal device's performance. Here's the issue, though - neither of which are properly calibrated. For a battery of this size, you would expect something better, but it's not. Looking through system files, there are several thermal config files, which simply tells to the thermal-engine binary at what state device should throttle. Sounds like a good idea, right? Yeah, not quite.
Upon opening these exact config files you see several lines, which describe temperature thresholds and which frequency device should start to use under it's determined state. There's more to this - GPU gets less voltage when it's getting warmer, thus reducing the ability for GPU to reach higher power levels. Everything would be great, if the device wouldn't start to throttle it's clocks at 45° degrees, which is basically nothing.
After seeing device's guts it's safe to state, that our phone has proper cooling done to it. Sure, it's basic, but with that much space and the price this phone is built for, you couldn't ask for better cooling.
So, what can you do to it?
There're several solutions.
-Deleting thermal-engine entirely and forgetting about the problem. Here's the thing, though - it's really dangerous, because not only you lose the ability to monitor your battery's temperature, you could cook your phone while charging. Charging throttling is a thing too, when phone slows down pulling current from a wall as soon as you are doing something intensive. It's a dirty fix with a lot of risks involved.
-Modifying thermal config tables. This way you could ensure device's safety while having optimal device's performance.
-Mounting hugeass cooler at the back of your phone or better yet - mounting cooler straight to the cpu silicon. Jokes aside, this is not recommended at all, unless you're insane and don't care about sacrificing device's size.
I'm actually thinking about making completely custom thermal tables and make them compatible with Magisk. Even if Magisk method wouldn't ensure enough confidence, I will make flashable zip for recovery flashers. Everyone's happy.
FAQ
-But the device has kernel level throttling monitor. That one works too, right?
-Yes, but it does not do anything related to cpu clock management and such. All it does is turning off your device entirely when undesirable temperatures are reached. Thermal-engine is what's monitoring temperatures and adjusting clocks when necessary. Kernel level throttling is there for additional safety, in case thermal-engine is not working properly.
-My device runs great. Why bother modifying what Xiaomi did?
-It's up to you to do anything with your purchase. If you are happy, it's fine, leave it as it is.
-I noticed hugely increased benchmark scores and games working smoother after deleting thermal-engine. Is that normal?
-Exactly. Here's your answer to what thermal-engine actually does. It hampers our device too much.
-When are you going to research this modification?
-As soon as I'm free from doing routinal things: school, job, home and such.
BONUS TIME!
Gfxbench 3 Manhattan 1080p offscreen scores for your reference:
With thermal-engine running: 15fps
Without thermal-engine running: 23fps
REMEMBER, I'M IN NO WAY RECOMMENDING YOU TO DELETE THERMAL-ENGINE. IF YOU DO THIS, IT'S YOUR RESPONSIBILITY WHEN DEVICE STARTS COOKING OR CATCHING FIRE. THIS IS SORT OF MY OBSERVATION OF A PROBLEM OUR DEVICE HAS. EVERYTHING IS AT OUR OWN RISK!
theres currently 2 thermalmod magisk module for our device, one is jthermal and the other is thermalx, currently using thermalx and it essentially removes thermal throttling, highest temp i got was 48°c playing pubg at smooth+extreme settings at noon.
aron11195 said:
theres currently 2 thermalmod magisk module for our device, one is jthermal and the other is thermalx, currently using thermalx and it essentially removes thermal throttling, highest temp i got was 48°c playing pubg at smooth+extreme settings at noon.
Click to expand...
Click to collapse
Might give a try. Thanks.
aron11195 said:
theres currently 2 thermalmod magisk module for our device, one is jthermal and the other is thermalx, currently using thermalx and it essentially removes thermal throttling, highest temp i got was 48°c playing pubg at smooth+extreme settings at noon.
Click to expand...
Click to collapse
Where are these? I can't find any mods like these for lavender., only for violet.
Morutimeru said:
Where are these? I can't find any mods like these for lavender., only for violet.
Click to expand...
Click to collapse
its on the official redmi note 7 telegram.
aron11195 said:
its on the official redmi note 7 telegram.
Click to expand...
Click to collapse
I've found them mods on 4pda forums, guys there are really on fire when it comes to this phone.
airidosas252 said:
I've found them mods on 4pda forums, guys there are really on fire when it comes to this phone.
Click to expand...
Click to collapse
Oh... I can't understand russian so that's a no for me :/
Morutimeru said:
Oh... I can't understand russian so that's a no for me :/
Click to expand...
Click to collapse
I do not recommend using these mods! They completely remove the limits, leaving the phone vulnerable to overheating during charging.
but here is the link JThermal, I took the file I downloaded from telegram and uploaded it. To install use MAGISK!
brundark said:
I do not recommend using these mods! They completely remove the limits, leaving the phone vulnerable to overheating during charging.
but here is the link JThermal, I took the file I downloaded from telegram and uploaded it. To install use MAGISK!
Click to expand...
Click to collapse
Thanks! I got a hold of the thermalX one. Registered on 4pda and downloaded it.
Install ACCA from magisk and you'll be able to control max temperature while charging
aron11195 said:
its on the official redmi note 7 telegram.
Click to expand...
Click to collapse
can you share the link? to the telegram group
I have tested the jthermodv3 and my battery icon went red warning. I have disabled it instantly and restarted and now vibration doesnt work.
do not make the mistake i did. f*ck this ****.
eraycetin said:
I have tested the jthermodv3 and my battery icon went red warning. I have disabled it instantly and restarted and now vibration doesnt work.
do not make the mistake i did. f*ck this ****.
Click to expand...
Click to collapse
Disable battery saver.
edit: where did u get v3? I'm using v2, I'd like to update.
Konduity said:
Disable battery saver.
edit: where did u get v3? I'm using v2, I'd like to update.
Click to expand...
Click to collapse
Someone posted it on the first page. I had to format data wierd things happend. Don't flash it!
eraycetin said:
Someone posted it on the first page. I had to format data wierd things happend. Don't flash it!
Click to expand...
Click to collapse
Yesterday I flashed this module on latest xiaomi.eu beta and everything is working perfectly. Maybe it works only with MIUI based ROMs?
Konduity said:
Yesterday I flashed this module on latest xiaomi.eu beta and everything is working perfectly. Maybe it works only with MIUI based ROMs?
Click to expand...
Click to collapse
It "works" on any rom, but it is not recommended, it lets your phone get really hot, and it destroys the battery life
brundark said:
It "works" on any rom, but it is not recommended, it lets your phone get really hot, and it destroys the battery life
Click to expand...
Click to collapse
Yes, I know. Everything in life is a trade. I'm willing to lose some battery lifespan for performance. Not being able to play games @60fps is unacceptable.
Konduity said:
Yesterday I flashed this module on latest xiaomi.eu beta and everything is working perfectly. Maybe it works only with MIUI based ROMs?
Click to expand...
Click to collapse
Possible. I was at PE 10 Beta. Maybe it's because of the SDK differences.
eraycetin said:
Possible. I was at PE 10 Beta. Maybe it's because of the SDK differences.
Click to expand...
Click to collapse
Most of the custom roms reuse already available binaries and stuff from MIUI, unless dev is clever enough to transition fully to CAF, so it should work, I think.
From what I've seen, MIUI is not any different when it comes to thermal configurations. Both of them use the same binary and same configs (they might be changed, not sure)
airidosas252 said:
While being owner of my Redmi Note 7, which definitely has some issues, I started experimenting with it.
Throttling is a common term amongst electronics and such. It's a way to safe our precious silicon from eventually cooking under high temperatures. There're a lot of issues involved here. Let me explain:
Xiaomi's belief is maximizing battery life while forgetting about optimal device's performance. Here's the issue, though - neither of which are properly calibrated. For a battery of this size, you would expect something better, but it's not. Looking through system files, there are several thermal config files, which simply tells to the thermal-engine binary at what state device should throttle. Sounds like a good idea, right? Yeah, not quite.
Upon opening these exact config files you see several lines, which describe temperature thresholds and which frequency device should start to use under it's determined state. There's more to this - GPU gets less voltage when it's getting warmer, thus reducing the ability for GPU to reach higher power levels. Everything would be great, if the device wouldn't start to throttle it's clocks at 45° degrees, which is basically nothing.
After seeing device's guts it's safe to state, that our phone has proper cooling done to it. Sure, it's basic, but with that much space and the price this phone is built for, you couldn't ask for better cooling.
So, what can you do to it?
There're several solutions.
-Deleting thermal-engine entirely and forgetting about the problem. Here's the thing, though - it's really dangerous, because not only you lose the ability to monitor your battery's temperature, you could cook your phone while charging. Charging throttling is a thing too, when phone slows down pulling current from a wall as soon as you are doing something intensive. It's a dirty fix with a lot of risks involved.
-Modifying thermal config tables. This way you could ensure device's safety while having optimal device's performance.
-Mounting hugeass cooler at the back of your phone or better yet - mounting cooler straight to the cpu silicon. Jokes aside, this is not recommended at all, unless you're insane and don't care about sacrificing device's size.
I'm actually thinking about making completely custom thermal tables and make them compatible with Magisk. Even if Magisk method wouldn't ensure enough confidence, I will make flashable zip for recovery flashers. Everyone's happy.
FAQ
-But the device has kernel level throttling monitor. That one works too, right?
-Yes, but it does not do anything related to cpu clock management and such. All it does is turning off your device entirely when undesirable temperatures are reached. Thermal-engine is what's monitoring temperatures and adjusting clocks when necessary. Kernel level throttling is there for additional safety, in case thermal-engine is not working properly.
-My device runs great. Why bother modifying what Xiaomi did?
-It's up to you to do anything with your purchase. If you are happy, it's fine, leave it as it is.
-I noticed hugely increased benchmark scores and games working smoother after deleting thermal-engine. Is that normal?
-Exactly. Here's your answer to what thermal-engine actually does. It hampers our device too much.
-When are you going to research this modification?
-As soon as I'm free from doing routinal things: school, job, home and such.
BONUS TIME!
Gfxbench 3 Manhattan 1080p offscreen scores for your reference:
With thermal-engine running: 15fps
Without thermal-engine running: 23fps
REMEMBER, I'M IN NO WAY RECOMMENDING YOU TO DELETE THERMAL-ENGINE. IF YOU DO THIS, IT'S YOUR RESPONSIBILITY WHEN DEVICE STARTS COOKING OR CATCHING FIRE. THIS IS SORT OF MY OBSERVATION OF A PROBLEM OUR DEVICE HAS. EVERYTHING IS AT OUR OWN RISK!
Click to expand...
Click to collapse
Can you tell me how to delete thermal engine please iam currently using redmi note 7 pro.Iam getting a lot of throttling please help

Categories

Resources