So.. why should someone Overclock his Phone? - Xiaomi Mi A1 Questions & Answers

So here’s what I got: It’s not about getting more power out of your processor. It’s something else..
As a PC Overclocker, getting more power out of the Phone’s processor was the first thing I thought was behind Phone Overclocking. But that should be impossible right? I mean.. how the hell are you gonna cool that thing? You can’t.. And it is. That’s why benchmarks give you the finger after you OC your phone. At best you’ll get about 15% more “points”, which of course is laughable.. But then again why do people do it? Is it a fetish? Well it’s not. It’s about these three things:
UI Reaction Time – UI Animations Smoothness – Memory Management
Three things that determine the quality of your daily experience with your phone.
Every vendor makes his own adjustments to the Kernel that determines the behavior of the hardware in response to the software. Some make it right, some don’t. And when you find yourself hostage to the latter, the only way you can break free and restore your phone’s true potential, is with a custom Kernel that allows you to adjust it. And it’s not even necessary to increase the frequencies! Tightening the timings, or even better eliminating them, can make all the difference in the world! But since a Mobile Chip has a built in protection against overheating, and raising the frequencies could not possibly harm it, then what hell.. why not?
And that god damn memory management that doesn’t let you have more that 4 - 5 apps in the background, even though you have 4 GB of RAM!.. what the hell is that! Well you can fix that too.
So you see, appearances can be deceiving. Phones are different creatures that the PCs. We do different things with them, so we need different things from them. And if you’re willing to give up just a little bit of your battery life.. man..you will NOT wanna go back!
So that’s it!
PS. Speaking of battery life, there is also that thing called “Underclocking”. It can even double your battery life, but of course the price would be the exact opposite effects of the above. Every man has his poison!..

Conpsycon said:
......But since a Mobile Chip has a built in protection against overheating, and raising the frequencies could not possibly harm it, then what hell.. why not?....
Click to expand...
Click to collapse
That is false. Even with thermal protection, the lifespan of the cpu can be drastically reduced if its overclocked and constantly running hot.
And many users overclock in order to get a better experience playing games.

barrack1 said:
That is false. Even with thermal protection, the lifespan of the cpu can be drastically reduced if its overclocked and constantly running hot.
And many users overclock in order to get a better experience playing games.
Click to expand...
Click to collapse
That is true in general. It's called electromigration and it doesn't matter if you cool it well enough. It is caused by the increased kinetic energy of the electrons. But phone CPUs don't get stressed enough to see it happening. Even in PCs where actual overcloking takes place, it would take many years to actually see it happening, and a very serious increase in performance.
So in comparison one could say that it really is safe.
One interesting thing that someone said to me last night though, is that custom kernels don't have a perfect temperature throttling when necessary. So one should check the temps in the initial stage of testing to make sure that it doesn't get too hot.

Two useful links about CPU Temperatures:
https://www.xda-developers.com/processor-temperature-results-for-tens-of-socs-how-hot-is-your-chip/
https://www.quora.com/What-is-the-safe-temperature-of-mobile-CPU

Conpsycon said:
That is true in general. It's called electromigration and it doesn't matter if you cool it well enough. It is caused by the increased kinetic energy of the electrons. But phone CPUs don't get stressed enough to see it happening. Even in PCs where actual overcloking takes place, it would take many years to actually see it happening, and a very serious increase in performance.
So in comparison one could say that it really is safe.
One interesting thing that someone said to me last night though, is that custom kernels don't have a perfect temperature throttling when necessary. So one should check the temps in the initial stage of testing to make sure that it doesn't get too hot.
Click to expand...
Click to collapse
I would've disagree with that. Overclocking used to work reasonably well a long time ago but for about 10-15yrs the headroom in desktop cpu's have shrunk and overclocking does not give stable results. My own experiences coincides with what I've heard with people who work in IT who build/sell pc's - overclocked cpus will encounter much more errors and intermittent problems and will require resetting back to stock clocks or even underclocking in order to be stable.

barrack1 said:
I would've disagree with that. Overclocking used to work reasonably well a long time ago but for about 10-15yrs the headroom in desktop cpu's have shrunk and overclocking does not give stable results. My own experiences coincides with what I've heard with people who work in IT who build/sell pc's - overclocked cpus will encounter much more errors and intermittent problems and will require resetting back to stock clocks or even underclocking in order to be stable.
Click to expand...
Click to collapse
Check this out for example: I underclock my CPU to 1689MHz from 2000MHz, but I change the governor to Performance, which keeps the cores constantly to that frequency WHEN operational. The result is a MUCH snappier behavior with no apparent latency, zero stress to the CPU, and battery compensation by the lower frequency.
Could you do that otherwise? That's what I'm talking about! Full flexibility!

Conpsycon said:
Check this out for example: I underclock my CPU to 1689MHz from 2000MHz, but I change the governor to Performance, which keeps the cores constantly to that frequency WHEN operational. The result is a MUCH snappier behavior with no apparent latency, zero stress to the CPU, and battery compensation by the lower frequency.
Could you do that otherwise? That's what I'm talking about! Full flexibility!
Click to expand...
Click to collapse
Thats not overclocking and you did not talk about flexibility. Even the title of your thread says "Overclock".
You said overclocking cpus on pc's and phones were safe and would take "many years" to see any degradation.

barrack1 said:
Thats not overclocking and you did not talk about flexibility. Even the title of your thread says "Overclock".
You said overclocking cpus on pc's and phones were safe and would take "many years" to see any degradation.
Click to expand...
Click to collapse
I said both of them, and they do not conflict with each other. I explained it in my first post.

Underclocking is a bad thing. unless you can't overclock much because of battery and temperature limitations, underclocking is not recommended. It depends on your governor, a stupid governor will slow you down and be visible on weak devices. your case is due to have stupid governor in phone, but suitable with strong device & want to save battery but don't need to lock freq at maximum

barrack1 said:
Thats not overclocking and you did not talk about flexibility. Even the title of your thread says "Overclock".
You said overclocking cpus on pc's and phones were safe and would take "many years" to see any degradation.
Click to expand...
Click to collapse
cpu will not degrade in life no matter how high frequency, overclock and runs continuously but its life cycle will be affected by high temperature and high voltage
{
"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"
}

Related

Is your G2 significantly faster once you overclock it?

I've recently bought a G2 and am thinking of overclocking it. Currently, its not even rooted yet, and I am quite new to all this.
Tried running Smartbench 2010 and got 1033 for the productivity index and 1380 for the games index. (Nexus One using CM6 seems to get 1000/1000). What kind of scores do you get when overclocked?
I recommend downloading linpack and quadrant standard
using linpack i got 36 mflops stock, now I get about 50 at 1.5 ghz
quadrant stock I got about 1600 and overclocked I get about 2500
flash content runs a lot better overclocked IMHO
Thanks for the response.
I looked at Quadrant but many people seem to say that the result is not great due to its heavy reliance on the file i/o speeds. Linpack seems to mainly stress the floating point speed of the processor.
Could you try running Smartbench on yours? At what clock speed are you running your phone?
I run 1.5 ghz overclocked
Running CM6, with stock speed of 800mhz, I get 988 and 1367
Overclocked to 1.5ghz, I get 1679 and 2121
Looks like G2 scales quite well! Has anyone gotten higher stable clock?
benching you see a difference but i do not notice anything else being faster. The phone is already fast.
Agreed. But when I do push my G2 to the limit (i.e. either running many apps or run few demanding apps), it does slow down. I figure it would make some differences in these situations.
I can definitely see a difference when overclocking, however only on cpu-taxing tasks:
- Kirby Super Star (a difficult game to emulate) on SNesoid is jerky on stock speeds and the sound stutters. On 1.5GHz it's (almost) totally smooth.
- psx4droid often achieves better framerates with overclocking.
- ADWLauncher EX is noticeably smoother during screen transitions on higher clocks, especially when widgets are involved.
etc.
806MHz doesn't quite cut it imo.
Running at 1500mhz definitely makes the phone faster. I have a pretty involved setcpu profile to optimize for every situation. As a result, I don't have poor battery performance, and the g2 runs better than any other android phone I've used.
The biggest boost is in the browser, its amazingly fast. Faster than chrome on my ubuntu netbook.
Sent from my HTC Vision using XDA App
Not to thread-jack, but how much does OCing at 1.5 impact the battery? I'd have it scaled down when the screen is off, but do you notice a big difference?
Sent from my Nexus S using Tapatalk
TheBiles said:
Not to thread-jack, but how much does OCing at 1.5 impact the battery? I'd have it scaled down when the screen is off, but do you notice a big difference?
Sent from my Nexus S using Tapatalk
Click to expand...
Click to collapse
I have mine OC to 1.42ghz and use the 1800mah Evo battery which fits inside the G2 using the stock battery cover so it doesn't add thickness to the phone, and can go from 5am till 9pm with moderate to heavy use and still have about 10% battery left.
Sent from my T-Mobile G2 using XDA App
joemm said:
I have mine OC to 1.42ghz and use the 1800mah Evo battery which fits inside the G2 using the stock battery cover so it doesn't add thickness to the phone, and can go from 5am till 9pm with moderate to heavy use and still have about 10% battery left.
Sent from my T-Mobile G2 using XDA App
Click to expand...
Click to collapse
Does it add considerable weight to the phone?
How do you heavily use your phone for an entire day?
Just flashed the virtuous rom to mine and have the CPU overclocked moderately to 1ghz. Have noticed slight increase in snappiness.
Sent from my HTC Vision using XDA App
ranzz said:
Does it add considerable weight to the phone?
How do you heavily use your phone for an entire day?
Click to expand...
Click to collapse
It adds no extra weight, atleast none I notice here's the thread for the battery( the red ones form chichitech), :
http://forum.xda-developers.com/showthread.php?t=799183
As fas as your second question, "How do you heavly use your phone an entire day?", if you cant figure that out I guess you are confused.
I have played around with the OC settings quite a bit. A few things:
Battery life is not much different than stock. If anything, it improves the battery slightly. I am currently running SetCpu with no profiles and have the governor set to 'on demand'.
The difference between 800mhz and 1.2 ghz is pretty significant. Not only have I seen higher benchmark scores but the phone feels even 'snappier'. No lag in transitions when switching between vertical and horizontal layouts. ADW EX transitions smooth between screens using effects, etc.
I have had it up to 1.5 ghz at times, and to be honest, I don't notice a huge difference between that and the performance at 1.2. Quadrant benchmark score is a few hundred points higher but in day to day usage on my device, the phone feels about the same.
I see very significant difference between my rooted, 1.5ghz OC'ed G2 and my wifes stock G2. My phone opens apps like maps, nav, earth, browser etc.. in half the time. I have noticed that the earth app is not as smooth on stock as it is oc.
I always thought that the G2 was fast stock, and it is, but when I use my wifes phone to hop on the web or whatever, I immediatly notice how much slower it is than mine.
Sent from my T-Mobile G2 using XDA App
joemm said:
I have mine OC to 1.42ghz and use the 1800mah Evo battery which fits inside the G2 using the stock battery cover so it doesn't add thickness to the phone, and can go from 5am till 9pm with moderate to heavy use and still have about 10% battery left.
Sent from my T-Mobile G2 using XDA App
Click to expand...
Click to collapse
sorry to be a lazy ass, but where is this 1800mhz evo battery the cheapest? or even safest to buy from?
I realize I could google it, and I will if i dont get an answer today haha thx again
There is a battery thread in the accessories section and I think there is a link to it there if I am not mistaken.
Sent from my T-Mobile G2 using XDA App
Mog said:
sorry to be a lazy ass, but where is this 1800mhz evo battery the cheapest? or even safest to buy from?
I realize I could google it, and I will if i dont get an answer today haha thx again
Click to expand...
Click to collapse
Heres where I got mine from:
http://www.amazon.com/Sprint-Battery-Charger-Verizon-Incredible/dp/B003XOUM12
You get 2 batteries plus a travel charger.
To get them to fit you have to remove the red wrapper off the battery heres a link to a video on how to get them to fit.
http://www.youtube.com/watch?v=WfDdDWttzRI&feature=player_embedded
Flashed this on my G2 running CM 6.1.1 and got a Quadrant score of 3037! I had it OC'd to 1.8 GHz with scaling set to performance, which would obviously mean that the G2 is incredibly faster if you OC it. However, I'm sure it will affect battery life, so I'm only OC'ing my phone to 1.4 GHz, not the fully capable 1.8 GHz.
{
"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"
}

Effects of CPU Frequency and Screen Brightness on Power Consumption

I've recently been curious as to how much influence my chosen CPU frequencies and screen settings have on power usage. Of course, it is common knowledge that an overclocked processor requires more current and that our displays are power hungry, but how apparent is the difference? Is the relationship between these variables and battery consumption linear, or is the increase in power demand perhaps exponential?
In an effort to explore these questions, I set up a controlled test that logged CPU frequency and screen brightness against current usage. This data should not be assumed to directly apply to all Incredibles; I was looking for correlations rather than exact calculations on how much energy a specific setting uses.
Methods (go ahead and skip down to the findings section if this is too dry for you)
AMOLED Inc
Kernel: Chad's 2.6.37.1-incredikernel-gb-2272011
ROM: CyanogenMod 7 - Nightly #11
And not that it should matter, but someone will ask so...
Radio: 2.15.00.11.19
HBOOT: 0.92
Recovery: ClockworkMod 3.0.0.8
CPU frequencies were controlled using SL4A scripts and Tasker. Max frequency can be manipulated through /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq, so a single echo command is all that is needed to adjust it. I set up a rudimentary if/then script to bump up the max frequency one notch at a time, then used Tasker to execute this script every two minutes.
When Tasker would execute the script, it would also write an entry to a log file shared with CurrentWidget. This ensured that the CPU frequency was being noted between milliamp readings. Multiple readings were taken at each level, and entries that immediately preceded or followed clock changes were discarded in an effort to use the more stable readings at each frequency.
To control for other influences, all other phone services were disabled, including GPS, Wi-Fi, and Bluetooth. The phone was also placed in Airplane Mode so that reception was not a factor.
The CPU governor was set at "performance," ensuring that it remained at its max frequency. To task the CPU, I put together another simple script which just ran md5sum commands in a never-ending loop. Before recording data, I verified that this script was enough to load the CPU and cause it to scale up to max frequency when using other governors.
The full range of frequencies was tested at the screen's dimmest factory setting (20/255) as well as the screen's brightest setting (255/255). Additionally, a separate test was run with the CPU fixed at 576 MHz and Tasker periodically bumping up the screen brightness.
Findings
My data showed an almost perfect linear correlation between clock speed and power use, as well as brightness and power use.
I was a little surprised at how big of a difference I was able to find. With the screen at its dimmest setting, 128 MHz consumed ~80 mA while 1,113 MHz consumed more than 300 mA. For a stock Inc battery (1,300 mAh) that marks the difference between (over) 16 hours of runtime and less than 4.5 hours of runtime. Of course, anyone who has tested the limits of underclocking knows that the phone is utterly useless at 128 MHz.
The influence of brightness was less of a surprise. At a given CPU frequency, the average difference that I noted between min screen brightness and max screen brightness was ~118 mA.
{
"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"
}
I would love to be able to test every single combination of CPU frequency and screen brightness, but it would take hours, and I cannot keep my phone out of use for that long. I did generate some additional projections, however, based on the trends observed in the combined tests. I've added these projected points to the graph below, based on screen settings of 80/255, 140/255, and 200/255.
Shortcomings / questions for further exploration
Given the sterile nature of the data collection, I can't just make a blanket statement like "underclocking your max CPU to 883 MHz will yield an average of 15% more battery life over leaving it set to 998 MHz." As convenient as it would be to make such a claim, it ignores a number of realities. None of us are using 100% of our processor at all times, so a more accurate prediction of runtime would need to take the phone's resting behavior into account.
We should also consider that the CPU's frequency will affect the speed at which it can perform tasks. If you clock your processor too low, you may actually end up consuming more power since your phone will have to work longer (resulting in more wake time and more screen-on time).
A more technically inclined member may want to chart CPU frequencies against power usage, while also including some sort of performance benchmark. Such a matrix would offer users an even better model on which to judge the cost/benefit of over or underclocking. But even the simple testing which I've completed above takes a fair amount of time and keeps the phone from being otherwise used, so it's not a very convenient thing to test.
What I find most interesting about the data is the implication that a user who overclocks his CPU max to 1,113.6 MHz and sets his screen brightness to 80 could use less power than a user who underclocks his CPU max to 883 MHz but uses a brightness of 200.
The easiest takeaway, of course, is an illustration of why users experience such wildly different run times on the same rom/kernel setup. Again, this should come as no surprise, but skepticism should be exercised any time someone promises you that switching to a specific rom/kernel combo will guarantee you X hours of battery life.
[This work may be freely distributed for non-commercial purposes, as long as it is passed along unchanged and in whole, with credit to author and a link back to this XDA thread.]
Wow, this is a shocker. I guess I should kill the habit of using higher brightness and be less afraid to OC.
pianoplayer said:
Wow, this is a shocker. I guess I should kill the habit of using higher brightness and be less afraid to OC.
Click to expand...
Click to collapse
I notice you have an SLCD DInc. It's worth mentioning that mine is AMOLED and I don't know if the two have similar power usage across the brightness scale.
byrong said:
I notice you have an SLCD DInc. It's worth mentioning that mine is AMOLED and I don't know if the two have similar power usage across the brightness scale.
Click to expand...
Click to collapse
Wow, I didn't notice that. I guess you're right. I read somewhere that darker themes help on AMOLED and lighter themes are better for battery on SLCD. My theme is quite light so I think that makes my on the less-consuming side for the screen.
byrong, this is awesome data. Just what I needed to see, thank you. I have the exact same setup and have been trying to figure out the best cpu/brightness settings for my device. With seeing this data its helped me determine some of the values I was going back and forth on. Now I guess I'll have to wait and see how my battery holds up.
Excellent. Thanks for the data. Even if I never put any of that into use it's still fun for the nerdy part of me (most of it ) to know! Excellent data, and just goes to show how important screen brightness is over the CPU Speed. Most users would probably get better battery life if they spent more time darkening the screen than installing a CPU governing program haha.
Great info! Kinda knew this already but cool to see it all graphed out.
Thank you for the graphs! I have the SLCD... but I think the results would be similiar regardless of screen type, especially in real-world usage.
Awesome work. Gives me some ideas about using setcpu...
Chad give me more description... what new features are in the new kernel....
Sent from my ADR6300 using XDA App
Gorilla* said:
Chad give me more description... what new features are in the new kernel....
Sent from my ADR6300 using XDA App
Click to expand...
Click to collapse
Ummm? It will enable you easier entrance/exit from the matrix.
Nice work btw byrong.
Horay for empiricism!
Thanks for the legwork and for sharing. Impressive.
[I for one welcome our new Android overlords ::: DInc::SLDC::x.7.28::CWM3.0.0.8::CM7::chad's]
phaze one said:
Ummm? It will enable you easier entrance/exit from the matrix.
Nice work btw byrong.
Click to expand...
Click to collapse
Aren't we calling it "the Grid" now?
[I for one welcome our new Android overlords ::: DInc::SLDC::x.7.28::CWM3.0.0.8::CM7::chad's]
Byrong,
Fantastic write up as always..
I'm curious if you would care to do one on the effects of a task killer on battery life/CPU and RAM usage?
Since Android launched cupcake, I've been strongly opposed to the use of task killers, as Android handles processes much better than we do.. but I would love to see if fact can back up that opposition or if these apps truly don't cause any harm.
I'm just not nearly as amazing with words and data as you clearly are.
via Incredible Redemption
Man, I wish I hadn't just given my mom my DInc and instead played with it some more. I would love to help do some more long term tests to help with this.
That and seeing the variation for screen type, mine being an SLCD.
Sent from my TBolt using XDA
Thank you all for the kind words.
AdhvanIt said:
I'm curious if you would care to do one on the effects of a task killer on battery life/CPU and RAM usage?
Click to expand...
Click to collapse
Great idea.
I'm game, but I would have to think long and hard about how to design the testing parameters. In all likelihood, I'd need some suggestions / ideas from you guys to carry out such a test.
I would need recommendations on a reliable logging program for RAM, CPU Freq, and CPU Load.
Unlike most of my other power testing, I would not want to run such a test in Airplane mode. One of the reasons that people give for aggressive task-killing is shutting down processes that may be using data in the background. As such, I would need conditions that actually allowed an observable difference in data usage.
Since the phone won't be in Airplane mode, my controls take a big hit. During a test, for example, I might receive a lot of calls, texts, and emails, all of which will have a major impact on CPU, RAM, and battery. The cleanest solution I can see is running the test across multiple nights: X full nights without a task-killer and X full nights with one.
The other big question is how to set up the task killer. I.E. should I allow it to kill everything, or set up an exclusion list? I'm leaning towards the exclusion list, but then we have to decide what to exclude. We should also think of a designated interval for the task killer.
If you want to help me brainstorm these issues, I'm happy to discuss them right in this thread.
TodesEngeI said:
Man, I wish I hadn't just given my mom my DInc and instead played with it some more. I would love to help do some more long term tests to help with this.
That and seeing the variation for screen type, mine being an SLCD.
Sent from my TBolt using XDA
Click to expand...
Click to collapse
The beauty of this type of testing is that it can be done by any motivated individual, on any device that allows scaling of the CPU. If you already have that new TBolt of yours rooted, you should be able to recreate the tests on that. I'm sure fellow TBolt users would love to see your findings.
byrong said:
Thank you all for the kind words.
Great idea.
I'm game, but I would have to think long and hard about how to design the testing parameters. In all likelihood, I'd need some suggestions / ideas from you guys to carry out such a test.
I would need recommendations on a reliable logging program for RAM, CPU Freq, and CPU Load.
Unlike most of my other power testing, I would not want to run such a test in Airplane mode. One of the reasons that people give for aggressive task-killing is shutting down processes that may be using data in the background. As such, I would need conditions that actually allowed an observable difference in data usage.
Since the phone won't be in Airplane mode, my controls take a big hit. During a test, for example, I might receive a lot of calls, texts, and emails, all of which will have a major impact on CPU, RAM, and battery. The cleanest solution I can see is running the test across multiple nights: X full nights without a task-killer and X full nights with one.
The other big question is how to set up the task killer. I.E. should I allow it to kill everything, or set up an exclusion list? I'm leaning towards the exclusion list, but then we have to decide what to exclude. We should also think of a designated interval for the task killer.
If you want to help me brainstorm these issues, I'm happy to discuss them right in this thread.
Click to expand...
Click to collapse
SetCPU has a lot of the logging/current system information that you would need, including:
CPU frequency: Time in state for each freq
CPU load average
Total and available RAM
As well as a number of other things. Screenshots are attached below.
As for controls, I would think auto kill at screen wake is a commonly used setting. Or maybe go at 1/2 to one hour intervals...
If you're going to set up a blacklist within the task-killer, I would recommend not adding too much to it, as the exclusion of too many apps would alter test results. But certainly add any auto-updating widgets you have to it.
I think leaving data on is also appropriate for this test-there are a number of apps that will not start/update (mostly web based like twitter/facebook and such widgets) if there is no data connection.
Edit: Forgot to add images. Lol.
AdhvanIt said:
SetCPU has a lot of the logging/current system information that you would need, including:
CPU frequency: Time in state for each freq
CPU load average
Total and available RAM
Click to expand...
Click to collapse
The info tab in SetCPU is indeed helpful for more of a 'snapshot view', but I don't know of a way to make it log that information to a text file over time, do you?
Time in state is helpful, and can be accessed via:
/sys/devices/system/cpu/cpu0/cpufreq/stats/time_in_state
But the problem is that it doesn't tell the story of when the CPU was scaling. It would be far more useful if I could plot CPU load and CPU freq over a time line, rather than just have averages or sums.
Same story with RAM: I want to be able to plot over the course of time.
AdhvanIt said:
As for controls, I would think auto kill at screen wake is a commonly used setting. Or maybe go at 1/2 to one hour intervals...
Click to expand...
Click to collapse
I like to automate my tests as much as possible, not only to (try to) remove myself as an influence, but so that I'm not tied to the phone for hours at a time, looking at a watch. If I did a screen on/off setting, it would require me to turn the screen on and off at select times, remaining consistent between tests.
So intervals would definitely be the way for me to go. I think I'll start a poll of some sort to get an idea of how most people set theirs.
AdhvanIt said:
If you're going to set up a blacklist within the task-killer, I would recommend not adding too much to it, as the exclusion of too many apps would alter test results. But certainly add any auto-updating widgets you have to it.
Click to expand...
Click to collapse
Agreed that I want to keep the exclusion list to a minimum. I will make this part of my poll as well. I don't want to create a strawman position by setting up the task killer to be more aggressive than anyone actually uses it. For example, I don't want to set it to kill Android OS, Launcher, etc, as that would just be excessive and probably does not reflect how anyone uses them.
This will be an extremely controversial one, and if I don't strike a delicate balance, people will be quick to throw out the data completely on the grounds of, "Well that's fine and all, but that's not how I use my Task Killer."
Thanks for helping me think this through.
byrong said:
The info tab in SetCPU is indeed helpful for more of a 'snapshot view', but I don't know of a way to make it log that information to a text file over time, do you?
Time in state is helpful, and can be accessed via:
/sys/devices/system/cpu/cpu0/cpufreq/stats/time_in_state
But the problem is that it doesn't tell the story of when the CPU was scaling. It would be far more useful if I could plot CPU load and CPU freq over a time line, rather than just have averages or sums.
Same story with RAM: I want to be able to plot over the course of time.
I like to automate my tests as much as possible, not only to (try to) remove myself as an influence, but so that I'm not tied to the phone for hours at a time, looking at a watch. If I did a screen on/off setting, it would require me to turn the screen on and off at select times, remaining consistent between tests.
So intervals would definitely be the way for me to go. I think I'll start a poll of some sort to get an idea of how most people set theirs.
Agreed that I want to keep the exclusion list to a minimum. I will make this part of my poll as well. I don't want to create a strawman position by setting up the task killer to be more aggressive than anyone actually uses it. For example, I don't want to set it to kill Android OS, Launcher, etc, as that would just be excessive and probably does not reflect how anyone uses them.
This will be an extremely controversial one, and if I don't strike a delicate balance, people will be quick to throw out the data completely on the grounds of, "Well that's fine and all, but that's not how I use my Task Killer."
Thanks for helping me think this through.
Click to expand...
Click to collapse
Definitely makes sense to automate...
I'm not really sure of any way to log these things regularly unless you know how to set up a script and have it pull the logs every so often...
Agreed-It will likely get met with debate regardless, as it's always had its fanbase and its opposition. Not sure really how to get a good balance, but I think the polls are a good way to start.
Survey is up: http://www.surveymonkey.com/s/8GD2F2H
(Posted here to reach a wider audience: http://forum.xda-developers.com/showthread.php?t=1032456)
I decided to keep it as simple as possible, both to encourage more people to respond, and to simplify the collection of responses.

supposed solution to heat during gaming .

i want to discuss the overlock or underlocking the cpu
so whats the effect on heat generated ????
overlock increase heat or dcrease heat ??
and is undervolting helps to decrease heat generated from the processor ??
many tests mentioned that overlock 1.6 is stable and undervolting scale (200-800) is also stable but non of the testers mentioned the heat effect all are reporting for stability and speed
i suffer from heat during hd games and i know it is normal but i dont want the phone to get hot during gaming i closes the game after around 30 minutes due to this problem
If you want to reduce heat during gaming, then you need to change the 1.2GHz speed (level 0) down to either 1.0GHz or 800MHz. Changing the Level 1 to Level 4 speeds won't solve your problem since your CPU will be at full speed Level 0 during gaming.
Sleepycat3 said:
If you want to reduce heat during gaming, then you need to change the 1.2GHz speed (level 0) down to either 1.0GHz or 800MHz. Changing the Level 1 to Level 4 speeds won't solve your problem since your CPU will be at full speed Level 0 during gaming.
Click to expand...
Click to collapse
srry i dont understand what will make the processor return to full speed 1.2 during gaming ?? if i set it to 1000 or 800 ???
Sleepycat3 said:
If you want to reduce heat during gaming, then you need to change the 1.2GHz speed (level 0) down to either 1.0GHz or 800MHz. Changing the Level 1 to Level 4 speeds won't solve your problem since your CPU will be at full speed Level 0 during gaming.
Click to expand...
Click to collapse
No that's just bull****. Go play around a few games and have CPU Spy running and after a gamsing session check the CPU states, rarely does a game go over 800MHz and most of the time spent is on 500MHz for most games. There aren't any games out there that need the computational power of higher frequencies and that's why their time-in-state is procentually so low.
If you really want to reduce heat, then you need to undervolt your GPU which right now is only possible with a Tegrak compatible kernel, which is frankly speaking out of date, or you wait a few weeks until the Ninphetamine team finishes GPU voltage control interface and then more kernels will be able to implement that.
While undervolting with Tegrak I was running 160MHz/750mV 266MHz/750mV full stable (default voltage 950/1000mV, 750 was minimum configurable). It was running full stable and while on default the back of the phone would get really hot, while undervolted it was just warm to the touch.
AndreiLux said:
No that's just bull****. Go play around a few games and have CPU Spy running and after a gamsing session check the CPU states, rarely does a game go over 800MHz and most of the time spent is on 500MHz for most games. There aren't any games out there that need the computational power of higher frequencies and that's why their time-in-state is procentually so low.
If you really want to reduce heat, then you need to undervolt your GPU which right now is only possible with a Tegrak compatible kernel, which is frankly speaking out of date, or you wait a few weeks until the Ninphetamine team finishes GPU voltage control interface and then more kernels will be able to implement that.
While undervolting with Tegrak I was running 160MHz/750mV 266MHz/750mV full stable (default voltage 950/1000mV, 750 was minimum configurable). It was running full stable and while on default the back of the phone would get really hot, while undervolted it was just warm to the touch.
Click to expand...
Click to collapse
so is there any effect on device performance ????
after undervolting the hd games performs properly as stock ?? and without heat ?? if yess i will root now and see what can i do
and how can i know that my kernel is tegrak compatible ?? it is kg2
Changing voltage has no effect on performance unless you undervolt into unstable values and make it crash. Stock should be compatible with Tegrak (I think???). Just check out the free app in the market. The stupid thing is you need to pay for the full version to get GPU control and that's why it's unpopular, so maybe it's just better to wait for Ninphetamine GPU control in a couple of weeks or soonish and have it for free and get a good kernel to go with it.
AndreiLux said:
Changing voltage has no effect on performance unless you undervolt into unstable values and make it crash. Stock should be compatible with Tegrak (I think???). Just check out the free app in the market. The stupid thing is you need to pay for the full version to get GPU control and that's why it's unpopular, so maybe it's just better to wait for Ninphetamine GPU control in a couple of weeks or soonish and have it for free and get a good kernel to go with it.
Click to expand...
Click to collapse
thanks bro
but i think i have small thing that confuse me undervolting is different from downlock ??
and what is the relation between them ??? i think certain volt will give related amount of "Mhz" ?? is that right ??? so set cpu is the same effect as tegrak or what ??
srry im new to android
xdafun4all said:
thanks bro
but i think i have small thing that confuse me undervolting is different from downlock ??
and what is the relation between them ??? i think certain volt will give related amount of "Mhz" ?? is that right ??? so set cpu is the same effect as tegrak or what ??
srry im new to android
Click to expand...
Click to collapse
Tegrak, basically gives you more options such as GPU clocking and custom tweaks. SetCPU is ok, but I just use it for controlling clock speeds. Also, you can set oc'ing, max at 2GHZ, but you have to play around with stuff.
Use a very thin case and you'll never notice.
Sent from my GT-I9100 using XDA App
AndreiLux said:
No that's just bull****. Go play around a few games and have CPU Spy running and after a gamsing session check the CPU states, rarely does a game go over 800MHz and most of the time spent is on 500MHz for most games. There aren't any games out there that need the computational power of higher frequencies and that's why their time-in-state is procentually so low.
Click to expand...
Click to collapse
Sure. I just fired up Plants vs Zombies, which is not the world's most demanding Android game and played 1 quick game, taking about 10 minutes. While playing, the back of the phone just where the camera is became warm through the phone case! Tell me why I have the phone running at 1200MHz for such a big portion of time if my post is just bull****. This was also the Last Stand level, where you get a pause in the middle of the attacks to replant.
{
"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"
}
Sleepycat3 said:
Sure. I just fired up Plants vs Zombies, which is not the world's most demanding Android game and played 1 quick game, taking about 10 minutes. While playing, the back of the phone just where the camera is became warm through the phone case! Tell me why I have the phone running at 1200MHz for such a big portion of time if my post is just bull****. This was also the Last Stand level, where you get a pause in the middle of the attacks to replant.
http://img825.imageshack.us/img825/5459/sc20110829170728.png
Click to expand...
Click to collapse
35% of time on full speed does not equal all the time on full speed like you said in your first post. It's still 50% on either 800 or 500MHz like I said. And you took a bad example as I dont even know if PvsZ is even using properly the GPU, and then might use only the CPU. Go play Gameloft HD games which are the most demanding atm and max CPU ratio to the rest falls down.
AndreiLux said:
35% of time on full speed does not equal all the time on full speed like you said in your first post. It's still 50% on either 800 or 500MHz like I said. And you took a bad example as I dont even know if PvsZ is even using properly the GPU, and then might use only the CPU. Go play Gameloft HD games which are the most demanding atm and max CPU ratio to the rest falls down.
Click to expand...
Click to collapse
so what ??? at the end undervolting will help or not ???
i need it for hd games
and if the cpu need to be at 50% of time so it need to use 1200 mhz 50 of time so if we fixed it to 800 or 1000 or 500 this will affect performance
no heat issues
I have ninphetamine kernel, underclocked to 800 and undervolted, I haven't noticed any heat issues since then when playing games and noticed no slow downs. Games tried: HeavyGunner, Zenonia 3. Also no heat issues anymore when playing flash video in a browser.
wslfung said:
I have ninphetamine kernel, underclocked to 800 and undervolted, I haven't noticed any heat issues since then when playing games and noticed no slow downs. Games tried: HeavyGunner, Zenonia 3. Also no heat issues anymore when playing flash video in a browser.
Click to expand...
Click to collapse
but sorry these games are not using graphics as eternal legacy or asphalt 6 hd or fast 5 hd
really it works
i rooted then set cpu then put the values 200-800 and asphalt 6 worked around 20 minutes no heat at all only slight warm
and the benchmark only decresed from 3300 to 2750

Undervolting drastically improves 4K video recording?

I've been fooling around with undervolting CPU frequencies for the past few weeks and I've noticed that it improves 4K video recording remarkably.
.402 firmware has something to do with it too since 4K video has been underclock to 800mhz rather then 1ghz on previous firmwares.
As you can see from the sceenies below I got 34mins of 4K which was about a 13GB file. Phone was also not very hot, I would say very warm.
{
"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"
}
I can't remeber how much I undervolted from default, so I've uploaded some screenshots of my undervolts below
*NOTE* that the important frequency to undervolt would be 800mhz since that is frequency it is locked at while using 4K video recording on .402 firmware. .69 firmware and lower would lock the frequencies at 1ghz so you would undervolt the 1ghz frequency.
I cannot confirm all this as I'm the only one I know that has tested it. I will need others to try and see if they get better results then they had previously, you don't need to get 30mins for good results. I just want to know if it improves 4K for you guys or if its still the same.
*NOTE* not all phones are made the same, some phone might not be able to undervolt as much as others and may cause your phone to slowdown or reboot.
*UPDATE* did some more tests to see the temperature and if I could do back to back recordings, these were the results
36mins >>
30mins >>
As you can see from the time difference in the screenshots there was less then 1min difference, this is because I went into system monitor to check the temperature before I started the next recording. Max temp was 44c and the current temp was 42c. Seems like I could do back to back recordings endlessly, the reason I didn't do more was because I had low battery at the time and 66mins of 4K recording drained about 45% or more of my battery.
*NOTE* Weather plays an important factor and it is winter here, my first recording was done @ 11C, the 2nd and 3rd ones where done @ 16c
I can conform it as it's logical, the camera shuts down due to overheating so less voltage means less heat obviously
Sent from my D6503 using Tapatalk
Mr.R™ said:
I can conform it as it's logical, the camera shuts down due to overheating so less voltage means less heat obviously
Sent from my D6503 using Tapatalk
Click to expand...
Click to collapse
Yeah, I understand it's logical and I was referring to the drastic impact it seemed to have for me. I haven't really noticed anyone confirming that 4K video is better on .402 since it should be because 4K is underclock from 1ghz to 800mhz.
Would be good to test the temperature differences really, see how far it is below the threshold once undervolted, I'm - 36 on all steps so I may test later busy with kids atm tho
Sent from my D6503 using Tapatalk
Pesetio said:
I've been fooling around with undervolting CPU frequencies for the past few weeks and I've noticed that it improves 4K video recording remarkably.
.402 firmware has something to do with it too since 4K video has been underclock to 800mhz rather then 1ghz on previous firmwares.
As you can see from the sceenies below I got 34mins of 4K which was about a 13GB file. Phone was also not very hot, I would say very warm.
[...]
*NOTE* that the important frequency to undervolt would be 800mhz since that is frequency it is locked at while using 4K video recording on .402 firmware. .69 firmware and lower would lock the frequencies at 1ghz so you would undervolt the 1ghz frequency.
I cannot confirm all this as I'm the only one I know that has tested it. I will need others to try and see if they get better results then they had previously, you don't need to get 30mins for good results. I just want to know if it improves 4K for you guys or if its still the same.
*NOTE* not all phones are made the same, some phone might not be able to undervolt as much as others and may cause your phone to slowdown or reboot.
Click to expand...
Click to collapse
It seems to be a good point. Does the bitrate remain the same when you are undervolting & underclocking your XZ2 ?
I don't know if the cpu is used for 4K encoding or if it's a special processing unit.
We can't lock the cpu frequencies without root ...
Sony should lock cpu power when the camera application is used because 1080p @60fps and numerous shot at 20.7mpix turn it into oven.:crying:
yeah it makes complete sense. the chipset will be using less voltage and current, wich means less power. less power = less heat
in my desktop core i7 4770k for example, if i undervolt 120mv from the original voltage, i can get like 13C diference in temperatures
Which app do you use to undervolt your phone?
Rilavalon said:
It seems to be a good point. Does the bitrate remain the same when you are undervolting & underclocking your XZ2 ?
I don't know if the cpu is used for 4K encoding or if it's a special processing unit.
We can't lock the cpu frequencies without root ...
Sony should lock cpu power when the camera application is used because 1080p @60fps and numerous shot at 20.7mpix turn it into oven.:crying:
Click to expand...
Click to collapse
The CPU basically has 100% usage on all 4 cores when it's recording in 4K, I'm not sure about the GPU but when I have underclocked the GPU to 200MHz, after 4K video recording it defaults back to the 575Mhz.
Mano1982 said:
Which app do you use to undervolt your phone?
Click to expand...
Click to collapse
You need to have doomkernel 6 or above installed. If you have Romaur, it comes with a default app (Performance Control). You can use Fauxclock, SetCPU etc..
Pesetio said:
I've been fooling around with undervolting CPU frequencies for the past few weeks and I've noticed that it improves 4K video recording remarkably.
Click to expand...
Click to collapse
have you done any stress testing ?
How is the stability. This will take time to assess.
One Twelve said:
have you done any stress testing ?
How is the stability. This will take time to assess.
Click to expand...
Click to collapse
Not really however it was getting a stress test @ 800Mhz frequency, since 4K recording stresses the CPU and that is for over 30mins and I have had now slowdowns, glitches or reboots since my undervolts.
Pesetio said:
I've been fooling around with undervolting CPU frequencies for the past few weeks and I've noticed that it improves 4K video recording remarkably.
Click to expand...
Click to collapse
Do you recall the surrounding ambient temperature under which you got this performance ?
One Twelve said:
Do you recall the surrounding ambient temperature under which you got this performance ?
Click to expand...
Click to collapse
11c
However I did another test that was done at 16c where I got 36mins, checked the CPU temperature, max 44c, current 42c then I went straight into 4K again and to my surprise I got 30mins.
44c must be the cutoff temperature, what was surprising is that from 42c to 44c took my phone half an hour of 4K recording to reach.
Seems like I could keep doing 30+ mins of 4K endlessly. Obvious battery goes down hard, 66mins of 4K was 45% battery drain.
It's winter here in Melbourne now so obviously that plays an important factor. I'll have to edit original post to include these factors so it doesn't mislead people.
* note
every CPU is slightly different some can handle more undervolt than others ( much like an overclock ) so YMMV with how far you can push the UV.
and hiyo from the west coast (perth) @One Twelve
pvy.
Pesetio said:
11c
However I did another test that was done at 16c where I got 36mins, checked the CPU temperature, max 44c, current 42c then I went straight into 4K again and to my surprise I got 30mins.
44c must be the cutoff temperature, what was surprising is that from 42c to 44c took my phone half an hour of 4K recording to reach.
Seems like I could keep doing 30+ mins of 4K endlessly. Obvious battery goes down hard, 66mins of 4K was 45% battery drain.
Click to expand...
Click to collapse
oh no, i was so hoping it was warmer when you did that test.
11c or 16 c is 20 degrees lower than where i am.
What this means is the Z2 will at least be able to match the rest unlike presently when it comes to heating.
Stabilisation causes faster heating because its resource intensive. Z2 matched the lumia 920 which is some feat but it won't be able to record for as long unless ambient temperatures are 10-20 degrees. At those temps camera works longer by a significant margin over stock. Even at elevated ambient temperatures I expect there will still be an improvement. P =V^2/R and all.
The next question i have is by how much is cpu performance affected say in benchmarks as a result of under volting ?
Pesetio said:
It's winter here in Melbourne now so obviously that plays an important factor. I'll have to edit original post to include these factors so it doesn't mislead people.
Click to expand...
Click to collapse
yes, would make things more clear.
One Twelve said:
Stabilisation causes faster heating because its resource intensive.
Click to expand...
Click to collapse
I was curious so I did another test, I enabled and disabled and didn't really notice any difference in CPU usage however when I did the test I noticed that the CPU cores are clocked @ 1.2ghz now but uses less CPU. Higher clock with less CPU usage seems to be better then lower clock with more CPU usage.
I have been on Romaur 3.0 which is also .402 and it was locked @866mhz and all the CPU cores were at 98 - 100 usage. Very strange indeed.
I have noticed some strange things though from prevous versions, such as being able to under or overclock each individual CPU core and have the settings always stick which I could not do before. Maybe it's because I updated to Doomkernel 9 but I'm not sure.
Have you got the system monitor app with the floating notifcation? I'd like to know what your 4K video is running @ and what kind of CPU usage u get.
I'm on Romaur 3.1 and Doomkernel 9
Pesetio, i don't have a Z2 yet. But am watching developments on it. Your work here is very promising :victory:
One Twelve said:
Pesetio, i don't have a Z2 yet. But am watching developments on it. Your work here is very promising :victory:
Click to expand...
Click to collapse
Ahh ok I thought you had one lol
You'll enjoy it when you do though
For others who do have the phone it would be interesting to know the CPU frequency that you 4K records in and what the CPU usages are, might be able to pinpoint the problems with 4K recording.
The only program I know of that has floating notification is System Monitor (paid version), I'm sure there is a free program that does this though I'm not sure myself.
had that issue.i have underclocked my CPU to 1.7 per core. after that i filmed 15 minute file, rec was stopped beacause my sd is full.
Pesetio said:
Ahh ok I thought you had one lol
You'll enjoy it when you do though
Click to expand...
Click to collapse
Initial impression
progressfx said:
had that issue.i have underclocked my CPU to 1.7 per core. after that i filmed 15 minute file, rec was stopped beacause my sd is full.
Click to expand...
Click to collapse
Good to see that it worked for you, I think undervolting plays a part but the CPU usage when recording 4K plays a major part, if you get very high CPU usage your phone will heat up very quickly and you won't get long recordings, however if you get low CPU usage (even @ high frequencies) you will get very long video recordings and phone will not be too hot.
One Twelve said:
Initial impression
Click to expand...
Click to collapse
I saw your post, it's actually better to record to the internal storage because it's apparently faster, however all of my tests were conducted onto the externalSD.
I did some more tests, this time I turned on my heater to try to recreate a hot climate lol.
I had the phone running @ 1.6ghz with the floating CPU notification on top and heater blowing hot air (1 notch down from max) from about a meter away. The CPU usage was quite low, so I immediately new that the recording would go on for at least 30mins which is did. Phone didn't feel any hotter then usual. And like usual I could go on and record 4K again straight after for 30+mins.
I'm confident to say that it is the CPU usage that is the main culprit. High CPU usage @ lower frequency will not get a long recording compared to low CPU usage @ higher frequency.
Just check the CPU usage when recording 4K and you will know right away if you are going to get good or bad results.
I know there is a mod that increases the heat threshold but personally I wouldn't use it since the threshold is there for a reason.

Major Speed Increase

I have noticed this phone underperforms compared to other Snapdragon 821 devices and I have found a method to majorly increase the speed of this device.
After the following changes the device remains running at cool operating temperatures and performance is increased over the board about 50%
Doing these changes will disable the useless thermal limits set by Xiaomi on this device. Xiaomi set the limits far to low unlike other manufacturers.
The device still has hardware thermal protection by Qualcomm so there is no chance these changes will damage your device.
Your will need:
ES File Explorer
Kernal Auditor
Root Access
Open ES File Explorer and navigate to System/etc
Scroll to the bottom and find file thermal-engine-8996.conf
Copy this to your sd card for a backup and delete this file. ( This will not take effect immediately )
Go to Kernal Auditor and change the big core CPU Governor to OnDemand and set the Maximum CPU frequency to 2150MHz.
Set the LITTLE core CPU Governor to interactive and Max Frequency to 1593MHz.
Go to GPU and change the GPU Max Frequency to 560MHz.
Restart your device for these changes to take effect, the changes are astounding and matches the performance of a stock Snapdragon 835 device while still running at standard operating temperatures.
are u sure, we will have same thermal sensation ? i mean in hand ?
Wont it make the device very hot and will have very fast battery drain?
Sent from my Xiaomi MIX using XDA Labs
The device doesn't get any hotter than usual with these settings.
It is important to under clock a bit so it can keep this sustained performance without breaching it's thermal limits.
The performance gains are very noticeable especially with Dolphin and PPSSPP.
You can see how hot the device gets with Kernal Auditor. Compare the heat listings before the changes and after these are done to check for yourself.
As long as you use these settings the device will not get hotter than stock.
omerbagi10 said:
Wont it make the device very hot and will have very fast battery drain?
Sent from my Xiaomi MIX using XDA Labs
Click to expand...
Click to collapse
The device doesn't get hot if you underclock the cores and battery life is even better because you are underclocking the GPU and LITTLE core which don't need to go up to 2.1GHz for standard tasks.
NGP-1001 said:
The device doesn't get hot if you underclock the cores and battery life is even better because you are underclocking the GPU and LITTLE core which don't need to go up to 2.1GHz for standard tasks.
Click to expand...
Click to collapse
So there are no downsides? :silly:
Sent from my Xiaomi MIX using XDA Labs
There are always down sides to system modifications that haven't been extensively tested, his mileage may not be what you get - be very careful changing core restrictions / settings.
I find there can be considerable variation based on the ROM. I have found EPIC to give the most consistent results. By results I mean Antutu, since general performance is hard to measure.
I set the cores to their max and the boost frequencies to max too. Governor is on demand for both. Response after touch is set to 100ms. In the thermal settings I have Temperature Throttle on and the GPU governor to performance. I get 155k In Antutu almost every pass. Sometimes it'll top 160k, other times down to 150k.
I don't care about the heat since I almost always use the included case and I'm not concerned with battery life. The main thing that still bugs me is the temp figure in Kernel Auditor. It'll show high 60s after Antutu, but other temp apps like CpuTemp show it never gets above 45C. What does your temp get up to?
From what I've read, even in the 60C+ range it's still within safe/reasonable working temps for the CPU. I've used these settings for around two weeks​ and I've gone out of my way to game when I rarely did before (sometimes 1-2 hours in PPSSPP which taxes it). Never had it freeze/lock/reboot once. My Nexus 6 would occasionally do that when I had it pushed to the limit.
If I didn't use the case then I probably would back off it a little as you have outlined as the temps when holding the bare phone can be pretty hot.
gavin19 said:
I find there can be considerable variation based on the ROM. I have found EPIC to give the most consistent results. By results I mean Antutu, since general performance is hard to measure.
I set the cores to their max and the boost frequencies to max too. Governor is on demand for both. Response after touch is set to 100ms. In the thermal settings I have Temperature Throttle on and the GPU governor to performance. I get 155k In Antutu almost every pass. Sometimes it'll top 160k, other times down to 150k.
I don't care about the heat since I almost always use the included case and I'm not concerned with battery life. The main thing that still bugs me is the temp figure in Kernel Auditor. It'll show high 60s after Antutu, but other temp apps like CpuTemp show it never gets above 45C. What does your temp get up to?
From what I've read, even in the 60C+ range it's still within safe/reasonable working temps for the CPU. I've used these settings for around two weeks​ and I've gone out of my way to game when I rarely did before (sometimes 1-2 hours in PPSSPP which taxes it). Never had it freeze/lock/reboot once. My Nexus 6 would occasionally do that when I had it pushed to the limit.
If I didn't use the case then I probably would back off it a little as you have outlined as the temps when holding the bare phone can be pretty hot.
Click to expand...
Click to collapse
Kernal Auditor gets the actual CPU/GPU temps , I'm not sure where apps like CPUZ or CPU Temp get their readings.
Even after running Dolphin or PPSSPP for a couple hours the CPU temp gets into the low 70's but that's as high as it goes.
These chips are safe to run up to 95 C for CPU , 100 GPU and 110 C for the battery.
Changing these settings keeps the device well within it's thermal limits.
I have had my phone set like this for months after running it many other configurations.
I just had to tell everyone about it because it beats my buddies OnePlus 3T quite a bit when it comes to gaming.
omerbagi10 said:
So there are no downsides? :silly:
Sent from my Xiaomi MIX using XDA Labs
Click to expand...
Click to collapse
Give it a try and see for yourself.
I recently upgraded to the latest epic ROM version and found my phone to be running extra slow and battery wasn't lasting as long.
Turns out the update had reverted my settings so I decided to post this thermal "fix"
I highly recommend everyone make this change really , there is a reason that OnePlus decided to underclock their OnePlus 3T and why it runs much faster than the Xiaomi Mi Mix before this fix is completed.
It's all because the thermal throttle kicks in much too early on the Mix due to the extra heat at max Mhz for the GPU and large cores.
The only way to change this and get the performance enhancements is the remove the thermal file as I posted and follow the steps to ensure good thermals.
Thanks. Done here. It got REALLY REALLY fast but so warm for now. I'll wait to see if it get colder...
Sent from my MI MIX using Tapatalk
Tried these mods right now.
Simply wonderful.
Antutu from 130.000 to over 157.000
Increased speed
Now all is very smooth
But the temperatures, incredibly, are gone down...this smartphone now is more fast and more cold.
Xiaomi/Qualcomm thermal control is so bad?
But there's a strange issue with Kernel Adiutor: I set the frequency, governor etc but, after reboot, I see these settings returning to original values. Is it normal?
{
"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"
}
alberto8523 said:
Tried these mods right now.
Simply wonderful.
Antutu from 130.000 to over 157.000
Increased speed
Now all is very smooth
But the temperatures, incredibly, are gone down...this smartphone now is more fast and more cold.
Xiaomi/Qualcomm thermal control is so bad?
But there's a strange issue with Kernel Adiutor: I set the frequency, governor etc but, after reboot, I see these settings returning to original values. Is it normal?
Click to expand...
Click to collapse
I also have this problem.
I have set everything and the "Apply on boot" is on but everything is going back to the original values.
Sent from my Xiaomi MIX using XDA Labs
alberto8523 said:
Tried these mods right now.
Simply wonderful.
Antutu from 130.000 to over 157.000
Increased speed
Now all is very smooth
But the temperatures, incredibly, are gone down...this smartphone now is more fast and more cold.
Xiaomi/Qualcomm thermal control is so bad?
But there's a strange issue with Kernel Adiutor: I set the frequency, governor etc but, after reboot, I see these settings returning to original values. Is it normal?View attachment 4181218
Click to expand...
Click to collapse
Really? It was so warm here. Like 80 degrees celsius. Would it be cause im using RR?
Sent from my MI MIX using Tapatalk
I had the same temperatures with miui, RR and Lineage.
If I delete that file in every rom, i think I have the same results: more speed, less temperatures
I will try this settings on My MIUI ROM...
@ all...
Tips to have Kernel Adiutor to apply @ boot value (on MIUI ROM i mean)
Go to security center and manage permissions for this apps like auto startup authorized after restart i see a notification with time which say, i will aplly your value in xx seconds... and it works...
@NGP-1001
Thx for your feedback and tips, will make some test on my daily usage and send my feedback too...
Cheers
NGP-1001 said:
Even after running Dolphin or PPSSPP for a couple hours the CPU temp gets into the low 70's but that's as high as it goes.
Click to expand...
Click to collapse
Dolphin the browser? The only other Dolphin I know of is the one for GC/Wii emulation.
Low 70s is as high as mine gets too but it can get uncomfortably warm without the case. Maybe it's just when the CPU and GPU are both being pushed that the heat builds more. If I just use a CPU-only benchmark, like GeekBench, the phone only ever gets moderately warm at most.
I was really disappointed with the general stutter and performance (just in the UI, not even gaming) before I noticed your old thermal post. I can't recommend this method enough. Getting rid of the thermal.conf makes a massive difference no matter if you underclock it or push it to the max, even if you're taking a hit on battery life.
For those on the latest EPIC, the integrated Kernel Auditor is gone, so it needs installed from the Play store. Even then it won't work on boot like the old one did. You need to go into Settings > App Settings > Permissions > AutoStart and toggle it on.
Great result with performance governor
gavin19 said:
Dolphin the browser? The only other Dolphin I know of is the one for GC/Wii emulation.
Low 70s is as high as mine gets too but it can get uncomfortably warm without the case. Maybe it's just when the CPU and GPU are both being pushed that the heat builds more. If I just use a CPU-only benchmark, like GeekBench, the phone only ever gets moderately warm at most.
I was really disappointed with the general stutter and performance (just in the UI, not even gaming) before I noticed your old thermal post. I can't recommend this method enough. Getting rid of the thermal.conf makes a massive difference no matter if you underclock it or push it to the max, even if you're taking a hit on battery life.
For those on the latest EPIC, the integrated Kernel Auditor is gone, so it needs installed from the Play store. Even then it won't work on boot like the old one did. You need to go into Settings > App Settings > Permissions > AutoStart and toggle it on.
Click to expand...
Click to collapse
O, i need agree, device is much smoother then before, if u don't play, u will have a standard temp ° but i try to play Battle Bay... Got damn, baby...
Hot like a woman in video clip from old west coast sound... 84.2°... based on demande request, if u play, device come hot, other it's cold...
Will try more time to check if i can adjust temp limit to reduce speed based on temp limit...
Cheers
If anyone is still getting higher temps I would try underclocking a bit more. You will still get the speed increase even at much lower MHz. I wouldn't use the performance Governor as it's a huge battery drain.

Categories

Resources