(I intended to post this in "general:, not q&a)
SUMMARY: I have been slowly decreasing my voltage. Wednesday, three weird things happened on my phone that never happened before (wakelocks, unintended reboot, instant drop in battery indication).
CONCLUSION: I attribute all three things (wakelocks, reboot, instant drop in battery indication) to reaching an unstable voltage.
DESIRED OUTCOMES (Why am I posting this):
1 - I think I understand what happened, but maybe you guys can point out if I'm mistaken.
2 - Maybe someone else will be interested in seeing voltage limits for my phone (-75mV appears to work, -100mV appears to be too low for me... I guess other Infuses may act differently)
3 - It seems to me that maybe excess UV should be considered when investigating causes of wakelocks (there are many other causes of course). This conclusion is based only on this one experience as reported below. I have heard obviously unintended reboot can be associated with too much UV'ing, but I never heard wakelocks could be associated with UVing. Open to comments.
DETAILS:
I’m using stock GB, rooted, Zen’s Infusion A/1600 kernel, with governor conservative 100-1200millivolts.
I have been decreasing my voltage settings by 25 millivolts every other day over the last week.
Wednesday, I got to 100 millivolts below max on all points (except 1600Mhz, which was 50 millivolts down)
Specifically, my levels on Wednesday were:
100 Mhz: 950mv max – 100mv = 850 millivolts
200 Mhz: 950mv max – 100mv= 850 millivolts
400 Mhz: 1050mv max – 100mv = 950 millivolts
800 Mhz: 1200mv max – 100mv = 1100 millivolts
1200 Mhz: 1275mv max – 100mv = 1175 millivolts
1600 Mhz: 1400mv max – 50mv = 1350 millivolts
Before Wednesday, I had performed stability test (using “stability test” program for about 30 minutes... it sweeps the frequencies I think) at 1600Mhz Fmax while 50 millivolts down and another stability test at 1200 Fmax while 75 millivolts down, both for around 20 minutes, no problems there and no other problems with my phone in the last week since I upgraded to GB. When I decreased to 100 millivolts down I only did a 30-second “stress test” from within setcpu at 1200Mhz. No problems there.
All day Wednesday I was running the conservative governor, with Fmin=100 and Fmax = 1200, cfq io scheduler, UV settings as above (all 100 down except 1600Mhz). I have one setcpu profile that puts the phone to 1600Mhz Fmax when I manually launch "Memento Database" program, but I did not launch that program at all on Wednesday, so I never got above 1200Mhz on Wednesday.
My battery profile for the day Wednesday is shown in attached graphic.
The label “phone awake with screen off” points to a time period approx 2-4pm when my phone was awake with the screen off. I have never seen this happen on my phone before, and I look at this screen pretty regularly.
I used the phone heavily during my hour-long ride home from work from 5pm to 6pm (no problems). I was surfing the internet, reading the news about the Hurricane in New Orleans (not tweaking my phone). Phone worked absolutely perfectly throughout this hour long heavy usage.
First thing on getting home, I connected to wifi and surfed for about 5 minutes. At about 6:10pm, while surfing with my battery level happened to be 47% (as I deduced later), suddenly appeared the black screen with Samsung logo ... phone booted up to normal home screen. I think that’s what you guys call the “screen of death”?
I assume the phone reboot may be my first indication that I’ve reached my limit for UV’ing?
But here’s what concerned me more than the SOD:
Immediately after rebooting, my battery now showed 8% !
How does it go from 47% to 8% that fast? I have several screens from “battery monitor widget” and from the Android battery screen, and they all confirm the battery dropped from 47% to 8% over this very short time period (less than 5 minutes). I also have audible low-battery warnings on my phone (generated by Tasker, a lass with a pretty British lady voice) at levels of 40%, then 30%, then 20%, then 10%. I didn’t get any of those audible warnings prior to this spurious reboot – seems to confirm either this was a real drop in battery, or else my battery signal feeding all my applications (battery monitor widget, Tasker, GB battery profile) somehow went bezerk.
I have a hard time imagining this could be a real drop in battery ....where would the energy go? The phone did not get hot.
Is it possible somehow the phone’s battery calibration got confused?
Attached is a screenshot of battery drop.
The rapid drop from 70% to 50% was when I was using the phone heavily for an hour... I consider that normal.
The vertical drop from 47% to 8% is what is completely unexplainable to me. Although I have seen in the forum reports of battery gage reading unreliably in certain circumstances.
By the way, I put my UV setting back to -75 Wed night, and phone working fine ever since. No wakelocks or other anomolies.
Questions:
1 – Is this what is known as “screen of death”
2 – Has anyone seen that type of rapid battery level decrease?
3 – What do you think caused it.... calibration problem or actual loss of battery? If calibration problem, it’s kind of weird that the voltage increased smoothly during recharge afterwards?
4 – Do you think the phone-awake-with-screen off occurences earlier in the day are related to my UV'ing? I didn't see them before and I haven't seen them since went back to -75mV, suggesting they are caused by UV'ing. But also note the phone worked fine during my hour long drive home after the wakelocks and before the SOD
5 – Do you agree I have probably reached stability limit and should stick with 75 millivolts down instead of 100 (I have been back on 75 again for awhile...no problems)
Attached below is my battery voltage over the course of the day, annotated with the items discussed above.
If you undervolted, your going to run into problems man. Simple as that. If you do anything to your phone not stock these things can and will happen.
Just take is as a fluke and move on. Everyone's phone uv's differently.
Imo what happened is the processor was making lots of errors and corrupted some part about your batteries actual percentage and was a lot lower than was presented to you.
Phone awake while screen on is a wakelock and quite common. Usually an app holds the phone turned on.
Edit: if uv of 75 fixes the problem. Then that definitely us what you should do. Note, benchmarks don't give stability reports. You would need run them for hours before you get accurate assessment.
Sent from my SGH-I997 using xda premium
1 – SOD is when the screen is black and phone won't unlock. Sometimes the volume buttons work. A long press of the power button will usually reboot the phone.
2 – The rapid power drains I've seen have typically been due to the rild process running amok and maxing the cpu. Typically the phone gets hot when this happens. In a SOD situation, you can typically still connect to the phone via ADB. If you can do this, run "adb shell" and then the following command: "top -m 5 -n 1" (spacing emphasized as all spaces are required).
3 – Dunno, but I'm with Elliot: the thing you changed was UV. Back that out and see if the problem goes way.
4 – Awake when screen off is usually related to an app that sets a wakelock and syncs in the background. Use Better Battery Stats to troubleshoot this. If it was UV causing this, I'm not sure why that would be - see #3
5 – FWIW, I never UV below -75... and truth be told I haven't seen a significant savings in battery life with UV, but that's likely because my phone is asleep most of the time...
Thanks for your response Elliot.
I think you interpretted it roughly the same as me except you lean more towards other possible causes of wakelocks unrelated to UV. That is certainly possible and I will keep a close eye on tha (moreso as a reuslt of your comments)
I am moving on with my phone at –75mv and don’t plan any drastic changes as a result of this unless I see more problems, but I also want to treat this as a learning experience. I spent awhile collecting the data and drawing my own conclusions, but wanted to bounce it off the others here for my own long-term learning.
Thanks again.
Thanks Zen, good info as always.
if you UV, i had better luck with higher UV on the top end:
ex: -125 or 150 @ 1600
-100 or 125 @ 1200
-75 or 100 @ etc
Basically, you are limiting power consumption and increasing battery life.
Once you hit the lower limit, your phone will let you know.
If you care to try, and you will have a slower phone:
set cpu max to 800 and then uv -75, -50, -25 stepping down. you will see the difference. it will make calls or mms fine. But you will notice slowdown with web/games/etc
every cpu is different. i've had 2 infuses because my screen broke, i replaced it and the replacement was defective, found another phone for about the price of a replacement screen.
one get's me about -100mv the other was completely stable at -225(maybe more) with stock clocks and -200 overclocked.
one thing that can cause instability is a big voltage differential from low to high especially with on demand governor, it's not only a matter of necessary volts to drive the chip per clock, the rapid fluctuations can cause issues. if you want performance you might want to uv from the top down leaving the lower freqs stock and seeing how much uv it will tolerate at 1600 before you mess with lower clocks, you might get more uv on the upper clocks that way which will reduce heat. if you want battery life then try less or no overclock still going from the top down.
for people that can't reach 1600, or think they can't try to set 1400 and 1600 at the same voltage so the voltage table has a plateau at the top. setting 1200 and 1400 high might smooth the transition to 1600, also try the voltages across the board, there may be a narrow range where the heat and required voltage balance out. i did this with my captivate and got it from 1200 to 1300mhz but 1300 would only work in a narrow 50mv range and 1200 needed to be enabled as well as set to the same voltage. the chip in that phone didn't like overclock at all compared to most infuses but sometimes thats how it goes.
still have wakelocks
if you UV, i had better luck with higher UV on the top end:
Click to expand...
Click to collapse
one thing that can cause instability is a big voltage differential from low to high especially with on demand governor, it's not only a matter of necessary volts to drive the chip per clock, the rapid fluctuations can cause issues.
Click to expand...
Click to collapse
Interesting. Both comments end up suggesting the same strategy to my way of thinking (more UV'ing at top frequencies will create a flatter voltage curve across the range of frequencies). I'll give it some thought when I get back to polishing my uv strategy. Maybe instead of putting all back to -75 I should do something like:
100 Mhz: 950mv max – 75mv = 875 millivolts
200 Mhz: 950mv max – 50= 900 millivolts
400 Mhz: 1050mv max – 75mv = 975 millivolts
800 Mhz: 1200mv max – 100 mv = 1100 millivolts
1200 Mhz: 1275mv max – 75mv = 1200 millivolts
1600 Mhz: 1400mv max – 100mv = 1300 millivolts
...and stick with conservative governor to avoid the big jumps. That way I'd never jump more than 100millivolts (except between 400 and 800 which jumps by 125... but was apparently already a 150mv jump in the stock?). But I'd want to do some testing on the at 1600 to see how it likes the 100mv down, first. Sound reasonable?
========
New subject
An "update" on my phone:
My battery display still shows today that there is quite a lot of time with cpu on and screen off (wakelock) even though now my phone is not undervolted at all today (I removed undervolting to do some testing and haven't put back the uv'ing yet).
I'm heading over to the market to download Better Battery Stats per Zen's suggestion.
Thanks
electricpete1 said:
Interesting. Both comments end up suggesting the same strategy to my way of thinking (more UV'ing at top frequencies will create a flatter voltage curve across the range of frequencies). I'll give it some thought when I get back to polishing my uv strategy. Maybe instead of putting all back to -75 I should do something like:
100 Mhz: 950mv max – 75mv = 875 millivolts
200 Mhz: 950mv max – 50= 900 millivolts
400 Mhz: 1050mv max – 75mv = 975 millivolts
800 Mhz: 1200mv max – 100 mv = 1100 millivolts
1200 Mhz: 1275mv max – 75mv = 1200 millivolts
1600 Mhz: 1400mv max – 100mv = 1300 millivolts
...and stick with conservative governor to avoid the big jumps. That way I'd never jump more than 100millivolts (except between 400 and 800 which jumps by 125... but was apparently already a 150mv jump in the stock?). But I'd want to do some testing on the at 1600 to see how it likes the 100mv down, first. Sound reasonable?
========
New subject
An "update" on my phone:
My battery display still shows today that there is quite a lot of time with cpu on and screen off (wakelock) even though now my phone is not undervolted at all today (I removed undervolting to do some testing and haven't put back the uv'ing yet).
I'm heading over to the market to download Better Battery Stats per Zen's suggestion.
Thanks
Click to expand...
Click to collapse
in my experience that will probably work well. but if the phone is at it's limits with the 1600mhz clock then you might have a specific voltage that it prefers so it's possible it doesn't work into what i was describing. try it, i can't say uv is safe but i haven't broken a phone yet by driving the voltage too low. it just tends to crash and will be fine on the next boot after you clear the settings in my experience so far.
Thanks. I went ahead and made the above changes in voltage. Then I manually rebooted (to allow voltage changes to take effect), went into setcpu to set my max frequency to 1600.
Successful stress test from within setcpu ~ 50 seconds.
Then went to the application named "Stability Test", selected "scaling" (which is supposed to vary the frequency) and has been successfully running the test for 14 minutes (11 successful cpu runs and 75 successful RAM runs). It has been on 1600Mhz the whole time according to the Stability Test display. The battery temperature increased from 37 to 42 pretty quickly then slowed, looks to be stable at 43C now.
Elliot warned above that it takes stability tests hours to find problems and Stability Test documentation mentions something about 5 hours on one frequency to find a problem. But if it's only running at one frequency as it appears from the display (stuck on 1600), I think even after running it for 5 hours I might not be confident that it will ferret out all the problems that might arise while switching frequencies (and I'm not sure I want to leave it running for 5 hours if it's stuck on 1600). I'm not sure exactly what is the purpose of "scaling" mode is, if it doesn't change frequency.
At any rate, I''m going to terminate the test and give it the real-life test now.
I will leave my governor at 100-1600 conservative governor for a few days to help shake out any glitches hiding inside there.
Will let you know if anything goes wonky.
Later on when I'm confident in results, I plan on setting up profiles for 100-400, 100-800, 100-1200, 100-1600 depending on the situation, but will stick with conservative in all cases.
Thanks again.
I wanted to provide an update of testing I have done with UV on my phone. Nothing conclusive for me yet, just a lot of data.
I’ll discuss 5 different UV settings labeled below as 0, A, B, C, D (going from most UV to least UV).
About notation - I think it is helpful to list UV settings in the full format I use (rather than just how many volts down) for two reasons:
1 – The reference voltages vary depending on kernel. So it could be misleading to report what the reduction in voltage is without reporting the reference. At 1200Mhz, Zen’s Infusion A/B uses 1275millivolts while Entropy’s DD uses 1300millivolts (voltages match at other frequencies). I’m not sure what the voltages are in stock kernel.
2 – Listing the final voltages allows you to see how much jump in voltage there is from one frequency to the next (large jump could possibly increase instability)... this aspect is not as obvious if we just list how many volts below the reference.
Setting set 0
100 Mhz: 950mv max – 100mv = 850 millivolts
200 Mhz: 950mv max – 100mv= 850 millivolts
400 Mhz: 1050mv max – 100mv = 950 millivolts
800 Mhz: 1200mv max – 100mv = 1100 millivolts
1200 Mhz: 1275mv max – 100mv = 1175 millivolts
1600 Mhz: 1400mv max – 50mv = 1350 millivolts
Duration: Tested less than one day.
Results: Reported in original thread: caused unwanted reboot while surfing. After that the battery indicator jumped.
Setting set A.
100 Mhz: 950mv max – 75mv = 875 millivolts
200 Mhz: 950mv max – 50= 900 millivolts
400 Mhz: 1050mv max – 75mv = 975 millivolts
800 Mhz: 1200mv max – 100 mv = 1100 millivolts
1200 Mhz: 1275mv max – 75mv = 1200 millivolts
1600 Mhz: 1400mv max – 100mv = 1300 millivolts
Duration: Tested 3 days heavy use
Results:
* “Rainbow Screen After Trap Game” – I tried the game “Trap” which was recommended in a thread somewhere as being a good test to find stability problems using UV. Sure enough upon exiting the game once, my screen turned to rainbow colors. I rebooted to restore screen to normal (same UV). I played the game for 15 minutes exiting a few more times, no more problems.
* “Widgets Lost” - Once after reboot I lost all my widgets (Tasker, Color Notes, Elixir). They did not come back even when I went to stock voltages and rebooted. Had to manually restore each of my widgets.
* “Discolor Upon Screen On” - Once in awhile when turning on screen, some colors are initially wrong (mostly I notice the white labels of my icons are yellow or pink). It goes away very quickly - either after I touch the screen or after scroll around a little. After that everything is normal.
* “Flash Upon Screen On” - Once in awhile when turning on the screen, I see a white flash for a fraction of a second (sort of like with an old TV tube when you turn it on or off). After that everything is normal.
These last two symptoms (“Discolor Upon Screen On” and “Flash Upon Screen On”) occurred probably once per day over the course of the three days relatively heavy usage.
Setting set B
100 Mhz: 950mv max – 25mv = 925 millivolts
200 Mhz: 950mv max – 25= 925 millivolts
400 Mhz: 1050mv max – 25mv = 1025 millivolts
800 Mhz: 1200mv max – 75 mv = 1125 millivolts
1200 Mhz: 1275mv max – 50mv = 1225 millivolts
1600 Mhz: 1400mv max – 75mv = 1325 millivolts
Duration tested: 3 days heavy use
Results:
* “Discolor Upon Screen On” – Same as above. 2 or 3 times over three days heavy use
* “Flash Upon Screen On” – Same as above. Once over three days.
setting set C
100 Mhz: 950mv max – 15mv = 935 millivolts
200 Mhz: 950mv max – 15= 935 millivolts
400 Mhz: 1050mv max – 15mv = 1035 millivolts
800 Mhz: 1200mv max – 50 mv = 1150 millivolts
1200 Mhz: 1275mv max – 25mv = 1250 millivolts
1600 Mhz: 1400mv max – 50mv = 1350 millivolts
Duration Tested: 3 days heavy use
Results:
* “Discolor Upon Screen On” – Same as above. 2 or 3 times
(Did not have the Flash upon screen on)
setting set D = 0 undervolt – use Zen’s default settings.
Duration: about 2 weeks experience with GB in this configuration
“Discolor Upon Screen On” – Never saw it
* “Flash upon screen on” – Never saw it, but did see a similar flash just once in my phone application when I removed phone from ear to look at keypad.
** EDITED TO ADD - Today after posting this, I noticed the flash upon screen on" while using no UV, 100-1200Mhz. Maybe I just wasn't looking for it before.
Other Testing Using Specific Applications:
I did a lot of testing with these programs on settings A, B, C and saw no anomalies. For settings A, B, and C, did the following:
SetCPU “Stress Test” for approximately one minute
“Stability Test” Applicaiton– “Scaling Stability Test” subtest for approx 20 minutes (battery gets up to about 115F).
“Stability Test” Applicaiton – “CPU+GPU Stability Test” subtest for approx 10 minutes (actually did this only in B, C, didn't try it in A)
I played the game "Trap" (that caused the problem in setting A) for 15-20 minutes in B, C, exiting and reentering a few times. Didn't see any problems.
Interestingly, when I run Quadrant it sometimes flakes out (returns to Quadrant start screen without any result) and somtimes run to conclusion. But this happens even on stock voltages, and I think I remember it's a problem with Quadrant that others have seen, so I’m inclined to think it is not an indicaion of instability I should worry about.
Analysis:
It looks like B is the most UV I can apply without significant problems appearing during 3-day test. But I still had some minor anomalies I notice occasionally when turning screen on (brief flash and brief discoloration),.but they occur to certain extent even at less UV (C) and at zero UV (D)
All of the above real-life testing in A, B, C was done with max frequency of 1200 (because I wanted to rule out 1600 as a problem).
I used 1600 only during the specific stability test appliations.. that ran fine and also in the D configuration that ran fine. So the problems (if they are problems) come from UV, not from using 1600Mhz.
Questions:
Do you think discoloration upon turning on the screen or flash upon turning on is cause for concern? The symptom itself is not an inconvenience but I’m wondering if it’s warning me about something else going on...(***)
I don’t really have a feel for what change in battery life occurs with these different settings. Seems very difficult to quantify. Any thoughts?
By the way, I'm using ULCB3 stock rooted GB. Using Zen's Infusion-A kernel. cpu control using setcpu: Mostly 100-1200Mhz cpu frequency, conservative cpu governor with the Up/Down threshholds tweaked to 95/45, noop io scheduler.
I installed this configuration about a month ago using qkster's Heimdall one-click.Never had any problems other than those mentioned above which may or may not be problems.
*** I have to confess, I wiped Davlick cache and wiped the other cache at that time, but never did the "factory data wipe" because I had data on my phone that wasn't backed up. I'm not sure if those particular instructions required factory data wipe or not but I remember it was mentioned as good practice. It occurs to me maybe these subtle brief strange symptoms upon turning on the screen are remnants from incomplete wipe?
I went back to stock and still had the occasional occurence of discolored screen upon waking from deep sleep.
It seems harmless, goes away if cycle the screen.
So I re-established UV. I have had my phone at the following settings since 9/17/12. No problems for me other than that occasional discoloration or flashing upon waking for sleep and one other problem where the phone would occasionally ring and I could see the call but not hear the caller (they could hear me). I eventually traced that to tweaking my governor settings... will post that in another thread.
100 Mhz: 950mv max – 50mv = 900 millivolts
200 Mhz: 950mv max – 25= 925 millivolts
400 Mhz: 1050mv max – 50mv = 1000 millivolts
800 Mhz: 1200mv max – 75 mv = 1125 millivolts
1200 Mhz: 1275mv max – 50mv = 1225 millivolts
1600 Mhz: 1400mv max – 75mv = 1325 millivolts
electricpete1 said:
I went back to stock and still had the occasional occurence of discolored screen upon waking from deep sleep.
It seems harmless, goes away if cycle the screen.
So I re-established UV. I have had my phone at the following settings since 9/17/12. No problems for me:
100 Mhz: 950mv max – 50mv = 900 millivolts
200 Mhz: 950mv max – 25= 925 millivolts
400 Mhz: 1050mv max – 50mv = 1000 millivolts
800 Mhz: 1200mv max – 75 mv = 1125 millivolts
1200 Mhz: 1275mv max – 50mv = 1225 millivolts
1600 Mhz: 1400mv max – 75mv = 1325 millivolts
Click to expand...
Click to collapse
1. discolored screen: zen told me about these. I've not seen it..no known harm effects though.
2. I've pushed the 1600 to -125mV; 1200 to -100mV without issues.
Related
Just started playing around with UV/OCing tonight and am pretty happy so far with the results. I'm just currently looking at getting some extra battery life with UVing.
Just curious what everyone else has achieved.
And yes I know, and everyone else should know, that just because someone else was able to get a certain OC/UV doesn't mean you will. So is the game of OCing anything.
Please post the full details of what you have achieved and/or what you run as a daily driver.
In this format:
(My current results, so far)
1120Mhz @ -125
1000Mhz @ -125
900Mhz @ -125
800Mhz @ -125
600Mhz @ -125
400Mhz @ -125
200Mhz @ -125
100Mhz @ -125
Rom: Continuum v4
Kernel: Onix - 2.0.1 #7
(I'm curious what the lowest voltage has been attained at stock/1120 clock)
bnevets27 said:
Just started playing around with UV/OCing tonight and am pretty happy so far with the results. I'm just currently looking at getting some extra battery life with UVing.
Just curious what everyone else has achieved.
And yes I know, and everyone else should know, that just because someone else was able to get a certain OC/UV doesn't mean you will. So is the game of OCing anything.
Please post the full details of what you have achieved and/or what you run as a daily driver.
In this format:
(My current results, so far)
1120Mhz @ -125
1000Mhz @ -125
900Mhz @ -125
800Mhz @ -125
600Mhz @ -125
400Mhz @ -125
200Mhz @ -125
100Mhz @ -125
(I'm curious what the lowest voltage has been attained at stock/1120 clock)
Click to expand...
Click to collapse
I have mine oc. to compensate with battery drain I was undervolting also here is my specs:
1400- 0 Quadrant - 2258 Linpack - 20.1
1000- 50
800 - 100
400 -125
200 -125
100 -150
Here's what I got
1400 @ -25 Quadrant ~1900 Linpack ~19
1200 @ -50
1000 @ -50 Quadrant ~1600 Linpack ~14
800 @ -75
400 @ -75
200 @ -100
100 @ -100
EDIT: I use Voltage control app, some people seem to have issues with set CPU
______________________
Samsung Captivate
ROM: designgear's Cognition v4.2.2
Kernel: eXistZ's KB1 #7 beta
i envy you guys that can run stable at those numbers. my results are
1280 -50
1000 -25
800 -100
400 -100
200 -50
100 -50
with less over clock i can get much lower voltages across the board and without over clock i can go lower still.
having the actual voltage the same for 1000 and 1280 seems to help stability. my chip deosn't seem to like voltage fluctuations. also the difference between the upper and lower voltage seems to be a problem. my phone is stable at -100+ across the board if i dont over clock. for that reason i tend to test things by undervolting from the upper frequencies down, if i go from the low frequencies i will see instability with very little undervolt untill the upper frequencies are also undervolted.
CM7 03182011, Kernel du Jour v4.2
Just UV, with governor set to "conservative"
1000: -100
800: -100
600: -125
400: -125
200: -150
100: -150
Cognition 4.3
Voltage Control App
Stability Test
Neocore
Linpack
Mhz UV Temp. Linpack
1300 -50 (20c/42c) 17.8~
1000 -50
800 -75
400 -100
200 -100
100 -100
Serendipity 6.3
Paragon 6 kernel @ 1344mhz
Linpack = ~18.8
1344 - 0
1000 - 50
800 - 75
400 - 100
200 - 125
100 - 125
I'm liking your numbers... I wonder if having 100 and 200 enabled makes a big difference. Eliminating 1200 and going from 1000 to 4000 seems quite a bit snappier.
Idk. Paragon kernel defaults the states.
jgrnght said:
I'm liking your numbers... I wonder if having 100 and 200 enabled makes a big difference. Eliminating 1200 and going from 1000 to 4000 seems quite a bit snappier.
Click to expand...
Click to collapse
Interesting thread...would be great to see what people are using for different roms..
I can't get 1400 stable, even at stock voltage.... but at 1300 I can get down to -75:
1300 @ -75
1000 @ -100
800 @ -100
600 @ -100
400 @ -100
200 @ -100
100 @ -100
Last time I tried going down to -125 it wasn't stable, but that was on a different kernel... gonna give it a try when I get a chance.
CM7 and The Escapist's kernel with Noop scheduler and 1300mhz limit enabled
1300 -50
1200 -75
1000 -75
800 -100
600 -100
400 -125
200 -125
100 -125
Nice to see everyone posting their numbers.
I'm finding that the UV setting I have are causing the phone to lock up once and a while. I did pass 50 passes of stability test and 30min of neocore. So are those programs really that accurate at testing stability? Is there anything better out there?
Playing Cordy for a bit seems to effectively stress the system, as I can get it to lock quickly even when seeming stable in the synthetic tests.
Cordy spends almost 100% of the time in top frequency state from what I've seen, plus its pounding the GPU.
bnevets27 said:
Nice to see everyone posting their numbers.
I'm finding that the UV setting I have are causing the phone to lock up once and a while. I did pass 50 passes of stability test and 30min of neocore. So are those programs really that accurate at testing stability? Is there anything better out there?
Click to expand...
Click to collapse
I've found that stability test is only accurate on the Mhz it is testing at. What i did was lock my Mhz at 1400, then 1300, then 1200, 1120, 1000, lastly 800. I had to have 5 frequencies ticked to get 800 to even work. I did each level to find the max UV for each level... then used those numbers.
I found having 1000 being my 2nd highest frequency, makes for a much snappier interface. Right now I have 1300, 1000, 800, 400, 200, 100.
Although I was able to use stability test all the way down to 800, I really couldnt test the lower frequencies with it. I found 125 was stable on the lowest 3 while doing benchmarking, however when moving back and forth quickly through settings options, mainly the drop down menus, it would cause it to lock up. So I raised the lower number up to 100 and no longer got crashes in option menus.
Here are the numbers I've settled with and get the quickest overall feel with.
noop scheduler.
1300 -50
1000 -50
800 -75
400 -100
200 -100
100 -100
Cognition 4.3 btw.
Tryng to underclock now. Blah, such a pain!
noop scheduler.
1200 -100
1000 -100
800 -100
400 -100
200 -100
100 -100
On Phoenix Ultimate Special Edition with Onix 2.0.3 kernal
No one else wants to brag?
Anyone aware of (or maybe could make?) an app that can test at each frequency? It would be very helpful to be able to test UV settings for frequencies in the 100-800 range. So far it seems the only "test" is once the phone locks up while navigating thew menus, bump up the voltage to all of the frequencies bellow 800.
(Thanks jgrnght for the suggestion on how to test each frequency 800+)
bnevets27 said:
No one else wants to brag?
Anyone aware of (or maybe could make?) an app that can test at each frequency? It would be very helpful to be able to test UV settings for frequencies in the 100-800 range. So far it seems the only "test" is once the phone locks up while navigating thew menus, bump up the voltage to all of the frequencies bellow 800.
(Thanks jgrnght for the suggestion on how to test each frequency 800+)
Click to expand...
Click to collapse
lol...nothing can beat with hands on experience...you can download cpu spy from the market it will tell you how your cpu stays in every frequency/states you have. From that data you will know which fequency needs more undervolting or tweaks.
Has anyone noticed a definite increase in battery life as a result of undervolting? Common sense points to that it would (less voltage needed, less battery power needed to power it) but does it really translate into solid real life results?
I'm hoping that all this UV business doesn't just give me an extra 15 minutes or something and that it actually translates into an hour or 2.
I got on CM7 TheEscapist's kernel:
1000: -125
800: -125
600: -125
400: -125
200: -150
100: -150
Scheduler: noop
Governor: Conservative
I have a few questions about OC/UV kernels or rather the theory behind them, particularly the UV part.
My natural instinct is to believe that when the voltage is dropped that the amps would increase so the CPU would still have the same amount of power to run the CPU at a particular frequency. This of course would negate any advantage in UV. I assume that my assumption is wrong or someone that really knows something would have pointed it out.
On the practical side of things, if the above is not true does the phone/cpu use the same amount of power if it is running at 216 Mhz with 770 mV as it does running at 400 Mhz with 770 mV. Same question would also apply to high frequencies that are UV and their relationship to heat generation.
Not sure about smartphones, but with my computer, I can drop 10 watts just from under volting when idle. Under clocking makes an even bigger difference. I'll get some exact numbers in a few mins, this got me wondering lol.
Edit:
At the wall my PC will pull 111 watts @ 1.6GHz idle. @ 4.5GHz it pulls 137 watts on idle.
I wish I had a link I could provide, but the basic science is this (learned this from Calkulin, Ziggy, and trial and error with my evo):
uV:
When you uV your cpu, each frequency step draws less power. For example, a non-uV'd cpu might draw 1300 mV at 1000 MHz whereas a uV'd cpu might draw 1225 or 1250 mV at 1000 MHz. There is no sacrifice in performance. You also get a cpu that runs cooler because it is drawing less power. The combination of these things means better batter life because you are drawing less power (obvious) and generating less heat (heat is one of the enemies of Li-ion batteries holding a charge).
But not all chips are the same, and some tolerate lower voltages better than others. If a particular voltage is too low, you get crashes, reboots, and other things like that because voltages that are too low make things unstable.
OC:
The theory behind OC is that the higher frequencies allow you to complete tasks faster. If the cpu can complete a certain task in a shorter period of time, it will be in use for less time overall, thereby allowing it to return to its idle frequency faster and stay in its idle frequency state longer. Again, this combination increases battery life because the power drawn from the battery is less. You can think of this as a "race to the finish" situation.
For example, if you OC to 1100 MHz, you might complete a task in 1.5s, whereas at 1000 MHz that same task might take 2s. The cpu actually works for a shorter time and returns to idle 0.5s faster. Add this up over an entire day, and your cpu actually spends more time at idle when OC'd than when not OC'd.
Again, just as with the uV, not all chips can tolerate maximum OC and just become unstable. That's why you've seen some G2x chips be able to run at 1500 MHz and be stable, while others can only tolerate 1300 MHz. Instability by trying to OC the cpu too much will lead to the same types of problems as lowering the voltages too far.
OC & uV together:
When you OC and uV, you allow the cpu to run at a higher frequency and draw less power at the higher frequency. This means you get tasks done faster and use less power to complete them, which allows the cpu to return to its idle frequency faster.
This is a real-world hypothetical: Imagine you have a cpu running at 1100 MHZ at 1300 mV and 1000 MHz at 1300 mV. When this is the case, your OC'd/uV'd cpu completes a task faster (in 1.5s rather than 2s) and actually ends up drawing less power than the non-OC'd/non-uV'd cpu because it completes the task faster and returns to idle sooner. When you add this up over an entire charge cycle, you should get better battery life.
So I'm not sure if this 4th G2x I received is a total beast or SetCPU is just f'ing with me, but I'm cruising at -200mV on top of the already undervolted Faux CM7 kernel.
The phone has absolutely no problems and it just keeps going.
Here's my setup:
1000Mhz Max 950 mV Current 925 mV (-25 mV)
800Mhz Max 850 mV Current 750 mV (-100 mV)
500Mhz Max 800 mV Current 650 mV (-150 mV)
400Mhz Max 770 mV Current 570 mV (-200 mV)
PS: Real men type it out
Edit: Poll added!
GideonX said:
So I'm not sure if this 4th G2x I received is a total beast or SetCPU is just f'ing with me, but I'm cruising at -200mV on top of the already undervolted Faux CM7 kernel.
The phone has absolutely no problems and it just keeps going.
Here's my setup:
1000Mhz Max 950 mV Current 925 mV (-25 mV)
800Mhz Max 850 mV Current 750 mV (-100 mV)
500Mhz Max 800 mV Current 650 mV (-150 mV)
400Mhz Max 770 mV Current 570 mV (-200 mV)
PS: Real men type it out
Click to expand...
Click to collapse
i really wonder sometimes if its really UVing with values so high x_X 216 running at 350mV x_X
{
"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 read in another thread by one of the developers (Faux?) that 770 mV is the lowest allowed voltage due to hardware restrictions and that any settings below that do nothing.
GideonX said:
So I'm not sure if this 4th G2x I received is a total beast or SetCPU is just f'ing with me, but I'm cruising at -200mV on top of the already undervolted Faux CM7 kernel.
The phone has absolutely no problems and it just keeps going.
Here's my setup:
1000Mhz Max 950 mV Current 925 mV (-25 mV)
800Mhz Max 850 mV Current 750 mV (-100 mV)
500Mhz Max 800 mV Current 650 mV (-150 mV)
400Mhz Max 770 mV Current 570 mV (-200 mV)
PS: Real men type it out
Click to expand...
Click to collapse
sincest said:
i really wonder sometimes if its really UVing with values so high x_X 216 running at 350mV x_X
Click to expand...
Click to collapse
You guys have amazing chips
The reason for 770mv @ 389/400 MHz is the limitation of ~100mv per "frequency jump" according to Tegra2 documentation. With the FakeShmoo implementation of overclock, we have a total of 8 slots (frequencies) to play with. If I start the lowest frequency let say at 500mv, then the max voltage at the eighth slot can only be ~1200mv. With 1200mv we can OC to 1.2~1.3 GHz, beyond that then it becomes unstable. So in order to get more juice @ higher clocks, we have to start a higher voltage at the low to get more juice at the high end. Of course, naturally, the next question would be why not start with 1000mv to begin with then you can get 1700mv at eighth slot? If we started with more juice at the low end, it would kill the battery 2x or 3x faster. So the current voltage table you see from my kernels is a compromise between battery life and OC Hope this clears it up a bit.
faux123 said:
You guys have amazing chips
The reason for 770mv @ 389/400 MHz is the limitation of ~100mv per "frequency jump" according to Tegra2 documentation. With the FakeShmoo implementation of overclock, we have a total of 8 slots (frequencies) to play with. If I start the lowest frequency let say at 500mv, then the max voltage at the eighth slot can only be ~1200mv. With 1200mv we can OC to 1.2~1.3 GHz, beyond that then it becomes unstable. So in order to get more juice @ higher clocks, we have to start a higher voltage at the low to get more juice at the high end. Of course, naturally, the next question would be why not start with 1000mv to begin with then you can get 1700mv at eighth slot? If we started with more juice at the low end, it would kill the battery 2x or 3x faster. So the current voltage table you see from my kernels is a compromise between battery life and OC Hope this clears it up a bit.
Click to expand...
Click to collapse
So there really isn't a hardware limit in place on how low this chip can go?
Sent from my HTC Vision (G2x in disguise)
I'm really starting to wonder if setcpu is busting our chops or not I'm
-100mV on all frequencies starting at 594mhz down to 216mhz on morfics jrcu 6/21 kernel, that can't be right can it is there a way to find out for sure???
for frequency 500mhz or 400mhz I can as low as I want ... 10mVs.. even 0mV. so i dont tihnk its accurate
c19932 said:
for frequency 500mhz or 400mhz I can as low as I want ... 10mVs.. even 0mV. so i dont tihnk its accurate
Click to expand...
Click to collapse
do you have any more than 100mV between frequencies?
Black6spdZ said:
do you have any more than 100mV between frequencies?
Click to expand...
Click to collapse
yes I do. should I not exceed 100mV between each frequency level then? would you happen to know the exact biggest voltage difference allowed between each level?
on another note, if I just leave them at 10mV, will that mean the processor will automatically take the voltage down to the lowest possible, or will it just do nothing but run at stock voltage?
c19932 said:
for frequency 500mhz or 400mhz I can as low as I want ... 10mVs.. even 0mV. so i dont tihnk its accurate
Click to expand...
Click to collapse
Yes, it seems possible to make settings that are not logically possible.
It would be nice if either SetCPU (which I am currently using) or Pimp My CPU would display in either real time or histogram what voltages are actually being used in the same way as they show frequency. Although I suspect this may be impossible due to hardware or this would be already implemented.
Added a poll up top.
I made it so you can vote either above or below -50mV. I'm basing this off the OC/UV kernel from Faux123.
I also skipped everything between 1.0 to 1.5Ghz. I'm going to assume at that speed, UV'ing will be very low.
double post
i actually use 1216max and uv it minus 150 all together so fauxs -50 plus i uv it -100, and my phone still flys and battery life is niceeee, oh and for the min i use 389, -200 on top of fauxs -50 so -250 all together
I am OC/UV up to 1200MHz -75mV on all frequencies below the undervolting built into Trinity JRCU kernel. This is stable and further UV gives me trouble. Your assumption about UV at higher frequencies is appears to be incorrect.
The poll actually confuses me, undervolting > -50 could mean I am not undervolting at all since 0 is greater than -50.
Interesting, I can UV more on the lower Mhz than higher. For example, at 1.0Ghz -50mV is already pushing it. Anything higher and I'll notice FC and random closes on certain apps.
First time polling, I guess there's no edit feature to change up the wording. Sigh.
SetCPU doesn't error check so everyone reporting -150mV and greater probably is incorrect. Remember there is only 800mV allowable from the highest frequency and only 100mV allowable between frequencies. So.. take my phone for instance.. I started at the highest 1.5Ghz frequency and tested -25mV at a time until phone rebooted and then went back up 25mV. This just happens to be 1300mV stable for me. This means the lowest 389Mhz can have at the lowest 500mV "1300mV - 800mV". I set everything -50mV across the board and then focused on 503Mhz since no matter what 389Mhz can only be -100mV lower. I was able to get -75Mv "725mV" stable. Then I was able to run 389Mhz stable at -150mV or 620mV. In actuality it is running 625mV since that is 100mV less than the frequency above it... setting it any lower in SetCPU would STILL only be 625mV regardless.
Also, this poll will be useless for everything but the 1500Mhz frequency as we wouldn't know if the user started with the SV or UV kernel. We need multiple polls with the final voltage eg 1100Mhz @ 1000mV,975mV,950mV
Test with a charger plugged in, I've had reboots with a battery at <50% because it cannot supply the high current draw of 100% cpu usage.
My specific settings are as follows as -?? means nothing because different kernels have a different starting point.
1075 mV @ 1200 MHz
975 mV @ 1100 MHz
875 mV @ 1000 MHz
775 mV @ 594 MHz
675 mV @ 432 MHz
575 mV @ 216 MHz
This is solidly stable having run it for several days without any bad behavior although any further reduction of voltage at 1200 MHz causes significant problems.
Phone: G2X
ROM: CyanogenMod 7.1 RC1
Kernel: Trinity 15 jrcu 06212011
Other relevant: SetCPU
I've gotten even better results with the SV kernel. After testing each speed for stability at lower and lower voltages here are my results:
1408Mhz @ 1175mV
1216Mhz @ 1075mV
1100Mhz @ 975mV
1015Mhz @ 900mV
816Mhz @ 800mV
655Mhz @ 725mV
503Mhz @ 650mV
389Mhz @ 595mV
lose 100Mhz from the UV/OC kernel but I get better UV settings with the SV kernel
Black6spdZ said:
I've gotten even better results with the SV kernel. After testing each speed for stability at lower and lower voltages here are my results:
1408Mhz @ 1175mV
1216Mhz @ 1075mV
1100Mhz @ 975mV
1015Mhz @ 900mV
816Mhz @ 800mV
655Mhz @ 725mV
503Mhz @ 650mV
389Mhz @ 595mV
lose 100Mhz from the UV/OC kernel but I get better UV settings with the SV kernel
Click to expand...
Click to collapse
too bad the lowest my phone can go for 1015mhz and 816mhz are 940 and 840mv respectively
Black6spdZ said:
I've gotten even better results with the SV kernel. After testing each speed for stability at lower and lower voltages here are my results:
1408Mhz @ 1175mV
1216Mhz @ 1075mV
1100Mhz @ 975mV
1015Mhz @ 900mV
816Mhz @ 800mV
655Mhz @ 725mV
503Mhz @ 650mV
389Mhz @ 595mV
lose 100Mhz from the UV/OC kernel but I get better UV settings with the SV kernel
Click to expand...
Click to collapse
We seem to be hitting about the same spot, almost exactly, I can OC to 1400 or 1500 but choose not to push my phone to that extent. I live in a very hot area and don't want to tempt fate.
I tried to search but was unsuccessful to find the information that I wanted.
Can someone tell me what are the recommended safe undervolt settings for the transformer?
I'm running prime 1.9 with clemsyn-blades 1.6a.
Thx
Sent from my Transformer TF101 using Tapatalk
steve.garon said:
I tried to search but was unsuccessful to find the information that I wanted.
Can someone tell me what are the recommended safe undervolt settings for the transformer?
I'm running prime 1.9 with clemsyn-blades 1.6a.
Thx
Sent from my Transformer TF101 using Tapatalk
Click to expand...
Click to collapse
I don't know what the stock settings are but I can tell you the process I followed when I OCed my tablet. Install the custom kernel and rom you want. Then install an app to modify frequency and voltages. I use setCPU but pimpmycpu works also.
1. Find the highest clock you can attain with the provided overvolt from the kernel. To do this just start at the highest frequency and see if the tablet reboots or locks during a stress test. Some people play games to stress. I use the built in stress tester in SetCPU. Make sure to apply the higher frequency for testing but not to apply settings on reboot or you could get stuck in a reboot loop (or so I am assuming the kernel might reset to 1 GHz when a hard lock occurs).
2. Once you have the maximum Frequency (for me it was 1.644 MHz) then start lowering the voltage for that frequency only until you loose stability. I can undervolt 1.644 by -75 mV. If I go upto -100 then I get random reboots. I go in -25 mV increments but you could narrow it down even more by going with -10 or even -5 intervals.
3. Take the offset you got (in my case -75 mV)on the highest frequency and apply it to every frequency above 1 GHz and see if this is stable. If not then slowely increase voltage until you gain stability again.
4. (Optional) Then start undervolting stock frequencies from 216 MHz to 1 GHz. Every tablet is different but (I have heard) this could cause sleep issues. I undervolted all frequencies from above 216 to 1000 GHz by -25 Mv without any issues. I may bring it down slowely by increments of-5 or -10 until I run into some issues.
5. (Optional) Undervolt 216 MHz as low as possible to save standby battery power. I undervolted my 216 MHz to 725 mV (even though SetCPu goes lower the kernel is locked at this). Which helps keep my battery levels up when the tablet screen is off and not in use.
Hints: To test that a voltage is okay at a certain frequency then make sure to set your max to the frequency level you want to test. If you set the max at 1544 and undervolt 1 GHz you will never know if it is okay because when you stress it it will blow right past 1 GHz upto 1544 and the voltage for that will take over.
If you have any questions please let me know.
Bjd223 said:
I don't know what the stock settings are but I can tell you the process I followed when I OCed my tablet. Install the custom kernel and rom you want. Then install an app to modify frequency and voltages. I use setCPU but pimpmycpu works also.
1. Find the highest clock you can attain with the provided overvolt from the kernel. To do this just start at the highest frequency and see if the tablet reboots or locks during a stress test. Some people play games to stress. I use the built in stress tester in SetCPU. Make sure to apply the higher frequency for testing but not to apply settings on reboot or you could get stuck in a reboot loop (or so I am assuming the kernel might reset to 1 GHz when a hard lock occurs).
2. Once you have the maximum Frequency (for me it was 1.644 MHz) then start lowering the voltage for that frequency only until you loose stability. I can undervolt 1.644 by -75 mV. If I go upto -100 then I get random reboots. I go in -25 mV increments but you could narrow it down even more by going with -10 or even -5 intervals.
3. Take the offset you got (in my case -75 mV)on the highest frequency and apply it to every frequency above 1 GHz and see if this is stable. If not then slowely increase voltage until you gain stability again.
4. (Optional) Then start undervolting stock frequencies from 216 MHz to 1 GHz. Every tablet is different but (I have heard) this could cause sleep issues. I undervolted all frequencies from above 216 to 1000 GHz by -25 Mv without any issues. I may bring it down slowely by increments of-5 or -10 until I run into some issues.
5. (Optional) Undervolt 216 MHz as low as possible to save standby battery power. I undervolted my 216 MHz to 725 mV (even though SetCPu goes lower the kernel is locked at this). Which helps keep my battery levels up when the tablet screen is off and not in use.
Hints: To test that a voltage is okay at a certain frequency then make sure to set your max to the frequency level you want to test. If you set the max at 1544 and undervolt 1 GHz you will never know if it is okay because when you stress it it will blow right past 1 GHz upto 1544 and the voltage for that will take over.
If you have any questions please let me know.
Click to expand...
Click to collapse
Cool. I'll give this a try.
Sent from my Nexus S using Tapatalk
I get a stable undervolt at -50 on all cpu steps which is pretty good. This should save loads of battery.
Sent from my Nexus S using Tapatalk
steve.garon said:
I get a stable undervolt at -50 on all cpu steps which is pretty good. This should save loads of battery.
Sent from my Nexus S using Tapatalk
Click to expand...
Click to collapse
I started at the top and went down every stepping from 1624 to 1000 and found out the lowest stable frequency and then added 10 to be safe to each one. I am currently at
1624 (-70)
1592 (-80)
1544 (-180)
1472 (-80)
1424 (-160)
1336 (-185)
1232 (-185)
1000 (- 110) which is 990 mV
I then set every stepping under 1 GHz frequency that was above 990 mV to 990 mV as if it can run at 1 GHz at that voltage then any speed under that will run fine at that voltage also. I have some more stability testing to do but I think I got it pretty much nailed down for my device. I may go through all the sub 1 GHz frequencies but it does take some time to run a stress test under each stepping.
Also I have never had any sleep issues but that is something to look out for. I also wonder how temperature affects stability as I have never gotten it above 40c. The pad might benefit from a 1 or 2mm thermal pad squeezed between the CPU and the back bezel (if it will fit).
Does anyone know that tjunction max for a Tegra2? I have a profile set up in SetCPU that sets the CPU frequency at 1 GHz if the CPU gets above 42 because I saw something similar in the Clemsyn rom but don't really know when the CPU will start to throttle itself at or shut off.
I am trying to find stock voltages for different devices because it seems like some mfgs are undervolting to save battery on some tablets.
Bjd223 said:
I started at the top and went down every stepping from 1624 to 1000 and found out the lowest stable frequency and then added 10 to be safe to each one. I am currently at
1624 (-70)
1592 (-80)
1544 (-180)
1472 (-80)
1424 (-160)
1336 (-185)
1232 (-185)
1000 (- 110) which is 990 mV
I then set every stepping under 1 GHz frequency that was above 990 mV to 990 mV as if it can run at 1 GHz at that voltage then any speed under that will run fine at that voltage also. I have some more stability testing to do but I think I got it pretty much nailed down for my device. I may go through all the sub 1 GHz frequencies but it does take some time to run a stress test under each stepping.
Also I have never had any sleep issues but that is something to look out for. I also wonder how temperature affects stability as I have never gotten it above 40c. The pad might benefit from a 1 or 2mm thermal pad squeezed between the CPU and the back bezel (if it will fit).
Does anyone know that tjunction max for a Tegra2? I have a profile set up in SetCPU that sets the CPU frequency at 1 GHz if the CPU gets above 42 because I saw something similar in the Clemsyn rom but don't really know when the CPU will start to throttle itself at or shut off.
I am trying to find stock voltages for different devices because it seems like some mfgs are undervolting to save battery on some tablets.
Click to expand...
Click to collapse
-180 in some frequencies!! Thats just insane. I've got to try this.
FYI, you don't have to lower voltage for all upper 1ghz frequencies. If your running at 1544 for exemple. The cpu will never touch the frequencies between 1000 and 1544. After the 1000 its a direct step to whatever frequency your overclocking too. You can monitor this in setcpu.
I cannot undervolt as precise as you can. For some reason setcpu only let me do 25mv increment...
Sent from my Nexus S using Tapatalk
I have undervolted by -175 on all frequencies, obviously 216Mhz is locked but all seems pretty stable after some testing. No sleep death either so I'm going to carry on testing.
steve.garon said:
-180 in some frequencies!! Thats just insane. I've got to try this.
FYI, you don't have to lower voltage for all upper 1ghz frequencies. If your running at 1544 for exemple. The cpu will never touch the frequencies between 1000 and 1544. After the 1000 its a direct step to whatever frequency your overclocking too. You can monitor this in setcpu.
I cannot undervolt as precise as you can. For some reason setcpu only let me do 25mv increment...
Sent from my Nexus S using Tapatalk
Click to expand...
Click to collapse
In SetCPU do not use the slider use the keyboard and type in the exact number you want.
I wanted to figure out every stepping in case I wanted to switch to a lower frequency but yeah most poeple don't have to do this.
The lowest stable I have gotten is -190 but feel like adding 10 padding will ensure stability.
I will probably do the sub 1 GHz frequencies today or tomorrow.
Bjd223 said:
In SetCPU do not use the slider use the keyboard and type in the exact number you want.
I wanted to figure out every stepping in case I wanted to switch to a lower frequency but yeah most poeple don't have to do this.
The lowest stable I have gotten is -190 but feel like adding 10 padding will ensure stability.
I will probably do the sub 1 GHz frequencies today or tomorrow.
Click to expand...
Click to collapse
I didnt even noticed I could use the keyboard...
I finally settled to -25 @216, -50 under 612 and -100 over 612. And everything is stable
Sent from my Nexus S using Tapatalk
I have tested each voltage. I am under the impression that 725 mV is the kernel minimum in the Clemsyn kernel so once I hit that I just set every frequency under that too 725. I will eventually go through and stress test every frequency @ 725 to make sure everything is a-okay but I am sure it is.
Here is my list incase anyone wants a place to start.
1680 MHz (-000) I can not overclock to this freuqency.
1624 MHz (-070) 1380 mV
1592 MHz (-080) 1345 mV
1544 MHz (-180) 1220 mV
1472 MHz (-080) 1220 mV
1424 MHz (-160) 1115 mV
1336 MHz (-185) 1065 mV
1232 MHz (-185) 990 mV
1000 MHz (-215) 885 mV
0912 MHz (-215) 835 mV
0816 MHz (-235) 790 mV
0760 MHz (-215) 760 mV
0608 MHz (-225) 725 mV
0456 MHz (-100) 725 mV
0312 MHz (-075) 725 mV
0216 MHz (-025) 725 mV
Clemsyn said the minimum mV in his kernel but I can't seem to find the thread, but I think it was 725. If I find out it is something lower I will go through the lower ones until I hit the minimum.
Also please note that everything above 1 GHz has +10 mV padding added to it above the lowest stable I could find. Everything below 1 GHz only has a +5 mV padding on it. So for example I actually hit a -240 w/o any issues @ 816 MHz.
I tested in +/-5 mV increments but if you wanted you could narrow it down even further.
EDIT: Also after some testing it looks like it does use frequencies between 1 GHz and your max. I have it set to 1644 and it does hit 1232 MHz occasionally.
Bjd223 said:
I don't know what the stock settings are but I can tell you the process I followed when I OCed my tablet. Install the custom kernel and rom you want. Then install an app to modify frequency and voltages. I use setCPU but pimpmycpu works also.
1. Find the highest clock you can attain with the provided overvolt from the kernel. To do this just start at the highest frequency and see if the tablet reboots or locks during a stress test. Some people play games to stress. I use the built in stress tester in SetCPU. Make sure to apply the higher frequency for testing but not to apply settings on reboot or you could get stuck in a reboot loop (or so I am assuming the kernel might reset to 1 GHz when a hard lock occurs).
2. Once you have the maximum Frequency (for me it was 1.644 MHz) then start lowering the voltage for that frequency only until you loose stability. I can undervolt 1.644 by -75 mV. If I go upto -100 then I get random reboots. I go in -25 mV increments but you could narrow it down even more by going with -10 or even -5 intervals.
3. Take the offset you got (in my case -75 mV)on the highest frequency and apply it to every frequency above 1 GHz and see if this is stable. If not then slowely increase voltage until you gain stability again.
4. (Optional) Then start undervolting stock frequencies from 216 MHz to 1 GHz. Every tablet is different but (I have heard) this could cause sleep issues. I undervolted all frequencies from above 216 to 1000 GHz by -25 Mv without any issues. I may bring it down slowely by increments of-5 or -10 until I run into some issues.
5. (Optional) Undervolt 216 MHz as low as possible to save standby battery power. I undervolted my 216 MHz to 725 mV (even though SetCPu goes lower the kernel is locked at this). Which helps keep my battery levels up when the tablet screen is off and not in use.
Hints: To test that a voltage is okay at a certain frequency then make sure to set your max to the frequency level you want to test. If you set the max at 1544 and undervolt 1 GHz you will never know if it is okay because when you stress it it will blow right past 1 GHz upto 1544 and the voltage for that will take over.
If you have any questions please let me know.
Click to expand...
Click to collapse
Thanks for this mate. I followed your guide exactly and I'm using clems 16? Kernel completely stable at all frequencies including the undervolting!
Based on my testing you need ~ a +15 to +25 mV over stable CPU voltage to get stable 3D GPU.
I found that I could run hours of "Stress CPU" without any issues at a voltage but during stressful 3D games that are using CPU + GPU cores, at that same frequency, the machine would FC the app, reboot or lock. I have also noticed that the closer you are to the stable voltage different things will happen.
Far off - machine will usually reboot
Closer - machine will usually hard lock
Even closer - Apps will FC, widgets will act strange (like invert colors, black blocks for backgrounds, etc.)
So what I do now is I test with Dungeon Defenders at the main menu. Its seems that if you can last 1 hr at the Dungeon Defenders menu you are GTG. If you actually play the game then 1 stage is what I use to test.
Bjd223 said:
I don't know what the stock settings are but I can tell you the process I followed when I OCed my tablet. Install the custom kernel and rom you want. Then install an app to modify frequency and voltages. I use setCPU but pimpmycpu works also.
1. Find the highest clock you can attain with the provided overvolt from the kernel. To do this just start at the highest frequency and see if the tablet reboots or locks during a stress test. Some people play games to stress. I use the built in stress tester in SetCPU. Make sure to apply the higher frequency for testing but not to apply settings on reboot or you could get stuck in a reboot loop (or so I am assuming the kernel might reset to 1 GHz when a hard lock occurs).
2. Once you have the maximum Frequency (for me it was 1.644 MHz) then start lowering the voltage for that frequency only until you loose stability. I can undervolt 1.644 by -75 mV. If I go upto -100 then I get random reboots. I go in -25 mV increments but you could narrow it down even more by going with -10 or even -5 intervals.
3. Take the offset you got (in my case -75 mV)on the highest frequency and apply it to every frequency above 1 GHz and see if this is stable. If not then slowely increase voltage until you gain stability again.
4. (Optional) Then start undervolting stock frequencies from 216 MHz to 1 GHz. Every tablet is different but (I have heard) this could cause sleep issues. I undervolted all frequencies from above 216 to 1000 GHz by -25 Mv without any issues. I may bring it down slowely by increments of-5 or -10 until I run into some issues.
5. (Optional) Undervolt 216 MHz as low as possible to save standby battery power. I undervolted my 216 MHz to 725 mV (even though SetCPu goes lower the kernel is locked at this). Which helps keep my battery levels up when the tablet screen is off and not in use.
Hints: To test that a voltage is okay at a certain frequency then make sure to set your max to the frequency level you want to test. If you set the max at 1544 and undervolt 1 GHz you will never know if it is okay because when you stress it it will blow right past 1 GHz upto 1544 and the voltage for that will take over.
If you have any questions please let me know.
Click to expand...
Click to collapse
I always wanted to learn more about the mV and CPU speed stuff. I always install rom, install kernel of the one I heard the better is and that's it. Never had touched the voltage measure in any of my android device. Yeah i know, so noob.
Thanks for this how to info, I think I will play around a bit later. Right now I'm struggling with my self on wish of the 3 roms available install and stay.
Had to remove undervolt this weekend since my camera wouldnt start anymore. I'll have to redo all the tests I was doing but test for camera too. The tablet would just freeze as soon as I opened the camera app
Sent from my Nexus S using Tapatalk
steve.garon said:
Had to remove undervolt this weekend since my camera wouldnt start anymore. I'll have to redo all the tests I was doing but test for camera too. The tablet would just freeze as soon as I opened the camera app
Sent from my Nexus S using Tapatalk
Click to expand...
Click to collapse
My camera does not work either. I just figured it was Clemsyns kernel not the OC. I don't know anything about cameras but maybe they use a lot of voltage so the voltage to the CPU dips.
I never used the camera personally, but I will test it some and see what additional mV you need over stable to get the camera working.
Also I noticed the more you undervolt (it seems to me) the harder it is to undock with an app open. usually instead of reloading they FC or lock the device so i just make sure to hit the home key before I undock then alt tab right back into the app.
Bjd223 said:
My camera does not work either. I just figured it was Clemsyns kernel not the OC. I don't know anything about cameras but maybe they use a lot of voltage so the voltage to the CPU dips.
I never used the camera personally, but I will test it some and see what additional mV you need over stable to get the camera working.
Also I noticed the more you undervolt (it seems to me) the harder it is to undock with an app open. usually instead of reloading they FC or lock the device so i just make sure to hit the home key before I undock then alt tab right back into the app.
Click to expand...
Click to collapse
My camera did not work at 912 MHz at stock voltage on 17b but now on 18 I can do my normal undervolt settings and the camera works fine.
Hi,
1) I have installed Kyorarom 0.5.2 OC kernel and here are my SetCPU profiles.
Main - Min: 300 Mhz, Max: 800Mhz
On Call - Min: 83, Max: 400, Priority: 100
Time - 11 PM to 6 AM - Min: 83, Max: 200, Priority: 99
Screen Off: Min: 83, Max: 400, Priority: 98.
My problem is, if I use my phone after 11PM, the time profile gets activated.
But when screen is on, I want to use the main profile. How can I do that?
2) Which governor has less stress on CPU? Conservative or Lagfree?
3) When in the main profile, my phone constantly runs at my max setting, (ie) 800 Mhz. It never runs lower than that. Why does it happen?
maverickgenius said:
Hi,
1) I have installed Kyorarom 0.5.2 OC kernel and here are my SetCPU profiles.
Main - Min: 300 Mhz, Max: 800Mhz
On Call - Min: 83, Max: 400, Priority: 100
Time - 11 PM to 6 AM - Min: 83, Max: 200, Priority: 99
Screen Off: Min: 83, Max: 400, Priority: 98.
My problem is, if I use my phone after 11PM, the time profile gets activated.
But when screen is on, I want to use the main profile. How can I do that?
2) Which governor has less stress on CPU? Conservative or Lagfree?
3) When in the main profile, my phone constantly runs at my max setting, (ie) 800 Mhz. It never runs lower than that. Why does it happen?
Click to expand...
Click to collapse
HI,
1) I dont think that's even possible man... would be nice though
2) conservative
3) how do you know that its for constant? and that's normal if you've chosen 'performance' gov'ner.
PersianSphinx said:
HI,
1) I dont think that's even possible man... would be nice though
2) conservative
3) how do you know that its for constant? and that's normal if you've chosen 'performance' gov'ner.
Click to expand...
Click to collapse
1) Okay. I thought that the main profile is something like a "screen on" profile, which would override all others.
3) I checked using CPU Spy. My phone is always at the highest frequency. I am using conservative.
try a different governor / kernel
maverickgenius said:
1) Okay. I thought that the main profile is something like a "screen on" profile, which would override all others.
3) I checked using CPU Spy. My phone is always at the highest frequency. I am using conservative.
Click to expand...
Click to collapse
I'm not sure about the constant 800. even if it is, as long as battery works well, everything is OK
PersianSphinx said:
I'm not sure about the constant 800. even if it is, as long as battery works well, everything is OK
Click to expand...
Click to collapse
Well, even if the battery works fine, running the processor constantly at 800 MHz will decrease its life.
Does your phone too behave the same way as mine?
maverickgenius said:
Well, even if the battery works fine, running the processor constantly at 800 MHz will decrease its life.
Does your phone too behave the same way as mine?
Click to expand...
Click to collapse
lol, how long are you planning to keep this phone? decreasing the life from 10 years to 8 wont bother me at all (I dont actually know the life)
PersianSphinx said:
lol, how long are you planning to keep this phone? decreasing the life from 10 years to 8 wont bother me at all (I dont actually know the life)
Click to expand...
Click to collapse
Agree with you on that point. But you still didn't answer my question
1) Not possible.
2) Conservative.
3) Problem with conservative governor in Kyoyakernel maybe.
Advice : - Always use On Demand.
Find Kryllios ROM instructioin, he have sugeested good settings to get maximum battery life...........
Max 300 - MHz and Min 1200 Mhz...what is this!
Hi, I have a Samsung Galaxy S II I9100G and rooted it with Kernel superatoms v3.0 XXLA1 and installed SetCPU 2.3.0.2 in it.
Now, when I set the max and min slider to 1200 MHz (with Governer hotplug) and goes back to the home screen, after that, when I again open SetCPU, the max slider gets to 300 MHz and the min to 1200 MHz (with Governer powersave).
What is this, man!!! How can the processor run with maximum speed of 300 MHz and the minimum of 1200 MHz ....
Plzz...I need help