Database Develop

[PRJ] The Manhattan Project - 1.6GHz - Shelved/Succeeded by Core

http://board.teamwhiskey.com/viewtopic.php?f=3&t=53&start=0
Project is simple in design and is a work in progress, overclock as far as we can go with ample supplied voltage for maximum stability. These are the screenshots taken last night from the last few builds of the evening, were up to something like 40 builds so far trying to get everything just the way we want it to run, look, and feel.
Currently we are overclocking both the GPU and CPU using both base frequency modifiers and multipliers. This is allowing us to directly push the GPU to its stable limit while adjusting out the CPU accordingly.
Project Developers
Morfic and Viralblack of Team Whiskey. This is and will continue to be Morfic's baby until the phone prints money and/or can transform into a jet.
The Short and Long term goals are the following:
-To optimize the kernel with the best toolchains available to us
-Inject vital code enhancements for stability and to extend the feature set of the SGS Vibrant line
-To determine Min/Max Voltage, Bus Clock, and Multiplier values for Maximum Overclock and Maximum Effecient Overclock
-Continue to push the GPU to its theoretical maximum independant of CPU clock values.
Target Numbers and Current Results.
CPU Target 2.0GHz, Current Maximum Overclock 1.6 Stable (Time limitation only, havent rebuilt to try more)
GPU Target 85 FPS Stable Bench, Current 78 FPS
At current this is a JFD Stable build thats more than fairly bulletproof on battery life. To use with JI6 builds just flash Eugene's JFD/JI6 support package.
{
"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"
}
This is NOT a hoax or scam, these are real DDMS screenshots from the build currently in progress, as it has not yet been released we do not have code or kernels available until we know its running smoothly. The following zImage is a prerelease alpha. Its strictly for proof of concept. This kernel is mildly optimized and includes the CPU and GPU overclocks. In three of three test phones the CPU was stable, in two of three phones the GPU was stable. One test phone was not able to complete neocore under this coding.
Battery life has proven about the same as traditional OC/UV kernels. This is NOT our finished product, this is an Alpha for proof of concept for other developers. Run at your own risk. There is no failsafe, this does not require SetCPU to engage. It comes out of the box at 1.6GHz with steppings of 1,2,4,8,16
Currently, tested incorporations with the zMod have shown that overclocking renders the zMod unstable and bricks the device. I personally preferred to use RyanZAs OCLF v2.2 APK.
I had one out of 20 quadrant runs fail because I switched scheduler away from deadline. CFQ proved to be less stable for one reason or another when running quadrant.
This thread will be updated as work progresses, Do not flash this if you do not know what youre doing. Suggestions are welcome for integrated features you would like to see in this kernel, we will not however be fielding bug reports as this right now is staged as a Prerelease Alpha or RC0.
If when this is flashed your phone crashes and is unable to work, my suggestion is odin the stock kernel back in place and wait for a lighter overclock version to be released.
Download -
Removed while core is being rewritten
EDIT 10/28:
To update the project and to iterate why the download link was removed, the kernel seems to be misbehaving and we havent yet nailed down the correction as of yet.
As many people pointed out once it was released, the kernel seems to slow down 3D graphics. With this information and taking a day to go through all the options and data available to me, we discovered this issue isnt limited to just 3D graphics.
Imagine if you were running at 10MPH, and while you were running time slowed down but the slowdown didnt effect you. So youre still running at 10MPH but to the rest of the slowed down world, now youre running at 15MPH, even though your speed and direction havent changed.
The changes in the code for one reason or another arent playing nice with the Android system. Instead of pushing the clock up, in most cases, what its doing is simply slowing the internal clock down to the tune of about 30%. This is what is actually providing the benchmark gains that were being recorded, not an actual gain. This doesnt mean that the project is dead, we are recompiling to get a stable overclock with the new configurations. This doesnt mean the project was a failure, it just means we need to rewrite the alpha.
Edit 10/30 -
So, we are having distinct issues removing the timing error from the kernel while maintaining the integrity of the time system. So, being that the timing is much more important, we began rewriting it all from scratch. This is currently marked as Project Core in the same forums. Core is taking the best of the overclocking and fine tuning that this kernel has and combining it with energy effeciency and stability a daily driver phone needs.
So, we will be balencing the high end power thats desired and the low power use that is needed in this new project. With that being said, I dont believe that 1.6GHz (in my own humble opinion) is even remotely necessary for 99 percent of phone, let alone for daily use. This project is being succeeded by Core to make a power friendly daily driver kernel.
The kernel will still be available from the forums Alpha/Beta Download Section, but will not be supported as its being replaced by Core.
We found the redline Now we need to find the happy medium. If you enjoyed this release, youll love Core and Core with zMod.
Thread Locked until release, OP PM me when you want to release it

Good work TW
You guys should start working on a 2.2 version (Once you get this one smooth)

Thanks, but....
The processor already gets too hot at 1ghz.
I don't feel like cooking it just to say "hey wow, look how high I pushed it before it died".
Y'all really need to let people know that overclocking creates more heat and in the confined area of a tiny little cellphone there is next to zero cooling.
It's just asking for trouble.
Not that I don't appreciate the effort you guys are putting into the tweaks tho....

n2ishun said:
Thanks, but....
The processor already gets too hot at 1ghz.
I don't feel like cooking it just to say "hey wow, look how high I pushed it before it died".
Y'all really need to let people know that overclocking creates more heat and in the confined area of a tiny little cellphone there is next to zero cooling.
It's just asking for trouble.
Not that I don't appreciate the effort you guys are putting into the tweaks tho....
Click to expand...
Click to collapse
My phone never gets hot?
But they did put a warning, and you are the one modding your device.

Oh wow! This is great! I have an Epic 4g, will I be able to benefit from this awsome project?

Eazail70x7 said:
Oh wow! This is great! I have an Epic 4g, will I be able to benefit from this awsome project?
Click to expand...
Click to collapse
Hopefully yes, the kernel will be ported to all SGS Variants if possible, Right now we have in our possession Vibrants and Captivates, were looking to expand and will probably be working with other developers to get this kernel ported across.

Master™ said:
My phone never gets hot?
But they did put a warning, and you are the one modding your device.
Click to expand...
Click to collapse
Yeah, mine will get warm, but I've never seen it get real hot.

testing now . just flash with clockwork? this is super cool guys. props

blackerwater said:
testing now . just flash with clockwork?
Click to expand...
Click to collapse
Yes, its CWM flashable, but be careful with it. There is no failsafe, if your phone cant handle the overclock itll just crash as its set automatically to 1.6GHz topend

So far so good. Just got a linpack of 12.062. Outstanding job guys!

doesn't work yet. just sits on vibrant screen. its cool ill just flash back ji6 zimage then it should boot. im on stock ji6..... I have a vibrant that can handle 1.3mhz for sure and voodo so ill keep playing with it.

Will this run with stock ji6?

Thank you so much for uploading! I reread about it in phandroid and was about to ask around for it
Sent from my SGH-T959 using XDA App

I flashed again just to make sure . it wont boot on ji6 stock but my ji6 kernel works just fine. so maybe if i was running a bionix rom?

blackerwater said:
I flashed again just to make sure . it wont boot on ji6 stock but my ji6 kernel works just fine. so maybe if i was running a bionix rom?
Click to expand...
Click to collapse
Right now ive only tested it on Fusion personally

Viralblack said:
Hopefully yes, the kernel will be ported to all SGS Variants if possible, Right now we have in our possession Vibrants and Captivates, were looking to expand and will probably be working with other developers to get this kernel ported across.
Click to expand...
Click to collapse
Excellent. I hope you guys can get an Epic soon and work some magic. Also good to hear thatyou'll be working with other devs. Good luck

blackerwater said:
I flashed again just to make sure . it wont boot on ji6 stock but my ji6 kernel works just fine. so maybe if i was running a bionix rom?
Click to expand...
Click to collapse
I have stock JI6 rom but have the "backwards compatibility for JFD kernels" flashed so maybe that's it.

Have you tried flashing the back kernel support file for JI6 then flash this kernel?

Whitehawkx said:
I have stock JI6 rom but have the "backwards compatibility for JFD kernels" flashed so maybe that's it.
Click to expand...
Click to collapse
ok ill give that a shot. thanks

Will this work with voodoo?
Bionix Fusion 1.0 Jacs Xmod Oc/Uv Kernel Color fix Voodoo

[HACK] Overclocking/Undervolting kernels with Voltage Control support

Hello,
I'm the author of Voltage Control, an application for Galaxy S series to control OC/UV kernels. I don't personally own a Nexus S (and I cant say I will, unless someone want to Nexus S donate one for me ;p).
Voltage Control Lite 4.00 on Android Market!
{
"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 know there are many apps to control OC, but I dont know any doing UV work.
If anybody is interested in making UV kernel for Nexus S there are some working kernels for SGS (given the hardware similarities some knowledge may be taken from current solutions).
Authors of most OC/UV code are raspdeep and Unhelpful, heres the topic: http://forum.xda-developers.com/showthread.php?t=822027 and second: http://forum.xda-developers.com/showthread.php?t=762171
If you are interested in creating UV kernel working with Voltage Control, I'll give brief explanation how VC works, and how are things done on SGS.
My application communicates with kernel via sysfs interfaces via cat and echo:
mandatory:
Code:
/sys/devices/system/cpu/cpu0/cpufreq/stats/time_in_state // to read all states
/sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq // may be obsolete, didnt make a decision yet
Code:
/sys/devices/system/cpu/cpu0/cpufreq/UV_mV_table
key in UV procedure, it represents undervolt for each frequency from highest to lowest; "25 25 50 50 100" means UV highest freq by 25, and lowest by 100
providing additional functionality:
Code:
/sys/devices/system/cpu/cpu0/cpufreq/frequency_voltage_table
Containing all frequencies and corresponding voltages, format:
FREQ STOCK CURRENT
Code:
1120000 1300 1300
1000000 1300 1300
This is useful for determining stock voltages used by the device, because they are needed in voltage calculation. Stock values are hardcoded, but If you want to change them I suggest to implement this table, to let user know at exactly what voltage his device is running.
Third column is obsolete, since VC calculates current (stock-uv) voltage itself might be removed in the future.
Code:
/sys/devices/system/cpu/cpu0/cpufreq/states_enabled_table
This is table where we store enabled ("1") and disabled ("0") state for each frequency;
"0 1 1 0 1" means we disabled highest frequency and second to lowest.
This is useful when kernel developer wants to produce ONE kernel for all OC lovers.
This means there could be one kernel supporting OC frequencies up to 1600Mhz (lets say 1600,1500,1400,1300,1200,1120), and user could choose which ones he wants to use (eg 1400 and 1120, disabling the rest).
I'm aware that Nexus S will receive tremendous attention from developers so I believe that someone might actually come with better implementation (single table consisting of all information needed by VC?)of this mess.
Best idea would be to make one table for all information, so it can be easily parsed.
Current version of app is 4.0 beta, which is a complete rewrite.
I'm open for ideas, suggestions regarding additional features.
Happy OCing.

Thanks xan for this invitation
Unhelpful also played a big part with the initial code to load dynamically frequencies and voltages using a sysfs interface.
BTW, your app source is welcome on project-voodoo github !
I know several people motivated to contribute, it would be a place to start.

Right, Unhelpful, adding topic to OP.
Yes, this needs to be open-sourced.
No date yet, because there are monsters inside.
The thing that struck me now, is why is all of that so complicated?
I think I like single-table idea more and more.

ill volunteer to help.

Very exciting to see some of the best dev names from the SGS community showing interest in the Nexus S. Many goog things to come for us if they remain interested...

So when are we gonna see a 1.2GHz Voodoo OC/UV kernel

My phone recieves a noticeable speed increase with the voodoo kernel for Nexus S with max and min set to 1000 with performance governor Can't wait for a super optimized kernel OC/UV!

looking forward to a UV kernel to extend battery life.

jroid said:
My phone recieves a noticeable speed increase with the voodoo kernel for Nexus S with max and min set to 1000 with performance governor Can't wait for a super optimized kernel OC/UV!
Click to expand...
Click to collapse
Anything on battery?
defysword said:
looking forward to a UV kernel to extend battery life.
Click to expand...
Click to collapse
+1

why is no one compiling a kernel supporting xans app? i really loved it on my galaxy s and when i saw this threat i was hoping for a supported kernel for my nexus s. it really can make oc more stable if u r patient enough to test the best setzings for ur phone
Sent from my Nexus S using Tapatalk

Can't wait for this too. Xan's app is really good on SGS.

shrooms90 said:
why is no one compiling a kernel supporting xans app? i really loved it on my galaxy s and when i saw this threat i was hoping for a supported kernel for my nexus s. it really can make oc more stable if u r patient enough to test the best setzings for ur phone
Sent from my Nexus S using Tapatalk
Click to expand...
Click to collapse
This died the 22nd of December, I got my Nexus S around that time, kept it stock while on vacation over the holidays, missed out there.
I *may* check this out, don't want to turn into Patches....
But I could continue to build Honity stable for Will, but others run lower than he does.
Instead of stock + UV build it would be only 1.
And Honity14 could be the UVOC some want.
Problem in that is that I take a risk of people not understanding the "use at your own risk" part involved.
Thanks for bumping this thread, didn't know xan had this.
Sent from my Nexus S

morfic said:
This died the 22nd of December, I got my Nexus S around that time, kept it stock while on vacation over the holidays, missed out there.
I *may* check this out, don't want to turn into Patches....
But I could continue to build Honity stable for Will, but others run lower than he does.
Instead of stock + UV build it would be only 1.
And Honity14 could be the UVOC some want.
Problem in that is that I take a risk of people not understanding the "use at your own risk" part involved.
Thanks for bumping this thread, didn't know xan had this.
Sent from my Nexus S
Click to expand...
Click to collapse
sounds good. i like honity kernel but for my phone it needs a little fine tuning.
i think people would understand "use at ur own risk" imo most people understood it on the sgs forum.
so i really would love to see honity14 with uv options
Sent from my Nexus S using Tapatalk

Is there any progress?

xan said:
I'm open for ideas, suggestions regarding additional features.
Happy OCing.
Click to expand...
Click to collapse
Hi. great stuff here. would this app work (or made to work) using snq-'s undervolting via sysfs here.
I use a kernel with the above patch enabled and my ROM users would love to be able to do this using an app like this.
Thanks!

pongster said:
Hi. great stuff here. would this app work (or made to work) using snq-'s undervolting via sysfs here.
I use a kernel with the above patch enabled and my ROM users would love to be able to do this using an app like this.
Thanks!
Click to expand...
Click to collapse
Im a noob when it comes to kernel developping, but i hope this will help u:
From what i understood this app undervolts using following command:
echo "0,0,50,100,0...." /sys/devices/system/cpu/cpu0/cpufreq/UV_mV_table
Where the numbers stand for the amount of undervolting in mV for each freq. (I think it is from lowest to highest but im not sure. If u set the values at boot it just writes a startup script in /system/etc/init.d/
Sent from my Nexus S using Tapatalk

Fantastic,
The nexus s dev community is definitely growing now
JD
Sent from my Google Nexus Magic Wand S

JupiterDroid said:
Fantastic,
The nexus s dev community is definitely growing now
JD
Sent from my Google Nexus Magic Wand S
Click to expand...
Click to collapse
wtf? fancy seeing you here

Stea1thmode said:
wtf? fancy seeing you here
Click to expand...
Click to collapse
WTH young man, did you get lost

Just thought that I'd mention this.
I develop a kernel for the Captivate for CM7. Our cm kernel is based off Nexus S code, so anything that works under it will work under your kernels as well (probably). I've implemented this successfully under my kernel with no real problems.
Just add a sysfs interface to the cpufreq driver to store the uv amount, then subtract that amount * 1000 from arm_volt before feeding it into the regulator.
Sources are here --> https://github.com/TheEscapistxda/CM7-OCUV
PS I apologize for my terrible commenting.

[DER KERNEL] NS+4G Trinity (E)UV and OC ICS + JB roms

It's DoW if you owned a Vibrant before, Trinity is the name in its passport though.
-------------------------------------------------------------------------------------------
Files and source:
www . Der Kernel . com
-------------------------------------------------------------------------------------------
​
ICS:
Voodoo Sound V10 (Thanks supercurio)
Trinity Spice (based on voodoo color V1 by supercurio)
BLN (Backlight notification) (Big Thanks Neldar)
Backlight dimmer by Ezekeel
Based on Samsung git source.
Safeboot to lower clock, higher need a tool like setcpu or no-frills CPU control
New Trinity bootanimation by @rascarlo
TUN module (Tunneling/VPN support)
CIFS module (Samba support)
EXT4 tweaks
Most lean build achievable.
Not a rebooter.
JB: WIP
For CM, to use this kernel flash the CM Black Screen Fix (Thanks JackpotClavin and lurtis)
Thanks to @RyanMacG for euroskank hosting​
When "Just Flash It" didn't work:
Flash rom.
Flash kernel.
Power off.
Pull battery for 2 minutes.
Put battery back
Now boot phone.
---------------------------------------------------------------------------------------
GPS woes? Try this .apk, worked wonders on GB.
Endursa shared this GL fixing link. Follow it, see 65fps on Quadrant planet like it was meant to be.
T for Trinity
UV for undervolt (you would want to uninstall setcpu like apps and just run it)
EUV represents a 880MHz top speed power savings kernel, snappy still? You Bet'cha!
144 for 1.44GHz (I will rethink this when the 10GHz phones hit the market)
-Any means it uses Koush's anykernel scripting to install trinity using the original ramdisk, making it possible to use it on m(any) roms w/o needing a special version.
(X)A new kernel just born by my slacker hands, feel free to use it, we think it's good even if it was eXperimental just moments ago. (X) comes off when we feel it's a daily driver.

maybe i need this later

Hmmm, interesting :3
I'll test it right now
edit: Freeze after playing games or using the 3d drawer

dario3040 said:
Hmmm, interesting :3
I'll test it right now
edit: Freeze after playing games or using the 3d drawer
Click to expand...
Click to collapse
Launcher Pro? Or do you mean adw's 4D?
Sent from my Nexus S

morfic said:
Launcher Pro? Or do you mean adw's 4D?
Sent from my Nexus S
Click to expand...
Click to collapse
Google's stock android 2.2/2.3 launcher.
{
"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"
}

Yea I will say this is the fastest kernel I have ever used, scored averaging 3700, best being almost 4000...Solid work morfic.

adewey234 said:
Yea I will say this is the fastest kernel I have ever used, scored averaging 3700, best being almost 4000...Solid work morfic.
Click to expand...
Click to collapse
Thanks for some positive feedback.
This is the longest tested of all kernels I built.
Simms22 is not much of a tester, his phone runs about any crazy oc and uv I throw at it, so I added time to testing it
Unfortunately no oc kernel will ever run stable on all phones, if they did, samsung would charge us for that.
Sent from my Nexus S

morfic said:
Thanks for some positive feedback.
This is the longest tested of all kernels I built.
Simms22 is not much of a tester, his phone runs about any crazy oc and uv I throw at it, so I added time to testing it
Unfortunately no oc kernel will ever run stable on all phones, if they did, samsung would charge us for that.
Sent from my Nexus S
Click to expand...
Click to collapse
Unfortunately not stable on my device. Ran a couple of quadrants (4015) but froze during smartbench.

morfic said:
Thanks for some positive feedback.
This is the longest tested of all kernels I built.
Simms22 is not much of a tester, his phone runs about any crazy oc and uv I throw at it, so I added time to testing it
Unfortunately no oc kernel will ever run stable on all phones, if they did, samsung would charge us for that.
Sent from my Nexus S
Click to expand...
Click to collapse
Naw it runs 100% on full OC 1440. I only have issues with radio occassionally, but thats because i cant make up my mind which is better, KB1 or KB3. But yea, mos def a daily driver. Ive drained my battery in 1/3 of the time i usually do, playing with my phone so much. Much appreciated sir, if you need more experienced feedback, let me know, I've done some work with android, just dont have the time as of late, but logcats etc are easy and not that taxing on my time

morfic said:
Thanks for some positive feedback.
This is the longest tested of all kernels I built.
Simms22 is not much of a tester, his phone runs about any crazy oc and uv I throw at it, so I added time to testing it
Unfortunately no oc kernel will ever run stable on all phones, if they did, samsung would charge us for that.
Click to expand...
Click to collapse
except for that damned 1.7ghz kernel. one of these days..
jlevy73 said:
Unfortunately not stable on my device. Ran a couple of quadrants (4015) but froze during smartbench.
Click to expand...
Click to collapse
the phone commited murphacide

adewey234 said:
Naw it runs 100% on full OC 1440. I only have issues with radio occassionally, but thats because i cant make up my mind which is better, KB1 or KB3. But yea, mos def a daily driver. Ive drained my battery in 1/3 of the time i usually do, playing with my phone so much. Much appreciated sir, if you need more experienced feedback, let me know, I've done some work with android, just dont have the time as of late, but logcats etc are easy and not that taxing on my time
Click to expand...
Click to collapse
i used(on uv now, will use again) this kernel, as a daily driver, longer than any other. it definitely deserves respect. btw, i use KB1, and also had a hard time decided which radio i liked best

Hey Morfic, I have a question, when you were applying the patches for the kernels with BFS, were you using this file or a different one? 2.6.37-sched-bfs-363.patch

adewey234 said:
Hey Morfic, I have a question, when you were applying the patches for the kernels with BFS, were you using this file or a different one? 2.6.37-sched-bfs-363.patch
Click to expand...
Click to collapse
359 and later 363, but been a while.
Curious to see if superuser still hangs in "VM_WAIT" with 363 now or not.
Sent from my Nexus S

Mine was stable at 1.4ghz, but crashing constantly at 1.44ghz.

terryhau said:
Mine was stable at 1.4ghz, but crashing constantly at 1.44ghz.
Click to expand...
Click to collapse
Exactly the same issue I had.

How can you guys get almost 4000 in quadrant scores.
The maximum I got w/ this kernel was 3100, for me the phone is reallly fast but if you have more I want this more that I don't have.
Thanks,

liquelt said:
How can you guys get almost 4000 in quadrant scores.
The maximum I got w/ this kernel was 3100, for me the phone is reallly fast but if you have more I want this more that I don't have.
Thanks,
Click to expand...
Click to collapse
set your cpu to 1440/1440, then test 4 times in a row. the 1st is always much lower then the rest. the 3rd or 4th test should be more accurate.

First - 3091
Second - 3247
Third - 3770
Forth - 3813
Fifth - 3591
So here is my results, thanks for this FANTASTIC KERNEL.

liquelt said:
First - 3091
Second - 3247
Third - 3770
Forth - 3813
Fifth - 3591
So here is my results, thanks for this FANTASTIC KERNEL.
Click to expand...
Click to collapse
there you go, good ole #4

@800MHZ is the voltage the same or lower than 1GZ?

[JB]Result of Benchmark Comparison with different ROM&Kernels

this is my first thread in xda. sorry if i break any rules. and plz tell me if i did.
============================================================
Hello committee! I just registered here a few days ago and find it a really nice place. i found a lot of useful infos and learned skills here. enjoyed every second reading the posts.
i just spend the afternoon testing with some different kernels and ROMs. trying to figure out which one should i choose for daily use. i know it would be a long term testing to get a conclusion but this is a start anyway.
i'm posting my result here hoping could do a little contribution to this committee. and PLEASE share your ideas with me too!
============================================================
I used the app AnTuTu Benchmark on my Nexus S. dunno if it's the best benchmarking app but i think as long as i keep all the other settings the same. the result still tells some truth.
The results below ONLY show the performance of roms and kernels at BENCHMARKING. it does NOT say anything about battery life, usability or stability.
View attachment 1265857
=======================UPDATE===============================
AOKP seems not working properly with AirKernel. trying to figure out why. anyone can help?
*Air186B65f = AirKernel #186 BlueColor with 65fps
*Air186V = AirKernel #186 VoodooColor without 65fps
(both the above are High-BigMem versions.)
*marOTA4.2= marmiteKernel v4.2 for stockOTA ROM
*mar4.2 = marmiteKernel v4.2 for Custom ROMs
*bfq is not currently available on marmiteKernel v4.2 so i only tested deadline. dunno if it's a proper choice.
*Adroxd = Androxide 3.1.0
*NotPass = benchmarking was interrupted by force close of the app.
some conclusions:
1. ROOT does NOT affect the performance on stock ROMs.
2. performance does NOT do better at benchmarking, as it's supposed to, than some other governers.
3. NotPass does NOT happen every time with the same settings. i.e. i tested 2 times at 1.4GHz with smartassV2 and bfq on stock ROM. the first time it gave an result of 4090. but the 2nd time it force closed.
4. schedulers seem to be not affecting benchmarks on a big scale.
5. The result shows stock ROM with custom kernels DOES better performance than custom ROMs. (I like things to be stock. it makes me feel safe. But at the same time i prefer those mods in custom ROMs like the volume rocker wake-up and advanced power menus.)
6. I love xda-developers.com :laugh:

New benchmark expert arrived!
Welcome!
Edit: aha...not in-progress like bedalus:/
Sent from my Nexus S using xda app-developers app

Since you are new to this let me say something that most users are unaware of: the governor is absolutely irrelephant. The CPU will always be at 100%, otherwise it's not a true benchmark, thus rendering the delay to reach such % (basically what a governor does) a parasite variable.
The I/O scheduler performance is also tricky to measure as some are better in sequential operations, some are not.
So if you want to test ROMs & kernels (which makes much more sense, not I/O schedulers and governors), then here are a few tips:
More ROMs (make sure to include CM as it has many small code optimizations).
More kernels.
Drop the governor column and do everything in performance (this will force the max CPU frequency and eliminate the aforementioned parasite variable).
Choose one I/O scheduler and use it for every kernel (homogeneity).
Don't do OC.
Use more benchmarks (CF-Bench, Sunspider, GLBenchmark).
Run at least 3 times and note down the average result.

ioplkj13 said:
New benchmark expert arrived!
Welcome!
Edit: aha...not in-progress like bedalus:/
Sent from my Nexus S using xda app-developers app
Click to expand...
Click to collapse
LOL no i'm not a benchmark expert, yet. ;D but thanks for the advise. i think in-progress makes ppl excited and helps me to do the work better. just like knzo's post!
knzo said:
Since you are new to this let me say something that most users are unaware of: the governor is absolutely irrelephant. The CPU will always be at 100%, otherwise it's not a true benchmark, thus rendering the delay to reach such % (basically what a governor does) a parasite variable.
The I/O scheduler performance is also tricky to measure as some are better in sequential operations, some are not.
So if you want to test ROMs & kernels (which makes much more sense, not I/O schedulers and governors), then here are a few tips:
More ROMs (make sure to include CM as it has many small code optimizations).
More kernels.
Drop the governor column and do everything in performance (this will force the max CPU frequency and eliminate the aforementioned parasite variable).
Choose one I/O scheduler and use it for every kernel (homogeneity).
Don't do OC.
Use more benchmarks (CF-Bench, Sunspider, GLBenchmark).
Run at least 3 times and note down the average result.
Click to expand...
Click to collapse
these are very helpful! i'll take the advise and make my test better.
but i don't think i'll do A LOT ROMs and kernels. since i'm just testing those ROMs and kernels which i might be using for daily. but i WELCOME all other guys here to post their results! :fingers-crossed:
again, many thanks to your help!

So the best rom+kernel is the stock+air kernel?

We need more numbers. A benchmark thread with nearly only conclusions is just asking people to believe you without any proof. Not gonna happen in a place full of skeptics.
Knzo suggestions are very good too. You need more homogeneity in your benchmarks. Being all over the place only adds variables in the equation that favor one kernel or the other. And more tests. Pretty much just follow all what he said.
knzo said:
Since you are new to this let me say something that most users are unaware of: the governor is absolutely irrelephant
Click to expand...
Click to collapse
{
"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"
}
Irrelevant*

polobunny said:
Irrelevant*
Click to expand...
Click to collapse
You don't really think I'd write irrelephant unintentionally right? I'm meme-aware! ;p

knzo said:
You don't really think I'd write irrelephant unintentionally right? I'm meme-aware! ;p
Click to expand...
Click to collapse
Haha! You got me there. I kinda knew of your meme ways, but that was some good trolling right there.

•Run at least 3 times and note down the average result.
Click to expand...
Click to collapse
Try to make the conditions the same for each run. For example, restart the phone and leave idle for 3 minutes, etc.

Andrew1998 said:
So the best rom+kernel is the stock+air kernel?
Click to expand...
Click to collapse
remember there's nothing called "best". as i said above the result only shows performance at benchmarking. as we know good benchmark score may result in poor battery life so it's not recommended for daily use.
if you could help this work it's much appreciated!
maybe you just do a benchmark (takes only 3 mintes) and post your rom and kernel with the score.
best wishes.

vikingsail said:
this is my first thread in xda. sorry if i break any rules. and plz tell me if i did.
Click to expand...
Click to collapse
Well since you asked nicely, you did post in the wrong section.
Thank you for the data, and there have been some good suggestions about tweaking the method.

Stock Rom.. air kernel #186..
Sent from my Nexus S

siddhesh.raikar said:
View attachment 1267462
Stock Rom.. air kernel #186..
Sent from my Nexus S
Click to expand...
Click to collapse
that is very impressive! any tweaks?

polobunny said:
We need more numbers. A benchmark thread with nearly only conclusions is just asking people to believe you without any proof. Not gonna happen in a place full of skeptics.
Knzo suggestions are very good too. You need more homogeneity in your benchmarks. Being all over the place only adds variables in the equation that favor one kernel or the other. And more tests. Pretty much just follow all what he said.
yes that's indeed what we need. i wish more ppl could take part in this.
and i'm trying to take more homogeneity into my testing. thanks for the advise!
Irrelevant*
Click to expand...
Click to collapse
sysadmn said:
Try to make the conditions the same for each run. For example, restart the phone and leave idle for 3 minutes, etc.
Click to expand...
Click to collapse
very funny...

siddhesh.raikar said:
View attachment 1267462
Stock Rom.. air kernel #186..
Sent from my Nexus S
Click to expand...
Click to collapse
Haha you over clocked to 1.4ghz. Here is mine.
Sent from my Nexus S using xda app-developers app

Good luck!

..

Kernels battery benchmarks

Hello everyone...i know we all share the same question...which kernel is better in terms of battery...so i decide (I had done this before but an admin lock the threat) but i decide to do it again..cause i really want this anwer, at least a part of it, to be answered...so i decide to stick with one rom, difficult don't try this, and simply flash all available kernels there is out there...run AnTuTu battery benchmark and compare the results
Im not so good with words so here it goes...
Rom Aokp Preview...I know all you start saying that this is a preview rom i should test this with a more stable rom...but all i want is to test the kernel it self...not the rom...so none or if a bit of rom bugs/drains will cause this test to fail cause i didnt use the phone at all, the whole 2 hours of the benchmark...
All my apps..i have at least Android is upgrating said 150 apps Titanium sais 98 without the system...i want to see the kernel working at a normal setup...maybe its nothing but who knows...
Wifi on 3G off...as i said i want this to be run at normal usage..we all have wifi open if not 100% 85% of our usage...
And for the end...i left all kernel settings pure default...i dont use FKU,Fauxclock,Semamanager or not even Trickster...if a dev as for example @morfic wants Trinity to work 2cores its fine by me...I didnt change nothing no UV no OC no UC no nothing....
The test needs 100% battery to run...of course..and a straight 2H and some minutes depenting on kernel to run from 100% to 73% to 46% and finaly to 19% by mesuring the time it took it to drain all this juice we found the drain per minute that we need to compare our kernels...
Test also mesures battery temp..so we mesure average battery temp...i have cpu temp widget but i can't stare the screen all this hours to check the temp when a sertane percedance is reached...also...for those who wanter why i haven't test francos yet...i also said this to @franciscofranco on google+ at 85%+- of the test i notice that the cpu gets really hot..im talking about 86C so the test stops..i manage to test it once by leaving my phone to my floor that cold but i mistakanly forgot 3G on and the test finish on 40minutes!
Anyway..here is my spreadsheet feel free to comment if i made something wrong specially on sums..and on the attachments i will add 3 pics of each kernel showing the end screen of the test,the rating of the test and the drain line on settings...
26/09/13:I decide to enable UnderVolt to the kernels from here and on...Matrix 11 just like franco...and semaphone for some strange reason don't have a proper thermal throtle...at 86C the test just stops...so im hopping i can keep it below that temp to be able to run the test!!
Thanks and remember this is just benchmark..feel free to test kernels yourself to see which one fits you best...and sorry that i kill your language!!
Next kernel: Bricked
21/09/13 Base rom changed to PSX....all kernels will be re-tested...sorry for the delay...
22/09/13:Finish benchmark for Faux123 26m
23/09/13:Finish benchmark for Trinity T4h
24/09/13:Finish benchmark for Stock 4.3
25/09/13:Finish benchmark for Franco 184*Retest needed
26/09/13:Finish benchmark for Matrix 11**-112mV enabled
26/09/13:Finish benchmark for Sema 1.6.5**-100mV enabled
27/09/13:Finish benchmark for AK 25+
28/09/13:Finish benchmark for NEO 12**-112mV enabled
28/09/13:Finish benchmark for Hells-core B38**-125mV enabled
29/09/13:Finish benchmark for Franco r184**-125mV enabled
30/09/13:Finish benchmark for Reborn90 r184**-125mV enabled
1/10/13:Finish benchmark for Bricked 20/9**-125mV enabled
{
"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"
}

Wow I can already tell this is going to be very useful and answer a lot of questions, thanks in advance for this!!

a thought maybe.. how about adding the default nexus 4 android kernel to the graph? it would make the comparison more interesting

I think this is very utility, wating for franco kernel.

Wow OP.. By any chance after these test could you use the best kernel out of these test and also do a analysis on the radios of mako. I think 3g test can predict the difference on battery drain using each one of them.

simms22 said:
a thought maybe.. how about adding the default nexus 4 android kernel to the graph? it would make the comparison more interesting
Click to expand...
Click to collapse
I could use the stock 4.3 kernel to check the difference...its a good advice....
arjunv said:
Wow OP.. By any chance after these test could you use the best kernel out of these test and also do a analysis on the radios of mako. I think 3g test can predict the difference on battery drain using each one of them.
Click to expand...
Click to collapse
With 3G open the results are gonna hurt many devs...as i said the only success test with franco kernel with 3G open but not in use cause of open wifi...the test from 2.04hours of faux it goes to 40minutes with franco....im using hybrid .34 modem right now....even than im not be able to get any 4G signal... :/
Also thanks for you kind words....i thought by now the threat cause be locked by a admin like the 1st one!!

Nice thread! Will solve some general problems regarding kernels maybe.

3mperror said:
Nice thread! Will solve some general problems regarding kernels maybe.
Click to expand...
Click to collapse
As i said they are just benchmarks...its gives an idea of what each kernel can do...at least of a non stop benchmarking point of view...
I finished semaphones benchmark but after the reboot i saw that from 18% the battery goes to 36% so there is something wrong with the batterystat.bin file...so i have to redo this benchmark so i can be as accurade as i can!!!

Very usefull thread,good job
Sent from my Nexus 4 using Tapatalk 4

Correct bench of Semaphone 1.6.5 next kernel again..franco r185!!!

intresting.....
waiting for Matr1x results

Some one out there have been doing the exact same thing over the past week with me then ! hahaha - I have Carbon Rom as a base, I have had it on Franco, Faux, Sema & Matrix - All same setup, same apps, same usage - And the best results I have had so far is Carbon + Matrix Kernel. Looking forward for your experience.

Ensifolk said:
Some one out there have been doing the exact same thing over the past week with me then ! hahaha - I have Carbon Rom as a base, I have had it on Franco, Faux, Sema & Matrix - All same setup, same apps, same usage - And the best results I have had so far is Carbon + Matrix Kernel. Looking forward for your experience.
Click to expand...
Click to collapse
you didnt do trinity kernel, id consider it one of the major kernel :crying:
just teasing
did you do the stock kernel, to use as a base?

simms22 said:
you didnt do trinity kernel, id consider it one of the major kernel :crying:
just teasing
did you do the stock kernel, to use as a base?
Click to expand...
Click to collapse
Simms mate! Hehehe, first, love your pics on twitter! Second, I am still in the testing process - Gonna flash Trinity on a day I know Im gonna be on full wifi and free on time - I want the test to remain objective and under the same past conditions.

S2 has/had a similar thread. Great!

Excellent thread, I cannot wait till you finish your table, and btw, thanks for doing this man, big big help for all of us
Sent from my Nexus 4 using xda app-developers app

Thanks for all your kind words!!!

Is the antutu battery tester only causing heavy load? Because that would ignore kernel's idle drain
Also, I'd chose always the most recent kernel version to test. hellcore is at b37 right now (maybe just a typo?)

ccab said:
Is the antutu battery tester only causing heavy load? Because that would ignore kernel's idle drain
Also, I'd chose always the most recent kernel version to test. hellcore is at b37 right now (maybe just a typo?)
Click to expand...
Click to collapse
Good question !
Sent from my Nexus 4 using xda app-developers app

This is a very use full piece of work, thank you for your time, and I look forward to seeing how matrix does. one point to note as well as the comments about idle, is that different kernels behave differently sometimes depending on the ROM that they are paired with, so possibly repeating the test with a cyanogen based ROM and some 3rd ROM for all of the kernels before taking an average might be an interesting test. But of course it is up to you, I don't know how much time you have :thumbup:
Sent from my Rooted NEXUS 4