[ROM][GEN2]CyanogenMod 7 (Android 2.3.7)

[Guest emc02]

Ok. The CPU uses flip all power. :D To test the test, I ran the stability test for 70 minutes at normal voltage at 729mhz. It's completing the tests in its sleep (it logs the progress) but it started at 50% and is still at 50%. The power used by the CPU appears to be insignificant whatever you have it doing!

729MHz ist too much, because there wont be some undervolting at this frequency (thats why I meant to test at 245MHz)

and we all know, that the battery monitor isnt accurate, so the only way to really test battery drain is to test from fully charged to 0%

[Guest targetbsp]

I know it's too high. But my point is, if the battery isn't going down after 90 minutes of flat out cpu use at 729 I start to wonder if the amount of power we save by slightly lowering the voltage at low frequencies has any significance at all.

Consider the amount you have to force the situation to even be saving this power. Idling, nothing is saved, in use nothing is saved because it's at high frequency. So now we're only saving power in a fairly narrow window of the cpu being used but not that intensely (we assume). How often does that occur in real use? And if the power the cpu is using is nothing next to things like the screen and wifi... do we even care? :D

The stability test does have a nice relevant test for this discussion though (which I've not tried). It has a specific Undervolt stability test which will loop through all the supported frequencies stress testing the CPU to make sure it doesn't die through lack of power. Might be worth a go for those going to stick with the UV kernel.

Edited January 27, 2012 by targetbsp

[Guest emc02]

well, if there is no amount of power that can be saved, it would not be neccesary to test the stability

and I think it would save power in idle mode on low frequencies (if it works as supposed to)

[Guest sumdumguy]

Can't you leave the screen on for a power test, just set it to never turn off.

btw, anyone else have wifi-lockup?? problems with cm7.. where programmes such as 'touch (pingchat)' stall when in sleep mode using wifi (but not when charging) meaning messages not getting through @50% of the time ? the same happened to me with the Lewa rom, but all seems fine using a standard type rom like fish and chips. It's sahme because i much prefer the latest Kang CM7.

[Guest jackal-sk]

A simple test would be:

1) charge to 100%

2) set to flight mode

3) set min and max clock speed to 245MHz

4) run an CPU Benchmark and loop till battery is empty and keep screen off (benchmark should run with screen off)

check the time needed till battery is empty

then take the non-undervolting-kernel and redo 1) - 4) and compare the time!

maybe someone has a Blade that is not in daily use to do the test

This is a great idea of confirming if the undervolting works.

[Guest targetbsp]

Can't you leave the screen on for a power test, just set it to never turn off.

Unfortunately I think the screen uses a vast percentage of the power so you wouldn't see much of a power saving if the only place your saving it is the CPU.

[Guest De@n]

Here ya go: http://www.mediafire.com/?a59h7piihvbonby

MD5# 55a58ff6956e99c2b816f1f9a2bc8f3e

A vanilla CM7 compiled from source with no changes. Includes all fixes up to and including "Rotation animation, fix for devices with ro.sf.hwrotation set to something other than 0." from http://cm-nightlies.appspot.com/?device=blade

Cheers :)

[Guest targetbsp]

Ok. The CPU uses flip all power. :D To test the test, I ran the stability test for 70 minutes at normal voltage at 729mhz. It's completing the tests in its sleep (it logs the progress) but it started at 50% and is still at 50%.

Ah ha! Good news. I had an idea which I tested and it panned out. Whilst the benchmark does indeed run whilst the phone sleeps its at a fraction of the performance of when it is awake. Hence the lack of CPU usage - it actually isn't doing anything most of the time.

I'm going to set up tasker to keep the phone awake but screen off during the test. We're potentially back in business again. :D

[edit]

And this does now hit the battery. :)

Edited January 27, 2012 by targetbsp

[Guest sm4tik]

1. I don't have a github payed account (to store more than 300Mb).

2. I prefer to keep my work (sources) out of all others. But I can share results.

I don't even have a Blade))

Though I respect your choice not to share the code, I really don't understand why. And to talk about a >300Mb diff is a bit of an overkill, don't you think? Anyways, keep up the good work, I see some people digging it here :)

edit: btw, when setting the lockscreen widget layout to right, does the carrier label follow along or is it just me that has it _left_ behind (ha.ha..)..? For me the carrier works as expected when aligned to left/center, but when going for right it will just go back left.

Edited January 27, 2012 by sm4tik

[Guest 569]

A fully charged battery has about 3.7 V. A depleted battery has say 3 V (?).

According to the theory of some guys here, would it mean: the lower the battery level, the less power the blade (or any device) will consume?

[Guest Akash P]

Has anyone tried this app??? Its a cpu control app which allow us to Set our UV settings, if uv is enabled in the kernel..My links Seems to be a good app coming from the developer of RAM Manager pro!.

EDIT:

Never mind does not work.

Edited January 28, 2012 by Akash P

[Guest Akash P]

"Weekly 2" (01.13.12)

http://www.mediafire...5hi6z7xaurxing9

Could you compile a newer weekly and post it in cm7 forum. Please.

[Guest Maxsas360]

Has anyone tried this app??? Its a cpu control app which allow us to Set our UV settings, if uv is enabled in the kernel..My links Seems to be a good app coming from the develore of RAM Manager pro!.

We dont have the undervolt driver in the kernel. Our kernel is undervolted manually in the kernel itself. None of the apps cannot change the undervolt values in the kernel without the undervolt driver. Because It causes instability.

[Guest Akash P]

We dont have the undervolt driver in the kernel. Our kernel is undervolted manually in the kernel itself. None of the apps cannot change the undervolt values in the kernel without the undervolt driver. Because It causes instability.

Is there any way of enabling undervolting in the kernel? Just for testing purpose.

[Guest 569]

I was looking for an explanation for what undervolting is and found this:

4) ***Undervolting***

In the main profile page, you will see different multipliers and voltages. What are multipliers?

Multipliers are incorporated with Intel SpeedStep Technology (see down below for full definition). Instead of the CPU running at full power all the time, multipliers are used to dynamically lower the clock speeds to make the CPU run more efficiently.

The more CPU power you need, the higher the multiplier your CPU will use. The higher the multiplier, the higher the corresponding clock speed will be. The higher the clockspeed, the more voltage it requires to run (vice versa)

i) First, We are going to start lowering the highest multiplier's voltage which is the 11x multiplier for me.

-Most people should be able to lower it by at least .100v to begin with the initial testing then slowly decreasing it from there on

ii) Always click Apply so the new voltage can take effect then do the stability test below. Keep doing this everytime you lower the voltage

When I understand it correctly then undervolting is just a another way to set the clockspeed.

It gives the possibility to change the clockspeed to a non standard clockspeed. That is logical to me.

First I mistakenly thought undervolting is to provide a lower voltage, but having the same clockspeed.

If my understanding is correct, then undervolting is something I definitely don't need.

[Guest jurrasstoil]

Then you haven't read your quote correctly.

It reduces the voltage (hence the name). The different clockspeeds (and their multipliers) will be the same with or without undervolting, but since different clockspeeds need different voltages to be stable you just can't set a general undervolting value.

For example let's say your blade clocks at 700mhz under load and at 122mhz when idling.

(I'm pulling voltages out of my a** now) Say the stockvoltages for 700mhz are 0.8V and for 122mhz 0.4V.

Now you could say "well, I undervolt it by -0.2V". Now 700mhz would be at 0.6V and stable, but 122mhz would be at 0.2V and not be stable. So you'd adjust it independently. So maybe 700mhz will work with 0.55V without crashing, but 122mhz needs 0.35V or it will get unstable.

It could also be that 700mhz works with 0.7V and would crash below but 122mhz is able to run stable with only 0.15V.

[Guest Maxsas360]

So I finally succeeded to build my own CM7 kang. now I want to build my own undervolt kernel. Where could I start? And how to use those tar.bz2 and .patch files from sej's server?

[Guest targetbsp]

When I understand it correctly then undervolting is just a another way to set the clockspeed.

You're reading it the wrong way around. Less power does not lead to lower speed. Lower speeds lead to the need for less power.

Jurrasstoil explained it well. Each speed has its own power needs but by default they tend to be set to receive more power than needed to keep them stable despite manufacturing differences between phones. For instance, mine receives enough power at 604 to keep it stable up to 729. Hardcore overclockers at this point then overvolt to clock further (there is a point where this will go horribly wrong :D). But you can equally well infer that at 604 (and probably lower frequencies) the CPU could actually be given less power and remain stable.

The question is: Does it work setting it in software or is it fixed in hardware no matter what you change? And is the CPU using enough battery anyway for the power saving to matter? Tune in tomorrow for the answers to these questions and life the universe and everything! :D

Edited January 28, 2012 by targetbsp

[Guest brickornot]

so its great that there is undervolting.

But still not much about usb host mode for current kernel ?

[Guest pan05]

Mirror for sej's builds:

http://vs247021.vserver.de/android/

Going to keep it up-to-date, so write it down if you like.

Could anyone please point me to the right direction on how to apply these patches to the cyanogenmod sources? I am successfully building from sources after following a very detailed guide found here, but i can't figure out how to apply the patches to my custom build.

Thank you in advance.

[Guest sej7278]

Could anyone please point me to the right direction on how to apply these patches to the cyanogenmod sources? I am successfully building from sources after following a very detailed guide found here, but i can't figure out how to apply the patches to my custom build.

Thank you in advance.

first you change into the directory where the patch is to be applied, then run "git apply /path/to/patch.patch"

cd ~/cm7/frameworks/base/

git apply ~/battery1.patch

git apply ~/battery_colour1.patch

git apply ~/softbuttons.patch


cd ~/cm7/packages/apps/CMParts/

git apply ~/battery2.patch

git apply ~/battery_colour2.patch


cd ~/cm7/vendor/tmobile/themes/Androidian/

git apply ~/battery3.patch


cd ~/cm7/vendor/tmobile/themes/Cyanbread/

git apply ~/battery4.patch

you might also want to use jacob's repository for the proprietary vendor files instead of the extract-files.sh script

git clone  git://github.com/HCDRJacob/proprietary_vendor_zte.git

cp -R ~/proprietary_vendor_zte/blade/ ~/cm7/vendor/zte/

the hardest one is the kernel as you have to patch tom_g's repository and then copy the zImage file it generates into the cm7 repository. edit: i've made a patch against the current kernel now, so its a bit easier than copying all the files into place (had to remove some whitespace from vocpcm.c to make it merge)

git clone git://github.com/TomGiordano/kernel_zte_blade.git

cd ~/kernel_zte_blade/

git apply ~/undervolt_recording_kernel.patch

cp ~/kernel_zte_blade/arch/arm/configs/cyanogen_blade_defconfig ~/kernel_zte_blade/.config


make ARCH=arm CROSS_COMPILE=~/cm7/prebuilt/linux-x86/toolchain/arm-eabi-4.4.0/bin/arm-eabi-


cp ~/kernel_zte_blade/arch/arm/boot/zImage ~/cm7/device/zte/blade/kernel

cp ~/kernel_zte_blade/drivers/staging/zram/zram.ko ~/cm7/device/zte/blade/prebuilt/

once that's all done its just the usual "source build/envsetup.sh && brunch blade"

Edited January 29, 2012 by sej7278

[Guest pan05]

Thanks sej7278.

I will try to follow your detailed instructions.

[Guest ilopez85]

Weekly 4 (01.13.27) (GEN2)

• ~160MB system partition needed
• ro.build.date=vie ene 27 21:15:21 CET 2012
• persist.sys.use_16bpp_alpha=0 corrects certain graphical errors (at the expense of a slower scroll, supposedly)
• Including all changes in the sources of Cyanogen on January 27 (to fix issues on large erase block size (android_bootable_recovery))
• Removed AndroidTerm, Development, Provision, QuickSearchBox, RomManager and SpareParts
• Replaced the folder media/audio with the original Gingerbread one, and ringtone, notification and alarm tones in build.prop
• Default in Spanish (I'm from Spain...)
• Added the gapps and updated all applications on January 27 except Maps 6.0.3 (6.2.0 integration still has given me problems)

http://www.mediafire...6597v6ka0jc66kv

Edited January 28, 2012 by ilopez85

[Guest sej7278]

Thanks sej7278.

I will try to follow your detailed instructions.

i made a monolithic patch for the kernel that adds undervolting, recording and turns off zram debugging, so now just apply the patch and copy the defconfig into place. its easier than copying/renaming a bunch of files. updated the instructions above, and the downloads

if you can compile a regular cm7 kang, then applying the battery bar and softbuttons patches should be no problem, then once you've got that working try the kernel patches.

you'll need to read up a bit on git, useful commands are:

git clone <then the github address>

git reset --hard HEAD 

git status

git diff

git pull

git apply --check <then the patch filename>

[Guest Akash P]

Weekly 4 (01.13.27) (GEN2)

http://www.mediafire...6597v6ka0jc66kv

WooHoo!!! That was fast, Thanks mate. Always preferred Gapps pre-integrated into the Rom.

Thanks Again.