di11i kernel beta .01 (unsecure boot.img- root shell, adb remount) ext4 tweaks/init.d

Search This thread

Lock-N-Load

Senior Member
Aug 25, 2010
1,618
392
Under Your Skin
UV is most of the times placebo, unless you UV by big values, that are most if the time not supported by CPU.

It's generally a source if instability rather than longevity. Some of our kernel devs made extensive tests and with secure UV there wasn't any difference in battery life (I speak of my phone if course but it's the same thing).

Calculate in % what does represent 25, 50 75mV less compared to CPU voltage....

Agreed. In phone testing dozens of kernels in high medium & low voltage across several phones there were never more problems - consistent, repeatable - than with low voltage kernels. As a rule I steer clear despite the obvious "supposed" benefits.... they just never really panned out reliably.
 

Diamondback

Retired Dev Committee Lead / Retired Senior Mod
Jan 17, 2010
4,476
6,631
virtuous-ten-studio.com
UV is most of the times placebo, unless you UV by big values, that are most if the time not supported by CPU.

It's generally a source if instability rather than longevity. Some of our kernel devs made extensive tests and with secure UV there wasn't any difference in battery life (I speak of my phone if course but it's the same thing).

Calculate in % what does represent 25, 50 75mV less compared to CPU voltage....

Agreed 100%.
My team has quite some experience over a wide range of different devices and UV never really showed any improvement over stock voltage tables.

It's not the CPU that drains the major power, it's the display and the radio.
Try putting you phone to airplane mode and just let it rest on table until it's dead. Then do the same test with wifi and data enabled.
You will be schocked :p

I for one won't use any UV'ed kernels at all for my ROMs, they only induce major stability problems to a device that has these problems already without UV ;)
Undervolting kernels is a huge myth, someone from google (Dianne?!) should really clear this up.
 

demandarin

Senior Member
Apr 7, 2010
7,021
2,038
Alexandria, Va
Agreed 100%.
My team has quite some experience over a wide range of different devices and UV never really showed any improvement over stock voltage tables.

It's not the CPU that drains the major power, it's the display and the radio.
Try putting you phone to airplane mode and just let it rest on table until it's dead. Then do the same test with wifi and data enabled.
You will be schocked :p

I for one won't use any UV'ed kernels at all for my ROMs, they only induce major stability problems to a device that has these problems already without UV ;)
Undervolting kernels is a huge myth, someone from google (Dianne?!) should really clear this up.

please don't mention the name Dianne before you know who comes in quoting everything she said since birth..lol there has other Google engineers who know just as much or more..lol
 

Diamondback

Retired Dev Committee Lead / Retired Senior Mod
Jan 17, 2010
4,476
6,631
virtuous-ten-studio.com
please don't mention the name Dianne before you know who comes in quoting everything she said since birth..lol there has other Google engineers who know just as much or more..lol

Hm? Dianne Hackborn did a great job so far on de-mythtifying a lot of things custom ROM developers used to do, including the wipe battery stats thing.

Not sure what your point is here...
 

demandarin

Senior Member
Apr 7, 2010
7,021
2,038
Alexandria, Va
Hm? Dianne Hackborn did a great job so far on de-mythtifying a lot of things custom ROM developers used to do, including the wipe battery stats thing.

Not sure what your point is here...

it was an inside joke. you didn't get it..lol. someone in general is always quoting her left n right. you are correct though...Dianne Hackborn did do a great job and is very knowledgeable on these things.
 

benefit14snake

Senior Member
Dec 23, 2011
819
174
Henrico, VA
Agreed 100%.
My team has quite some experience over a wide range of different devices and UV never really showed any improvement over stock voltage tables.

It's not the CPU that drains the major power, it's the display and the radio.
Try putting you phone to airplane mode and just let it rest on table until it's dead. Then do the same test with wifi and data enabled.
You will be schocked :p

I for one won't use any UV'ed kernels at all for my ROMs, they only induce major stability problems to a device that has these problems already without UV ;)
Undervolting kernels is a huge myth, someone from google (Dianne?!) should really clear this up.

Aww makes sense. Thanks for clearing that up for me diamond.

Sent from my Transformer Prime TF201 using Tapatalk
 

di11igaf

Inactive Recognized Developer
Sep 6, 2010
1,898
739
East Coast
Undervolting a little across the board will probably do next to nothing, but undervolting at low clock speeds where the device is sleeping will without a doubt increase battery in standby mode. With on-demand governor, the CPU is usually only at a few clock speeds, so undervolting at those specific clock speeds will make some difference, depending on the under volt.
If undervolting made no difference, oems would always run the CPU at the highest clock speed, like the second gen snapdragon.
A 1.5 ghz CPU that when released on the g2 ran at 800mhz.(lower voltage than 1ghz) Most of the rest released at 1ghz(less voltage than 1.5 ghz)
Same with the dual core version. Less voltage always = less heat, more battery. How much depends, could be next to nothing, at some clock speeds on some CPU's it could be significant.
Not to mention you can get to a higher clock speed with about the same amount of heat.
Instability depends on the device and specific CPU, two devices with the same cpus could have two very different voltage requirements, so by default the OEM has make the voltage high enough for EVERY device, when a lot of them could get by with a lot less, some will require the full voltage.
 
Last edited:

benefit14snake

Senior Member
Dec 23, 2011
819
174
Henrico, VA
Undervolting a little across the board will probably do next to nothing, but undervolting at low clock speeds where the device is sleeping will without a doubt increase battery in standby mode. With on-demand governor, the CPU is usually only at a few clock speeds, so undervolting at those specific clock speeds will make some difference, depending on the under volt.
If undervolting made no difference, oems would always run the CPU at the highest clock speed, like the second gen snapdragon.
A 1.5 ghz CPU that when released on the g2 ran at 800mhz.(lower voltage than 1ghz) Most of the rest released at 1ghz(less voltage than 1.5 ghz)
Same with the dual core version. Less voltage always = less heat, more battery. How much depends, could be next to nothing, at some clock speeds on some CPU's it could be significant.
Not to mention you can get to a higher clock speed with about the same amount of heat.
Instability depends on the device and specific CPU, two devices with the same cpus could have two very different voltage requirements, so by default the OEM has make the voltage high enough for EVERY device, when a lot of them could get by with a lot less, some will require the full voltage.

The cores towards the center of the wafer usually tend to clock the best.
Edit: unless there are donut shaped variations lol

Sent from my Transformer Prime TF201 using Tapatalk
 
Last edited:

benefit14snake

Senior Member
Dec 23, 2011
819
174
Henrico, VA
is this wafer you speak of edible?

No myself and a bunch of guys on a PC forum used to pick cpus based on their serial because that tells you when it was made and what part of the wafer it comes from. Its been a few years but I think the closer to 001 in a given batch had the best potential. That was in my sub zero cooling days lol.

Edit: took a single core opteron 64 from 1.8 to 3.4ghz. Locked multi =) fun stuff. That chip would do 2850mhz stock voltage.
Talk about running out of spec, its internal clock was running almost double.

Sent from my Transformer Prime TF201 using Tapatalk
 
Last edited:

di11igaf

Inactive Recognized Developer
Sep 6, 2010
1,898
739
East Coast
Some cpu's love overclocking, i had an inspire kernel(inspire was 1ghz) that could hit 2.0ghz.
And loks like a small overclock may be fairly easy, theres already code in the kernel for 1.7ghz, just a matter of making it an available userspace overclock like we are now. Ots just a matter of making it available in the kernel so we can use it. Might not make v1, but hopefully soon.
 
Last edited:

Diamondback

Retired Dev Committee Lead / Retired Senior Mod
Jan 17, 2010
4,476
6,631
virtuous-ten-studio.com
Some love overclocking, i had an inspire kernel(inspire was 1ghz) that could hit 2.0ghz.
And loks like a small overclock may be fairly easy, theres already code in the kernel for 1.7ghz, just a matter of making it an available userspace overclock like we are now. Ots just a matter of making it available in the kernel so we can use it. Might not make v1, but hopefully soon.

I guess you need to edit those cpu modes too? If this is the case i'd say you make the 1.5ghz mode to max out at 1.6 ghz and make the 1.6gz mode max out at whatever max frequency you set.
 

di11igaf

Inactive Recognized Developer
Sep 6, 2010
1,898
739
East Coast
I guess you need to edit those cpu modes too? If this is the case i'd say you make the 1.5ghz mode to max out at 1.6 ghz and make the 1.6gz mode max out at whatever max frequency you set.

Ya I still gotta look into it more, but that's a possibility. I have to first enable the 1.7 ghz mode in the kernel, then user space. If that don't work, I may be able to turn the 1.6 ghz mode into 1.7. Gonna look into it tonight a little
 

di11igaf

Inactive Recognized Developer
Sep 6, 2010
1,898
739
East Coast
Have you been able to compile a kernel from source yet? I am still having errors.

Yes, I haven't pushed the fixed Asus source to my github, but check AndroitRoot's github and you'll see tf201 source. Those guys fixed the compile errors and I got it to compile last night.
Clone theirs and you should be good, one of my last few posts in another thread I uploaded the stock .config I pulled from device. If you need it search my last few posts.
 
  • Like
Reactions: jermaine151

benefit14snake

Senior Member
Dec 23, 2011
819
174
Henrico, VA
Flashed perfectly. I actually flashed through terminal emulator and it worked great!

Getting 13 mbs write and 29 read. Not sure if those are supposed to change.

Sent from my Transformer Prime TF201 using Tapatalk
 
Last edited:

di11igaf

Inactive Recognized Developer
Sep 6, 2010
1,898
739
East Coast
I flashed this thinking i didnt need to be unlocked and bricked my prime :( im so stupid... Guys im 15 and my parents will kill me if they find out. Will Asus rma it or am i screwed? Im getting the Unrecoverable Bootloader error 0x00000008 :/

Oh boy, to make a backup you need to have recovery, which means unlocked boot loader. But what's done is done, and since your boot loader isn't unlocked rma may be your only option, but I can't tell you what to do. I know it sucks man considering your 15 and all, but bricks can and DO happen there is a risk no matter how small.
 

ThunderRootedDragon

Senior Member
Jul 21, 2011
349
35
Ahhhhaha...clearly the first dozen or so posts is chatter btw the pros -- interesting, but honestly can't fully follow.

Coming from motley and I'd like to flash this -- anybody else care to comment on how its going for them? Batt life is decent on motley but it spikes up to 1.8 for everything (I knew it was stuck at 1.8, but thought it would scale a little bit) and I think batt life is suffering more than it should for whatever benefit I get from a faster cpu.

Any thing I should know about concerning the tweaks before I flash?


Thanks.
 

Top Liked Posts

  • There are no posts matching your filters.
  • 6
    This basically (for right now) the stock kernel image with some tweaks to init.
    -unsecure boot.img-(perma- rooted adb shell)
    -adb remount(mount /system r/w)
    -init.d support added in for init scripts(survive after reboot)
    -ext4 filesystem mount tweaks-/system/data/cache(faster I/O)
    (Actually have a full ext2 version for,****s and giggles, but it boots into the encryption error screen, so cant release that. May have a fix though)
    -mem and cache tweaks coming soon

    still working on the actual kernel, will release that next, still have lots of work to do

    boot.img MD5- fca41dba8f4699b67fd461a1632b65cf

    MAKE BACKUP FIRST

    You will not have issues if you wipe data, chances are youll be fine if you dont, but if anything starts acting up just wipe data, then install boot.img
    This is the actual boot.img and for now has to be flashed with adb with these EXACT commands--
    Code:
    adb push boot.img /sdcard/
    adb shell
    su
    dd if=/sdcard/boot.img of=/dev/block/mmcblk0 seek=3968 bs=4096 count=2048
    reboot
    
    
    #Thanks Diamondback and friends for offsets




    This is not the ideal way to flash this, but i can not get the blob file to staging partition to actually write the image. Im pretty sure its an issue with the signature of the blob file, so if someone wants to pack this into a blob id be glad test it and then update the OP.
    My primes been using this boot.img for a few days, along with one other. Flash at your own risk.
    NOT COMPATIBLE WITH CM-9
    3
    if you copy the blob, while p4 is mounted to staging in cwm, you will have a problem when you reboot, as the filesystem umount, and that cause a small change to the mounted filesystem, to indicate it is umounted normally. and the p4 partition becomes invalid.

    so it can easily become a problem if p4 is mounted.

    at least thats how i experienced the problem when seen initially.

    That's why I said to always add an unmount("/staging"); to the top of your script. The attached flashable zip works perfectly on the official CWM; staging is not mounted by default, but I told the script to unmount it anyway in case someone with the unofficial CWM tries to flash it. If it's already unmounted, the script just continues.

    Anyone may use this as a template for consistently flashing a kernel to staging.
    2
    UV is most of the times placebo, unless you UV by big values, that are most if the time not supported by CPU.

    It's generally a source if instability rather than longevity. Some of our kernel devs made extensive tests and with secure UV there wasn't any difference in battery life (I speak of my phone if course but it's the same thing).

    Calculate in % what does represent 25, 50 75mV less compared to CPU voltage....

    Agreed 100%.
    My team has quite some experience over a wide range of different devices and UV never really showed any improvement over stock voltage tables.

    It's not the CPU that drains the major power, it's the display and the radio.
    Try putting you phone to airplane mode and just let it rest on table until it's dead. Then do the same test with wifi and data enabled.
    You will be schocked :p

    I for one won't use any UV'ed kernels at all for my ROMs, they only induce major stability problems to a device that has these problems already without UV ;)
    Undervolting kernels is a huge myth, someone from google (Dianne?!) should really clear this up.
    2
    Well, I seriously doubt any under volt will be in the first few versions unless I get lucky. One of the source files where I believe I may be able to get a simple over clock is well over 4000 lines of code. If it doesn't work as planned, who knows when . This isn't even accounting for any voltages, just clock speed. Tegra 3 is DEFINITELY a complex beast.
    Hopefully someone like bumblebee has already been looking into it and has a better grasp than me.(edit just looked at his git, nothing for tf-201. I'm sure there's more people than me looking into it)
    2
    Hopefully have the optimised kernel out soon, hopefully tomorrow depending how late I stay up tonight.