[Kernel]N8000 ICS/JB [1704Mhz OC kernel]

Search This thread

esgie

Senior Member
Feb 17, 2013
332
179
Well, my device hasn't got any issues when running 1600mhz at 1375mV, no throttling or temperature changes, and no reboot even after some stress tests ive made. It is working well using both xcstasy kernel and stock kernel+tegrak overclock app so i am completely sure anout what im saying.
Your 1300mV is just a little bit too small voltage for my CPU to work stable. As i discovered, differences between voltage settings of your kernel and my device are no longer visible in the range of 0 - 1200mhz, over 1200mhz and especially over 1400mhz my device needs significantly higher voltage setting than your; however it does not result in abnormal heating, thermal throtting, cpu core melting or other natural disasters ;)
In result, your kernel (even non-oced) is booting and is working well for a few minutes (both kernels never rebooted on boot sequence!!), however after launching some advanced tasks - it reboots or freezees because of staying too long in a frequencies which requires more voltage than it receiveso ; (

This truly makes testing hard and also impossible for everyday use, for me at least. As the default voltage values used by you are enough to boot the device, they are still not appropriate when it comes to heavy calculations. Desirable solution: to unlock the voltage table so users can modify voltages themselves if needeed, without modifying the default voltage values, as they are good enough to boot and work without reboot for long enough to let user change the hi-freqs voltages for - probably - every device...

Well is there any easy solution for locked voltages? Or (as i suppose) this feature has to be included in a kernel itself? Is it a lot of work to unlock it?

Good that youve interested in this frandom stuff, as it may be the near-ultimate solution for that crappy lagging issues we all hate... i am euphoric about possible testing, i must admit ;)

Thank you.

Sent from my GT-N8000 using Tapatalk 4
 

kcrudup

Senior Member
Mar 27, 2007
1,517
749
San Francisco Bay Area
... In result, your kernel (even non-oced) is booting and is working well for a few minutes (both kernels never rebooted on boot sequence!), however after launching some advanced tasks - it reboots or freezes because of staying too long in frequencies which requires more voltage than it receives
I don't see how this can happen on my non-OC kernel- it has exactly the same voltages and frequencies as Samsung does for the range 1.4GHz-200MHz, and all I've done is add 100/50/25MHz. If you're getting reboots with my non-OCed kernel, there's gotta be something wrong with your device (I can't think of any other explanation).

Now, if you're using my OCed kernel, remember that Samsung has got code in TW that queries the kernel for its highest possible speed, then explicitly sets your device to use that speed, independent of any apps that might change the speed ("SetCPU", et al.). I couldn't do anything to prevent this (as we don't have TWiz source), which is why I make explicit non-OCed versions of my kernel too.

Desirable solution: to unlock the voltage table so users can modify voltages themselves if needed
Yeah, but I'm probably not going to implement that.

Good that you're interested in this frandom stuff, as it may be the near-ultimate solution for that crappy lagging issues we all hate
I have to thank you for that hint! This has seriously speeded up my tablet noticably! I'll be modifying the kernel's character-device tables to use the "frandom" versions directly from /dev in an upcoming release (mostly so I won't have a possible race-condition that exists now when I flip the /dev entries in user-space in my boot scripts).
 

esgie

Senior Member
Feb 17, 2013
332
179
@kcrudup: does your kernel has exfat support?

I tried the frandom version and I couldn't see my ext sdcard anymore.

I have posted a solution in post #198. BUT - I have a better one, hehe ;) It uses original script from the modification which is not only the mount script, but it also cares for scanning the card, etc etc. and it let you mount ntfs sdcards, too! My single-command workaround posted in #198 was simple and working, but absolutely not stable. And Here's updated tutorial...
Working exFat SD card on kcurdup's kernel
1. Using a computer or stock kernel with exfat support, create a file called
Code:
.mounthere
at the root of exfat extSdCard (i.e. /storage/extSdCard/.mounthere - plz be aware that there is a dot (".") at the beginning of the filename...
2. Edit the file and input only such a text:
Code:
/storage/extSdCard
Save the file and leave it there for a while...
3. Reflash the kcurdup's frandom kernel of your choice. At this time, your exfat card - containing a file created in our previous steps - will disappear from the filesystem. Anyway...
4. You will need a init.d support with the kcurdup's kernel to run some code at boot - well, even if I had init.d support added to the stock kernel, after flashing kcurdup's one it stopped working... I had to add it once again. So you may want to look here:
http://xdaforums.com/showthread.php?t=1933849
for an kernel-independent init.d support. I can say that zip-init method is working well for me... so after flashing with CWM/TWRP or installing with Terminal Emulator, and testing if the init.d work (details in the topic above..) we may proceed to the essential point...
5. Download the newest version of FUSE mount mod from here:
http://xdaforums.com/showthread.php?t=2155363 d
and flash it with CWM/TWRP.

Ur exfat card is working isn't it? ;)

The script in the zip is fully automatic.. It mounts all exfat/ntfs devices for a "test mount", then it reads the .mounthere file from the device (if exist) to find out what is the proper path of mountpoint, it dismounts the test mount and performs a mount again, but to the mountpoint achieved from the created file...
after that, /storage/extSdCard shall be fully accessible, as usual...

Hope I helped.

@kcurdup
Well, please take my apologizes for such a late answer.
1. My device is absolutely ok, I just flashed the same, OC'ed kernel twice by mistake ;) Non-OC version is working absolutely ok.
2. Well, I have found a way to use your OC'ed kernel on my device, but it's tricky and needs a bunch of reflex...
A VERY RISKY AND DANGEROUS solution how to run kcurdup's OCed kernel along with custom voltages
So: before flashing your kernel, I have installed again a Tegrak Overclock Ultimate app (from Google Play, paid version needed as we will need a "start on boot" function... which does not work in free version) - it is an app which let me OC/UV even on stock kernel by using some custom kernel module (however, the changes are not reported "outside" the app, so you may be wondering if it isn't a placebo - but it is not...), and it will let me to change the voltages in your OC'ed kernel too (and this time the changes WILL be visible, lol...). It is no use to configure the app at the moment, as after changing the kernel to OC'ed - which offers different range of frequencies and different number of freq steps than stock - it will not preserve any settings and it even won't let you load a saved profile; I just run it once to give it root privileges, then I ve made a shortcut on the main homescreen - you'll see why it's a brilliant idea in the next step... Then, I do a NAND backup (!!! - important!!!) and after flashing your OC kernel - here comes a tricky part :)
Kcurdup's kernel will cause a system crash of my device (running average, deodexed, near-stock XXCMI2 with some script mods written by myself for my own purposes) in about 30-40 secs after first appearance of the homescreen. There is not much time - I must run the Tegrak (from the shortcut on the homescreen - ha!) and as quickly as possible try to click options in an order below:
a) load overclock module (it should confirm success with vibration);
b) optimization menu;
c)change upper voltages, for i.e.1600 - 1300 mhz, to ones accepted by my device, or at least similar ones - there is no time to be very precise moving the voltage slider, as even a little too low or too high voltages shouldn't cause a crash immediately without a large CPU stress....
If I manage to do it very fast - and it is possible :) - I am achieving enough stability to browse through optimization menu freely, without time pressure, however it is good idea not to wait because low frequencies (25 and 50) are causing crashes too on device of mine, although they are much more rare. So, first I am making a precise voltage setting (to the values I had obtained experimentally in the past) for 1600 - 1300 MHz voltages, so I am now almost-sure that no reboot for high frequency will happen. Then, I do change scaling setting to 100 - 1600, as 25 and 50 MHz also a reboots on my device (they are much less frequent but they WILL happen for sure in not more than in hour from the boot...). Then, I set the rest of 100 - 1200 MHz voltages to my own settings, then I save the profile + tick "start on boot" setting. This is however not the end...
...as the tegrak won't stick to the scaling settings at boot (as the author says: "it may result in a bootloop" ehh...), which may result in not loading the voltages profile at all on the boot or a reboot when using 25-50 MHz (I don't remember which effect it caused as I did above four-five days ago...). And this is where i make use of the second app - this time it's free - called No Frills CPU, but you may want to use any other app that is applicable for setting the standard CPU settings.. The only thing I use it for is to lock the scaling on 100 - 1600 and use the setting on boot - that is why I have chocen No Frills, 'coz it's the smallest one in the market I think.
After that, the device boots up every time (! - for sure. I have booted it about 200 times for the last week, every time with success... ;)
And since then, it's working absolutely stable and with overclocking (and tegrak's app tweaked voltages...) -- and with frandom...
Phew...
WARNING: I am not responsible for any damage of device and/or data caused by the method described above, and especially for the possible data corruptions caused by reboots which will happen if you're too slow.
WARNING 2: DO A NAND BACKUP - after one trial my device never booted again and stuck on "android is updating" screen for ever... no cache/dalvik wiping nor even rom re-flash helped as it was /data partition which was corrupted. There was no solution for similar freezees on the internet (at least I have not found them...), but as it was not the first time for me with such an effect, so I have spent some time (i.e. whole night...) comparing broken and non-broken backups, etc. and finally found the solution - using adb via recovery, or "File Manager" in TWRP recovery, one should delete /data/system/sync directory - the files inside gets corrupt if the crash/reboot happens when the system is accessing them; deleting the folder may result in some problems with data sync or accounts, but it's a way more easy to fix than installing and configuring all your programs again from the beginning... I had backups of previous system-states with /data corrupted and "android is updating [forever...]" problems caused, for example, by using "reboot" command from the terminal (= immediate reboot without closing the memory access = a risk of corruption...) and not closing the system with the menu... including one from the past that came with no backup around, and as I hadn't found solution that time, I was forced to configure whole tablet from the beginning; deleting /data/system/sync directory fixed the issue for every previous case of "android is upgrading" hang which couldn't be fixed by reflashing the /system nor by wiping any caches etc..., so it might be a very important observation for all users that like to play with their filesystems (that's why I am mentioning it here, even if it's an offtop)! Anyway, the best solution is to do a NAND backup before playing with stuff like that and this is the only option which I do recommend.
WARNING 3 If someone will try to repeat above steps, he should first check if maybe the kcurdup's OC kernel works WITHOUT all this crazy stuff, as above solution for OC + frandom refers to my device, which is a very hard partner to cooperate I must say... on the contrary, kcurdup's device does not need any gymnastics for OC to work so please try if it is working out-of-the-box, first...
3. Frandom surely is something. You are right - N8000 is working as it has never before. It seems that there is nothing to discuss, it's just a good piece of code and I can realize no other kernel than yours in my device - others are just LAGGY, no matter what entropy they are using :)

Well, good to hear that I have not wasted your time :) And - it is a pleasure to be an inspiration for you (for the second time - our very first conversation resulted in custom kernel + AllShare Cast solution ;P), as the final results are... enormous :) Well, I love to swallow and then store millions of information in my head - and good that I have found someone who is worth sharing the theoretical conclusions :)


Hope you'll continue developing your kernel as it is really blazing fast with frandom... and too bad it's the only frandom based solution for N8000 around... (well, to be exact, it's - almost - the only solution for N8000 at all...)

My next idea is to port a fiops (v2? or maybe there's further one?) scheduler module for IO operations... it gives me the very very best benchmark results on custom kernels on my N7100. However, it will not result with such a boost in performance as frandom, for sure, so I can't say that it is essential component ;]
 
Last edited:
  • Like
Reactions: sudisk

kcrudup

Senior Member
Mar 27, 2007
1,517
749
San Francisco Bay Area
1. My device is absolutely ok, I just flashed the same OC'ed kernel twice by mistake ;) Non-OC version is working absolutely ok.
OK, good to know- if the non-OCed version had still looped on you I'd've been worried.

c) change upper voltages ... to the values I had obtained experimentally in the past for [the] 1600 - 1300 MHz voltages, so I am now almost-sure that no reboot for high frequency will happen.
If you can give me your 1.3 GHz+ voltages, I'll look into it. FWIW, the older voltages (before I'd normalized them) were:

Code:
+        /*   ASV0,    ASV1,    ASV2,    ASV3,    ASV4,    ASV5,    ASV6,    ASV7,    ASV8,    ASV9,   ASV10,   ASV11 */
+        { 1450000, 1425000, 1425000, 1400000, 1400000, 1400000, 1400000, 1375000, 1350000, 1350000, 1350000, 1325000 }, /* 1.7 Ghz */
+        { 1325000, 1312500, 1300000, 1287500, 1300000, 1287500, 1275000, 1250000, 1250000, 1237500, 1225000, 1212500 }, /* 1.6 GHz */
+        { 1325000, 1312500, 1300000, 1287500, 1300000, 1287500, 1275000, 1250000, 1250000, 1237500, 1225000, 1212500 }, /* 1.5 GHz */
+        { 1325000, 1312500, 1300000, 1287500, 1300000, 1287500, 1275000, 1250000, 1250000, 1237500, 1225000, 1212500 }, /* 1.4 GHz */
+        { 1300000, 1275000, 1237500, 1237500, 1250000, 1250000, 1237500, 1212500, 1200000, 1200000, 1187500, 1175000 }, /* 1.3 GHz */
<snip>
and finally found the solution [for certain types of hung boots]: one should delete /data/system/sync directory
Thanks for this- I've had situations where I've also been hung up at the end of the boot sequence and have had to restore the entire /data directory.

3. Frandom surely is something. You are right - N8000 is working as it has never before.
I am really surprised how well it's working, especially since there's a lot of discussion as to whether any Entropy fixes are/n't placebo. What's nice about a kernel-based solution like "frandom" is it doesn't affect the system like the old "seeder" apps do.

I've also modified some of the kernel tunables using guidelines used in the "Supercharger V6" thread (w/o all the intrusive scripting used by that method, and ignoring some of the Low-Memory Killer changes and Dalvik VM Heap-size mods). I've been running it for a couple of days and seem to like it- I'll be pushing a kernel for that for people to try out soon.

I've also patched a potential kernel exploit listed here that'll be in my next kernel, too.

My next idea is to port a fiops (v2? or maybe there's further one?) scheduler module for IO operations
Do you have a site with source for it?
 

bucklee

Senior Member
Apr 16, 2011
183
19
Brooklyn
Hi Krudcup, was trying to use your kernels and my microSD (64GB) keeps getting ejected. Also all the Language is in Polish, this is probably something to do with ARHD 9.0 I am running. I'm also running 8013. Any ideas?

Now I'm just stuck in the samsung Logo.....:-\
 
Last edited:
  • Like
Reactions: sudisk

sudisk

Senior Member
Feb 7, 2012
823
311
Paris
I thanked your post by mistake.

The answer to your first question is in post #206.

For the second one, are you using super su v1.69? There is a bug in that version.
 

esgie

Senior Member
Feb 17, 2013
332
179
Thanks for this- I've had situations where I've also been hung up at the end of the boot sequence and have had to restore the entire /data directory.
Well, the problem is when you are getting this kind of error in 5 minutes before planned backup (after a month without doing it...) in a almost-perfectly-configured-device with 500+ apps installed :D
However, I hereby confirm again that this method IS WORKING and it let us to recover from corruption @ data partition...
If you can give me your 1.3 GHz+ voltages, I'll look into it.
Below is my profile (still working stable on your kernel, after dealing with the "change your voltages in 20 secs after boot or die" method!)
Code:
frequency=1600000,1500000,1400000,1300000,1200000,1100000,1000000,900000,800000,700000,600000,500000,400000,300000,200000,100000,50000,25000
core_voltage=1375,1325,1262,1212,1175,1125,1050,987,975,962,937,937,925,900,875,862,862,850
scaling_governor=pegasusq
scaling_min_freq=100000
scaling_max_freq=1600000
Please ignore two last settings (for 50k and 25k MHz) because those frequencies causes a reboot on my device after some time and it is no difference what voltage I have set for them.
Voltages for 1600 - 100 MHz had been measured experimentally (after a long night with 100+ crashes of the system hehe) long time ago. They shall be considered as an maximum undervoltage settings - every single setting will cause an unstability on my device if lowered, even a little bit (by "unstability" I mean crash when running 10 minute-full-stress-test, even if it happened as late as after a few mins... as you can see, I am well prepared :D )

As to the fiops scheduler:
https://github.com/DerTeufel/android_kernel_samsung_smdk4412/blob/dualboot/block/fiops-iosched.c
This is a repo of the kernel for N7100 (Note 2) which I am using at the moment and which offers a great performance, and a lot of features.
(incl. dual boot CyanogenMod up to 4.3.1 with Stock Samsung ROM), and as the dev is a very active person I am convinced that he has integrated all the essential patches/code updates until today...
I think that the CPU governor "devilq" (source should be available at the same git I suppose) is also something worth looking.
@bucklee
That's right, see the very beginning of my too-long post #206 for extFat solution
The second answer given to you by user sudisk is good too - today I woke up with greek language turned on, for example... It is not kernel related afaik. We should wait for an update...
 

kcrudup

Senior Member
Mar 27, 2007
1,517
749
San Francisco Bay Area
Also all the Language is in Polish

today I woke up with Greek language turned on, for example- it is not kernel related AFAIK
Well, now I'm not so sure- I boot up now in German, which I thought was the result of some corruption due to the "Supercharger" mods I'm testing reverting me back to the original language in the DBT (German) ROM I'm running (BTW, the "/data/system/sync" boot-hang fix was a great find- many thanks for that, it saved my butt here, too).

Now that I've seen two other people have the same problem, it's gotta be the kernel. Maybe it's a race-condition when I switch over the Frandom nodes, or a LANG issue with the version of Busybox I've included.

Can y'all do me a favor, and send me the output of "cat /proc/version" as well as the link you'd used to download the boot.img (if you remember)?
 

esgie

Senior Member
Feb 17, 2013
332
179
Now that I've seen two other people have the same problem, it's gotta be the kernel. Maybe it's a race-condition when I switch over the Frandom nodes, or a LANG issue with the version of Busybox I've included.
Uh u misunderstood us - it's a SuperSU 1.69 issue, and its not kernel or even device related..
Nothing to worry, then. Hope u didnt spent all night looking for a non existent bug ;]
 

kcrudup

Senior Member
Mar 27, 2007
1,517
749
San Francisco Bay Area
I've got an update that I like enough to publish. There's not many changes, though- it's still got the "frandom" stuff in it, plus I've added in some suggested changes from the Supercharger V6 thread (mostly Virtual Memory and Network Buffer changes, I didn't touch the Low-Memory Killer settings as Samsung's defaults actually work pretty well for us, as we have 2GB RAM). I've also added in preliminary support for the "fiops" scheduler that @esgie had asked for, but it's not complete yet (there's some fair-scheduling code that needs to also be cherry-picked in, but I have to update some parts of the I/O Block Layer to make those work) and I'll add it in later.

I also have to add support for the OC voltages @esgie wanted, but those will come later as well.

Here's the OC version.
Here's the non-OC version.

Development is likely to slow down, too- the 3.0.x kernel that our devices use has now been marked "End of Life" at version "101" by Linus and crew, so unless I see something compelling from upstream (which does happen) or a security issue comes up, I probably won't have anything new for a while.
 
  • Like
Reactions: GDT and m.wojcik

esgie

Senior Member
Feb 17, 2013
332
179
Great job!
However: guess I'll personally wait for finished fiops / stable voltages version, as I use Supercharger (along with zeppelinrox's network tweaking scripts) for everyday use, so those tweaks will be not a big change for me. Of course, it would be nice to get rid out of those scripting mess Supercharger makes in system, but not in the cost of repeating the nightmarish procedure of setting my voltages in first 20 secs after boot, which took 6 tries before I did it succesfully = 5 crashes of unsuccesful trials = 5 chances of messing up my data last time ;]
As you seem to use - or at least analyze - the performance scripts messing around, maybe you'll look into CrossBreeder (look for it on XDA, very easy to find...)? It's a thing that implements another method of generating entropy (worse than frandom BUT the best one which can be used without additional kernel module - at least this is what they write...), BUT also some dns caching + light ad blocking stuff which seems to be worth looking (however, as it's working on system layer it's just for your knowledge; i don't think it's a good idea to implement those in kernel or something...).
And as to the developing slowdown - it's okay, as the elements essential for performance as well as stability - which are the most important thing to final user - seems to be on place ;] well, I think we're all waiting for 4.3 stock system, so I agree that it is pointless to spend whole free time on upgrading something which will be depreciated in the nearest (hopefully, Samsung!!!!) future...
 

GDT

Senior Member
Mar 22, 2007
1,043
72
Bologna
I've got an update that I like enough to publish. There's not many changes, though- it's still got the "frandom" stuff in it, plus I've added in some suggested changes from the Supercharger V6 thread (mostly Virtual Memory and Network Buffer changes, I didn't touch the Low-Memory Killer settings as Samsung's defaults actually work pretty well for us, as we have 2GB RAM). I've also added in preliminary support for the "fiops" scheduler that @esgie had asked for, but it's not complete yet (there's some fair-scheduling code that needs to also be cherry-picked in, but I have to update some parts of the I/O Block Layer to make those work) and I'll add it in later.

I also have to add support for the OC voltages @esgie wanted, but those will come later as well.

Here's the OC version.
Here's the non-OC version.

Development is likely to slow down, too- the 3.0.x kernel that our devices use has now been marked "End of Life" at version "101" by Linus and crew, so unless I see something compelling from upstream (which does happen) or a security issue comes up, I probably won't have anything new for a while.
This release doesn't work for me. Hangs my note at startup. I'm speaking of OC version. The previous one was ok.
 

esgie

Senior Member
Feb 17, 2013
332
179
Works fine for me
Re-download and try again

Sent from my GT-N8000 using XDA Premium HD app

If youve been reading carefully you'd know that the oc version works on some devices only, including the kernel dev's one.
For example: it is working long enough for my device (about 20-30 seconds after boot) to set different voltages by tegrak overclock app (and only this app seems to be capable of changing the voltages, as it uses its own module; voltages are blocked by default so no other app can handle it). For other's device the voltages may be too low at all to even boot. For others they seem to fit perfectly.
 
  • Like
Reactions: aoryx

GDT

Senior Member
Mar 22, 2007
1,043
72
Bologna
I'll be more precise. With old releases I can use the kernel @Stock settings voltages without issues. Now it locks on startup.
 

Top Liked Posts

  • There are no posts matching your filters.
  • 38
    Here is a kernel build for the Samsung Galaxy Note 10.1.
    Running on 1704Mhz. With extra 100, 50 and 25Mhz step. So it runs 25 to 1704 instead of 200 to 1400.
    Custom bootanimation support. (Standard bootanimation wont work)
    Triangle away support.

    Update N8000BRV2 (ICS)
    -better voltage settings
    -better wifi drivers
    -BFQ scheduler back
    -updated linux version
    -other minor tweaks

    Update N8000Jelly_OC
    -Build for JellyBean

    edit.
    N8013 and N8010 support pulled back because of the wifi problems

    source https://github.com/brieuwers/N8000Kernel

    Thanks to Gokhanmoral, Espenfjo, Entropy512 and Franciscofranco.


    Flash tar with odin or zip with CWM.

    Everything at your own risk.

    Greetz Bas.
    6
    Thanks!
    Yes, and no... I do get the occasional boot-loop situation which after a few loops usually boots and allows me to power off normally so that the device can cool down. If I do not power off and let it cool down before restarting, after having had bootloops, it will be unstable and reboot into a new cycle of bootloops.

    However, it seems as if the CPU frequency both for min. and max. is set to 1.7 GHz at boot, and remains that way.
    If I get a successful cold-boot, I use System-Tuner Pro to change min. to 100 MHz and keep max. at 1.7 GHz. I do not think I have had a reboot/bootloop situation with the changes settings from cold-boot.

    I forgot to set the minimum frequency.
    Here's a beta2 for you for testing with other voltage settings.
    Maybe this one will work fine.
    Let me now if it works.
    4
    N8013 kernel loaded. No problems installing. SPen good. Using Nova launcher - extremely snappy. Did not change governor (default is pegasusq). Cleared background tasks before each run. Benchmarks so far:
    Linpack - average 190
    Quadrant - average 6100
    RL Bench (SQL) - average 18.2
    Antutu - average 13900
    Wifi - when out of range of 5Ghz, automatically switched to 2.4Ghz. No problems with hitting top speeds on either. Signal range for 5Ghz seems a little shorter than before, but not drastically.
    Seems stable. Wouldn't be surprised if this can be OC'd higher.
    Will use throughout the day and report any problems.
    4
    Thanks, and yes, the first kernels worked fine.

    Could you try this kernel for me and report back if it still bootloops or not?
    3
    XDA seems to have some problems right now.
    Editing and uploading new kernel wont work.
    I will try again later.

    Sent from my GT-N8000 using XDA Premium HD app