Alot of people have been posting in these forums, "Which kernel is the best"?
I was intrigued and decided to work this out for myself and share it with the rest of the community.
This is not meant to hurt the makers of these kernels.
But FIRST, it is not a fact that there is a 'best' kernel. 'Best' is an opinion and everyone has a different opinion.
On the Nexus 7, I personally believe there is no 'bad' kernel and none of these kernels I have tested perform badly or lag heaps.
People still like to push their N7's as hard as they can though.
The Main 2 Features people look for in a kernel are: Performance & Battery Life
Some people prefer performance over Battery, or vice versa, or an all-round kernel.
I will test another kernel if someone requests me too. I will not post other's results because they most likely have a different setup and therefore different results making it inaccurate.
Nice! This should be a sticky to help reduce unnecessary threads about kernels. My only suggestion would be to run Quadrant 3 times and take the average number. Only because 3 runs will have different results every time
I am testing several kernels here. The fastest for me so far is trinity, but the battery is draining really fast. I then tried the 1400 MHz trinity and things got a little better with a good performance in quadrant.
I am testing motley now and seems good but a little slower than trinity. Home screen doesn't look as buttery as in trinity. I will keep testing a little longer.
Sent from my Nexus 7 using xda premium
---------- Post added at 06:19 PM ---------- Previous post was at 05:54 PM ----------
Crap...i just had a reboot while using motley and started playing dead trigger. I returned to trinity 111 and it is definitely smoother than the others for me. Maybe it is because it is the first one i flashed and now when i try others, even wiping, something remains and the other kernels don't work right. Who knows....
The best way to test CPU performance would be to make sure that all kernels are at the same exact core frequency max, and all cores enabled, on performance governor. Performance governor alone doesn't enable all cores, and benchmarks sometimes don't trigger all 4 cores as on either (it can, but not always, and that can lead to totally different numbers on different kernels).
Some kernels have higher max frequencies then others, some use the new CPU quiet driver and some don't (although this wouldn't really matter if all cores are forced on for testing).
Some kernels have higher GPU clocks then others, some have higher refresh rates then others, some have both higher refresh rates and clocks, etc.
Personally, the best way to judge a kernel is by what features are implemented in it for example, is it linaro compiled? Does it use the old hotplug method for CPU cores, or the new quiet driver? Updated WiFi driver? fsync toggle? Was it built with RNDIS drivers?
Quadrant isn't a stable means of benchmarking (run it 3 times in a row and watch those numbers change), but generally speaking, if the same kernel is compiled with linaro, vs no linaro, the linaro build would be faster (don't quote me on this). Not all those kernels are linaro that I'm aware of.
Also, Jay's buttery smooth tweaks, his fsync doesn't work on franco.Kernel. That kernel specifically uses a different command to toggle fsync, and unless that method was used, any I/O results would be void (basically comparing fsync kernels to non-fsync, where fsync would obviously dominate). And there is even the possibility the ROM you were using may interfere with CPU clocks and voltages unknowingly.
So with all that said, trying to find the "best" kernel, is going to take a bit more effort then just benching with Quadrant
And for my own personal opinion (no hard-testing done) I like franco.Kernel Stock clocks and voltage, fsync toggle, new CPU quiet driver and WiFi driver, works awesomely faux's kernel would be my second preferred kernel, only because lack of fsync control, and getting rid of the CPU quiet driver.
The best way to test CPU performance would be to make sure that all kernels are at the same exact core frequency max, and all cores enabled, on performance governor. Performance governor alone doesn't enable all cores, and benchmarks sometimes don't trigger all 4 cores as on either (it can, but not always, and that can lead to totally different numbers on different kernels).
Some kernels have higher max frequencies then others, some use the new CPU quiet driver and some don't (although this wouldn't really matter if all cores are forced on for testing).
Some kernels have higher GPU clocks then others, some have higher refresh rates then others, some have both higher refresh rated,
Personally, the best way to judge a kernel is by what features are implemented in it for example, is it linaro compiled? Does it use the old hotplug method for CPU cores, or the new quiet driver? Updated WiFi driver? fsync toggle? Was it built with RNDIS drivers?
Quadrant isn't a stable means of benchmarking (run it 3 times in a row and watch those numbers change), but generally speaking, if the same kernel is compiled with linaro, vs no linaro, the linaro build would be faster (don't quote me on this). Not all those kernels are linaro that I'm aware of.
Also, Jay's buttery smooth tweaks, his fsync doesn't work on franco.Kernel. That kernel specifically uses a different command to toggle fsync, and unless that method was used, any I/O results would be void (basically comparing fsync kernels to non-fsync, where fsync would obviously dominate). And there is even the possibility the ROM you were using may interfere with CPU clocks and voltages unknowingly.
So with all that said, trying to find the "best" kernel, is going to take a bit more effort then just benching with Quadrant
And for my own personal opinion (no hard-testing done) I like franco.Kernel Stock clocks and voltage, fsync toggle, new CPU quiet driver and WiFi driver, works awesomely faux's kernel would be my second preferred kernel, only because lack of fsync control, and getting rid of the CPU quiet driver.
Yeah. I know. I'm going to add a other features section. If someone else could do it that would be good. If not, I'll do it later.
I was going to do a HD video playback test. See how long it lasts on each.
Sent from my Nexus 7 using xda premium
Who JUST watches video... would be as effective as leaving it on a white screen for hours. Like I said there is no point and would just give false hope.
*Cue Dramatic Music*Put on your red and blue latex overalls and cape because it’s time to … more
XDA Developers was founded by developers, for developers. It is now a valuable resource for people who want to make the most of their mobile devices, from customizing the look and feel to adding new functionality. Are you a developer?
Introducing XDA:DevCon – A Conference For Developers By Developers
>
[GUIDE] *11/3 UPDATE* 4.2.2 RESULTS! N7 'Best' Kernel? (Performance & Battery Tests)
Print
Email
[GUIDE] *11/3 UPDATE* 4.2.2 RESULTS! N7 'Best' Kernel? (Performance & Battery Tests)
View Profile
View Forum Posts
Follow on Twitter
View Facebook page
Follow on Google+
View GitHub profile
Visit Homepage
Linear Mode
