Originally Posted by pulser_g2
The issue is that these performance gains seem to come from holding the CPU in a higher state- I think that the actual improvement is a byproduct of the means of applying the fix - the fix itself shouldn't give this performance boost.
EVERYTHING I've read is indicative of it being due to CPU scaling - someone said their Note (using Exynos 4210) doesn't have home button lag on turning the screen on - this happens when the phone isn't sleeping fully. I therefore suggest that if you want to test this properly, you should also do a test with a locked CPU clock, and see if it's faster...
Most likely it will be identical, and you can decide between forcing up your clock speed (and having 10 minutes of battery life), or using the regular clock-speed governors and accepting a little bit of lag when ramping up.
Now we're getting somewhere. I reran my tests and can conclude the following.
Some apps were faster with the sysctl fix and the CPU locked. Maps was a big one and connecting to certain sites.
Most of the apps I had seen improvement on showed the same or even better improvements with the sysctl fix out of the way and the CPU locked.
That means, at least on my hardware, that the sysctl fix works on some things, but only a small fraction of what i was originally claiming. The rest seems to be because the CPU is kept awake and in a higher state.
This could also be why some people aren't seeing improvements like others. It could have a lot to do with how your kernel governs CPU behavior.
Sent from my Nexus 7 using Tapatalk HD