Yes, i have the same opinion.
Second graph in this thread by Ezekeel, shows the difference between lower and higher frequency steps' battery consumption with and without LPA/deep idle. As it occurs, higher frequency supports race-to-idle better there
Again, "100 mhz bad compared to 200 mhz" depends on we're using lpa or not. (sched_mc=2 or 3).
Anyhow, since only first core can hit lpa, screen-off profiles with: different governor, hard limiting frequency, switching to single core mode, etc actually does a little harm than good by delaying for race-to-idle.
SmartassV2 does not have a screen-off profile. (Smartass has). It has only a "sleep_ideal_freq" to which the governor ramp down CPU aggressively when screen is turned off.
If some task demands more power during screen off, the frequency can go all the way upto maximum frequency depending on the load/demand.
Lulzactive has screen-off profile and the max-freq for scree-off state is configurable as well.
Yup - thanks to stuff like idle states and multiple cores, many of the old "rules of thumb" have become FAR more complicated.
In my opinion - hard-limiting to 200 is bad. I hard-limit to 500 as a safety measure - voltage increases a decent amount at 800, and if something drives more than a blip at 800, something is wrong.
Deciding when to turn on/off the second core is much tougher - having it on gets you "done quicker" in theory - but turning it on takes time, and some tasks run less efficiently when spread across cores. Also it may be better to hit 800 MHz than to run both cores at 500.
I wanted to post feedback for geko: raising all your UV settings with 50mV has kept my device all day running without a single problem. Maybe I will try to go down 25 mV again to get even closer to your setup. To bad my device can't handle your config, but like this it's close
|cpu, gpu, neak, ocuv, redpill, siyah kernel config settings tweak, voltage|
|Thread Tools||Search this Thread|