I’ll do it in some spare time. But I’m not sure what would it tell me beyond what I can see even without it. Even just Battery Buddy logs or readings from /sys/class/power_supply/battery/current_now show that the XA2 Ultra simply consumes much less energy in virtually all conditions. In sleep it manages to go as low as 11-12 mA (some 10-20% of time) averaging at 19-25 mA (70-80% of time) while the 10 III never gets below 36-40 mA and averages at 45-55 mA. On the XA2 Ultra, when it goes to sleep there are occasional spikes up to 48-62 mA (and literally a few higher but still below 100 mA), whereas on the 10 III spikes are almost always in the 120-250 mA range, and some of them even reach 300-600 mA. I’m afraid that powertop won’t show anything new in this regard.
That’s really difficult to detect, unless someone starts stopping (maybe removing) services until to find the culprit. It might not be a SFOS and the issue is in lower level binaries?
Maybe powertop can reveal the processes which consume more power and wakeup the cpus this often. Maybe there is something misconfigured in the kernel settings.
It certainly isn’t the stock Android because all 10 III reviews give it much much better battery life and talk times. It might be the AOSP, but I haven’t seen any data. With not even any LineageOS release for the 10 III so far, there isn’t much to compare…
I’ll certainly check it. But since the 10 III is now my daily driver, I’ll need to find proper time for it when I can leave it for some hours intact to collect the data.
Powertop seems to not work correctly uner SFOS when the display is off. I couldn’t generate a long time report using ssh. But for comparison just put the both phones in flight mode, close all apps and change the display settings to turn it off after 2 minutes. Start the terminal and generate a 30 second report:
sudo powertop --iteration=1 --time=30 --html=POWERREPORT.HTML
Or enable wifi and execute the command via ssh.
30 seconds should be enough for a comparison.
I have a “spare” XA2 Ultra I could try to compare with. It’s not at 4.4.0.72 but should still work for comparison purposes.
Aight, since the forum doesn’t let me upload a .zip file, let alone .html files, here they are:
- SUSE Paste for XA2 Ultra
- SUSE Paste for X10 III
It looks like X10III is a bit more efficient, but then again, the Gold Cores seem to spend their time…somewhere. I don’t have the knowledge to dissect those, so go wild!
Thank you,
only thing i noticed, that the x10iii is always using the highest clock speed of the used core. The XA2 uses only the lower frequencies.
This seems to be caused by schedutil governor. With ondemand it is also using the lower frequencies.
Sadly all this does not seem to help to find the cause of the battery drain.
OK, and here’s mine:
On the 10 III the infamous voicecall-ui -prestart seems to be eating CPU while actually doing nothing.
On a different note… Just out of curiosity, I’ve just switched all cores of the XA2 Ultra to “performance” governor, so now four cores run all the time at 2.2 GHz and the remaining four at 1.8 GHz. It made the XA2 noticeably snappier, almost as smooth as the 10 III. Power consumption has increased, but still within some 80-110 mA range, i.e. actually still lower than the 10 III… It’s been 20 minutes now and it hasn’t even lost a single % of charge.
Your XA2Ultra has a way lower wakeup count than your X10iii. Even if we subtract the wakeups caused by voicecall-ui-prestart, it is significantly higher than your XA2ultra.
Yeah, I’ve noticed it. But the question is why. Both devices are actually mirrors - same data, same software installed, and the same (i.e. nothing) running while taking the measurement.
Maybe I’ll give it a reboot and try to take the reading once more on the 10 III.
OK, here’s a new one, taken right after reboot and without activating the SIM card (it wasn’t active on the XA2 Ultra, either). Slightly less wakeups, still more than XA2 but I guess comparable.
P.S. @direc85, your 10 III still has much less wakeups. What state was it in? WLAN, cellular radio enabled or disabled? Thank you.
OK, last one, after yet another reboot, everything (WLAN, cellular radio, GPS, Bluetooth, etc) disabled:
Wakeups dropped to 192/s.
Thank you so much! The
mcetool --set-low-power-mode=enabled
command makes such a difference. Wow!
I admire all who try to improve the battery consumption!
Out of curiosity, I’ve just installed White on Black ambience (which I further tweaked to be even more monochromatic). Sadly, there is no noticeable reduction in power consumption because only the main launcher screen is really black (pixels are really turned off). Both the Events screen and backgrounds of all applications are actually grey - it is clearly visible that those grey pixels are active and backlighted. This grey color is caused by the Silica UI automatically applying some kind of effect over the black background, which in this case makes it grey-ish. So, without a deeper modification (removing that effect) it is not possible to get any noticeable power saving by using a black theme.
Nice to hear. I checked with mcetool |grep power, that after the update to 4.4.0.72 low power mode is still enabled
I’ve created myself a black background image at the resolution of the phone screen, and I also applied a script from x10ii-x10iii-color-banding-in-low-light-conditions which fools the system and gives me true black colour. Still tho many apps, especially the browser, have bright or white colours which does not help much.
In any case power consumption on the screen is not the main issue, the phone keeps consuming lots of power while idling, in my case it’s about 1% per hour if not more.
I think the OLED screen is not used in the way the Nokia N9 did it. It looks like the whole screen is always powered on and not only the pixel which are used and not black.
Regarding the general power consumption, maybe the open source drivers do not contain all the power saving mechanics than the Android ones.
I very much doubt it. Set the screen brightness to maximum and then slowly swipe from the home screen to the Events screen and notice how it switches from real black to gray. Same way, compare the background of the home screen (where the covers of minimized apps are shown) or the lock screen (which are the only screens where the blur effect isn’t applied to the background, so a black background stays really black and on OLED displays causes that pixels are really OFF) with backgrounds of applications which seem to have black background - if you have max brightness set, you will notice that their background is actually nowhere near as black as on the home screen and the pixels are all active and emitting some light.
Only the Lock screen and the home screen show real black color and allow the pixels to really switch off. Everywhere else, the blur effect automatically applied to the background by the Silica UI, causes that the background is no longer black (let’s say its color is no longer rgb #000000 but something like #010101 i,e, very dark grey) which causes that the whole screen remains powered on as only full black (#000000) powers off the pixel. So making SFOS OLED friendly would require disabling that blur effect.
How this blur gets automatically applied is clearly visible when one very slowly swipes between Home and Events screen - on colorful themes you can see how the background gets blurred in between, and on a black theme (if you set display brightness to max in a dark room) you can see how real blackness turns into grayish and pixels start emitting light.
we might get slightly off topic, but there is no colour banding issue with XZ3 (community port) which is using similar screen OLED technology.
This is not connected with the color banding issue of the 10 III display, it is about the blur filter that SFOS applies to all backgrounds of all screens and applications (except for home and lock/LPM screen), whose side effect is that black is no longer entirely black and thus on all screens (but home and lock/LPM screen) pixels remain powered on. In the last paragraph of my previous post I described how to check this.
