Questions for Jolla Tablet owners (about battery and CPU usage)

Question

21

Questions for Jolla Tablet owners (about battery and CPU usage)

asked 2016-02-24 13:22:41 +0200

Moshroom

1563 ●19 ●24 ●26

http://www.univelka.net/n...

updated 2016-02-24 16:06:34 +0200

As I've stated in another thread, I managed to get an unofficial Tablet (sorry about that). It looks genuine, but... What I'd like to do is to gather some user experience mainly from people who have the real one. The main reason is to find out whether my device behaves like genuine so that I'll know whether it's relevant for me to answer Tablet related questions in this forum and develop workarounds ;). So far it looks like the answer is yes (at least regarding the brightness issues). But we'll see.

What number do you have in Settings/About Product/Hardware Adaptation?
How often do you have to charge?
How long is your battery life in idle?
If you are using System Monitor (https://openrepos.net/content/abranson/system-monitor-tablet-edition), what percent of CPU usage you'll get when the device is on idle?

(And my answers:

0.9.0.9
Every other day
48 h
76 %, but if the charger is attached the number drops to approximately 1 %)

edit retag flag offensive close delete

Comments

1

Idle battery life is only 48 hours? Really? My now over four year old Android tablet can lay idle for several weeks.

nthn ( 2016-02-24 16:12:01 +0200 )edit

3

@nthn Well, that 48h is probably with wlan connected and I can assure that it won't do any better with Android.

tigeli ( 2016-02-24 18:34:31 +0200 )edit

1

I am more conserned over the 76% CPU usage while idle. There's something happening there.

jollailija ( 2016-02-24 18:45:10 +0200 )edit

@tigeli, disconnecting wlan doesn't affect the battery life nor CPU usage while idle.

@jollailija, yes that's my assumption too. If someone with an official tablet will answer, we'll hopefully find out whether this is an error in Tablet version of Sailfish OS (or a misunderstood feature of the System Monitor app) or something that shouldn't be there in the first place (which would make this just my personal problem to be handled outside this forum).

Moshroom ( 2016-02-24 18:55:16 +0200 )edit

1

@Moshroom Well disabling wlan does affect on my tablets battery life. Can't tell about System Monitor app though..

edit: Just checked that my tablet's cpu is 97% idle when running top in terminal (without charger being connected).

tigeli ( 2016-02-24 19:02:08 +0200 )edit

1

@tigeli, to be absolutely precise, I can see some change in battery life too, but so little that it wouldn't affect my answer to the question two. Do you have an official Jolla Tablet (meaning that you are an IGG backer who received your perk from Jolla)? If so could you please tell me the number you see in Settings/About Product/Hardware Adaptaion?

Moshroom ( 2016-02-24 22:45:58 +0200 )edit

1

@Moshroom Yeah, I have one.. "Device adaptation: 0.9.0.9".

(And if you have anything higher than that you are not running the latest software (2.0.1.7) available but something else..)

tigeli ( 2016-02-24 23:01:14 +0200 )edit

1

@Moshroom Lol :D tigeli is a Jolla employee (I hope he is still on board). :)

Alex ( 2016-02-24 23:06:32 +0200 )edit

@tigeli, thanks. So probably my Tablet is as genuine as it can get without being an official one. Therefore it seems likely that either there is a problem in the System Monitor app or that the tablet version of the Sailfish OS runs some unneccessary battery-eating processes during idle...

Moshroom ( 2016-02-24 23:13:06 +0200 )edit

5

@Alex Yeah, the correct tense is 'was' as currently I'm working with some other stuff but you never know what future brings.

tigeli ( 2016-02-24 23:15:50 +0200 )edit

@Alex, oh I didn't know that. Well, I guess that's the other possible way to have an official Jolla Tablet.

Moshroom ( 2016-02-24 23:19:42 +0200 )edit

2

@Moshroom 17 hours idle and battery has depleted from 100% to 92% (without wlan being connected). With the current pace the battery would last about 8 days..

tigeli ( 2016-02-25 16:57:29 +0200 )edit

I set my device to flight mode, turned off the Android support and rebooted. An hour after that the battery shows 98,64 %. So this way it'll last three days (not quite eight, but 33 % more than two). Hovewer the battery drain curve might not be linear so better leave it alone for now and check again tomorrow morning.

Moshroom ( 2016-02-25 20:39:00 +0200 )edit

1

@tigeli: WLAN is connected, it notifies me of new emails right as they arrive. I don't know how it works, but it does.

nthn ( 2016-02-25 23:02:06 +0200 )edit

2

@tigeli, I managed to get some improvement with flight mode on, Android support (and System Monitor!) off and all notifications off. It dropped 4 units in 8 hours (from 67 %* at 11 pm to 63 % at 7 am). This way it'd last about 8 days, so same as you.

= in active use before going to sleep it dropped about 30 % in 2 hours

Moshroom ( 2016-02-26 07:14:22 +0200 )edit

3

Can only comment that the battery life is laughable. With idle tablet it's less than two days for a dead battery, have to keep it in charger all the time. Haven't dug out what's happening, good comments in here so have to take a look. Current state is a joke anyway :)

jsievikorte ( 2016-02-29 05:45:10 +0200 )edit

In this thread linked below: http://askubuntu.com/questions/33640/kworker-what-is-it-and-why-is-it-hogging-so-much-cpu?answertab=votes#tab-top

There are instructions on how to use perf, which can tell you what kworker is doing with those CPU cycles. On Sailfish 'perf' is a separate package of its own, but I had to install 'less' to view the report, and that doesn't seem to be available by default in the i486 repo so I had to add the mer-tools one for that. My results seem to point to a component of the wireless driver, but more reports would only help.

Andy Branson ( 2016-04-12 01:13:05 +0200 )edit

2

The tablet cannot reach deep-sleep mode, thats why the battery drains to fast. It is always over 1% per hour. Jolla please investigate it and help fixing this!. I would say it is a software or firmware problem.

You can install "cpu spy" from fdroid (needs android) to see cpu states graphicaly, or use command from terminal (does not need android, maybe need devel-su):

cat /sys/devices/system/cpu/cpu0/cpuidle/state?/usage

The "?" stands for [0-4], the cpu has 5 states, But is only working in "state0" (POLL). Thats why its draining.

Frequency:

cat /sys/devices/system/cpu/cpu?/cpufreq/cpuinfo_cur_freq

I did "mcetool -F enable" on console (force powersave), but it does not help very much. In the journal I can find entries that maybe the cpu should go to deeper sleep states, but something always wakes it up.

Apr 30 09:21:49 Jolla kernel: Disabling non-boot CPUs ...
Apr 30 09:21:49 Jolla kernel: smpboot: CPU 1 is now offline
Apr 30 09:21:49 Jolla kernel: smpboot: CPU 2 is now offline
Apr 30 09:21:49 Jolla kernel: smpboot: CPU 3 is now offline
Apr 30 09:21:49 Jolla kernel: Enabling non-boot CPUs ...
Apr 30 09:21:49 Jolla kernel: smpboot: Booting Node 0 Processor 1 APIC 0x2
Apr 30 09:21:49 Jolla kernel: Skipped synchronization checks as TSC is reliable.
Apr 30 09:21:49 Jolla kernel: CPU1 is up
Apr 30 09:21:49 Jolla kernel: smpboot: Booting Node 0 Processor 2 APIC 0x4
Apr 30 09:21:49 Jolla kernel: Skipped synchronization checks as TSC is reliable.
Apr 30 09:21:49 Jolla kernel: CPU2 is up
Apr 30 09:21:49 Jolla kernel: smpboot: Booting Node 0 Processor 3 APIC 0x6
Apr 30 09:21:49 Jolla kernel: Skipped synchronization checks as TSC is reliable.
Apr 30 09:21:49 Jolla kernel: CPU3 is up

And I think this does not look good, too:

Jolla kernel: acpi PNP0A08:00: ACPI _OSC support notification failed, disabling PCIe ASPM
Jolla kernel: acpi PNP0A08:00: Unable to request _OSC control (_OSC support mask: 0x08)

Next I will try "https://openrepos.net/content/rinigus/system-monitor-unofficial-edition" it can show failed suspend calls.

jota ( 2016-05-02 22:22:25 +0200 )edit

Idle battery life is only 48 hours? Really? My now over four year old Android tablet can lay idle for several weeks.
nthn ( 2016-02-24 16:12:01 +0200 )edit
@nthn Well, that 48h is probably with wlan connected and I can assure that it won't do any better with Android.
tigeli ( 2016-02-24 18:34:31 +0200 )edit
I am more conserned over the 76% CPU usage while idle. There's something happening there.
jollailija ( 2016-02-24 18:45:10 +0200 )edit
@tigeli, disconnecting wlan doesn't affect the battery life nor CPU usage while idle.
@jollailija, yes that's my assumption too. If someone with an official tablet will answer, we'll hopefully find out whether this is an error in Tablet version of Sailfish OS (or a misunderstood feature of the System Monitor app) or something that shouldn't be there in the first place (which would make this just my personal problem to be handled outside this forum).
Moshroom ( 2016-02-24 18:55:16 +0200 )edit
@Moshroom Well disabling wlan does affect on my tablets battery life. Can't tell about System Monitor app though..
edit: Just checked that my tablet's cpu is 97% idle when running top in terminal (without charger being connected).
tigeli ( 2016-02-24 19:02:08 +0200 )edit
@tigeli, to be absolutely precise, I can see some change in battery life too, but so little that it wouldn't affect my answer to the question two. Do you have an official Jolla Tablet (meaning that you are an IGG backer who received your perk from Jolla)? If so could you please tell me the number you see in Settings/About Product/Hardware Adaptaion?
Moshroom ( 2016-02-24 22:45:58 +0200 )edit
@Moshroom Yeah, I have one.. "Device adaptation: 0.9.0.9".
(And if you have anything higher than that you are not running the latest software (2.0.1.7) available but something else..)
tigeli ( 2016-02-24 23:01:14 +0200 )edit
@Moshroom Lol :D tigeli is a Jolla employee (I hope he is still on board). :)
Alex ( 2016-02-24 23:06:32 +0200 )edit
@tigeli, thanks. So probably my Tablet is as genuine as it can get without being an official one. Therefore it seems likely that either there is a problem in the System Monitor app or that the tablet version of the Sailfish OS runs some unneccessary battery-eating processes during idle...
Moshroom ( 2016-02-24 23:13:06 +0200 )edit
@Alex Yeah, the correct tense is 'was' as currently I'm working with some other stuff but you never know what future brings.
tigeli ( 2016-02-24 23:15:50 +0200 )edit
@Alex, oh I didn't know that. Well, I guess that's the other possible way to have an official Jolla Tablet.
Moshroom ( 2016-02-24 23:19:42 +0200 )edit
@Moshroom 17 hours idle and battery has depleted from 100% to 92% (without wlan being connected). With the current pace the battery would last about 8 days..
tigeli ( 2016-02-25 16:57:29 +0200 )edit
I set my device to flight mode, turned off the Android support and rebooted. An hour after that the battery shows 98,64 %. So this way it'll last three days (not quite eight, but 33 % more than two). Hovewer the battery drain curve might not be linear so better leave it alone for now and check again tomorrow morning.
Moshroom ( 2016-02-25 20:39:00 +0200 )edit
@tigeli: WLAN is connected, it notifies me of new emails right as they arrive. I don't know how it works, but it does.
nthn ( 2016-02-25 23:02:06 +0200 )edit
@tigeli, I managed to get some improvement with flight mode on, Android support (and System Monitor!) off and all notifications off. It dropped 4 units in 8 hours (from 67 %* at 11 pm to 63 % at 7 am). This way it'd last about 8 days, so same as you.
= in active use before going to sleep it dropped about 30 % in 2 hours
Moshroom ( 2016-02-26 07:14:22 +0200 )edit
Can only comment that the battery life is laughable. With idle tablet it's less than two days for a dead battery, have to keep it in charger all the time. Haven't dug out what's happening, good comments in here so have to take a look. Current state is a joke anyway :)
jsievikorte ( 2016-02-29 05:45:10 +0200 )edit
In this thread linked below: http://askubuntu.com/questions/33640/kworker-what-is-it-and-why-is-it-hogging-so-much-cpu?answertab=votes#tab-top
There are instructions on how to use perf, which can tell you what kworker is doing with those CPU cycles. On Sailfish 'perf' is a separate package of its own, but I had to install 'less' to view the report, and that doesn't seem to be available by default in the i486 repo so I had to add the mer-tools one for that. My results seem to point to a component of the wireless driver, but more reports would only help.
Andy Branson ( 2016-04-12 01:13:05 +0200 )edit

Answer 1

6

answered 2016-03-04 06:24:31 +0200

codeandcreate
1115 ●10 ●18 ●21

updated 2016-03-07 12:33:40 +0200

I have the official one and it doesn't really matter if i use it or not if it is on. besides carddav/caldav sync (daily) and 3 mail accounts (30min) i dont't have any background tasks aktive. since i got it on monday i don't have any long time view on battery life but i think there is no real powersaving mode (Jolla, please comment on this!), when you turn the display off.

0.9.0.9
less than 24 h (read rss feeds, surfing,... ) ~ normal use
24 - 48 h ?
around 1.5%

my last tablet was a Samung Galaxy Tab 7.7. with the same normal usage i got about a half till a week battery life and with no usage more than 2 weeks.

Update: The Android Runtime really eats a lot of power. if i close it, the battery lasts about 100% longer. Still, it i don't have battery times like on my old device, but hey... :)

Update 07.03.: SysMon now show me also nearly 100% Cpu usage all the time. i have not installed anything... just dropped a few mp3s and some reboots (i shut the tablet down over night). But for example since 7:00 only 10% battery drain today. if i look via terminal what top says, kworker is "the bad guy" (http://askubuntu.com/questions/33640/kworker-what-is-it-and-why-is-it-hogging-so-much-cpu?answertab=votes#tab-top).

edit flag offensive delete publish link

Comments

Sorry guys didn't find this thread before I created mine. I got a report out of perf as instructed by that askubuntu page I found to look into kworker (which I see you've linked too) and it looks like it's something to do with the wireless driver:

https://together.jolla.com/question/132947/jolla-tablet-high-standby-cpu-when-not-charging-becomes-unresponsive/

Andy Branson ( 2016-04-12 01:07:15 +0200 )edit

Answer 2

4

answered 2016-05-16 09:58:42 +0200

rinigus

4011 ●42 ●58 ●59

updated 2016-05-16 22:58:19 +0200

When you check CPU %, don't forget to check how much CPU was sleeping. You could do it with mcetool (that will give you sleep and total time from boot) or SysMon with CPU sleep monitored (unofficial version available in OpenRepos, https://openrepos.net/content/rinigus/system-monitor-unofficial-edition ; code at https://github.com/rinigus/harbour-systemmonitor ). [Disclosure: I wrote this SysMon extension for CPU sleep when checking out Nexus 4 port].

I would expect that the tablet has a rather large CPU sleep %. Now, as far as I understand, CPU utilization reported the % at the time when tablet is non-suspended, i.e /proc/stat statistics is related to non-sleep time. So, I assume that when your tablet is mostly sleeping and waking up once in a while, it runs housekeeping processes and then goes to sleep as well. This would correlate with high CPU sleep and high CPU utilization %. To get some idea of CPU utilization against the wall clock, try to multiply ( 100 - CPU_sleep_% ) / 100% * CPU_utilization_% .

When you plug in the charger, CPU sleep is disabled. As a result, since the tablet has almost nothing to do, CPU utilization would be very low.

So, that all fits with your case as well as with relatively modest battery consumption.

Now, usually, we see ~10% CPU [corrected the value] utilization in sleep in Nexus 4. It has 4 CPUs and I suspect that its mainly one threaded program doing the housekeeping. In your case, you have 75% CPU usage or higher. Either it means that housekeeping is done on multiple threads or there is something else going on.

When compared to Android tablets, I have Nexus 7 and that can stay idle for a week, no problem. But there is a difference in CPU architecture and many other aspects.

edit flag offensive delete publish link

Comments

1

Thanks for this, your CPU sleep addition to SysMon has helped see what's going on here. These values of ~75% usage would correlate with 3 of the 4 CPUs being suspended, and the other being awake but with low load. Do you think it would be a good idea to subtract CPU sleep from utilization in SysMon?

The suspend success and failure graphs show 'no data' on the tablet. I'll have a look see what's going on there.

Andy Branson ( 2016-05-16 12:43:24 +0200 )edit

1

Re suspend success/failure: In several ports, file /sys/kernel/debug/suspend_stats contains information regarding suspend stats (see end of https://www.kernel.org/doc/Documentation/power/basic-pm-debugging.txt for example). If you could find where you have this file mounted in the tablet, then please let me know - I'll modify the source accordingly.

Re CPU sleep and utilization: Its seems to create a lot of confusion and I have been confused as well. Perhaps we could extend the utilization graphs by wall clock equivalents.

On the other hand, I'd like to keep current stats as well. Just imagine the case where your device is suspending and having CPU usage low. Such scenario would indicate that it could have suspended more - no point in sitting with low CPU usage and not sleeping :) [unless something else useful is going on]

rinigus ( 2016-05-16 12:56:43 +0200 )edit

1

It seems identical to the phone:

[root@pootle ~]# cat /sys/kernel/debug/suspend_stats
success: 23384
fail: 2424
failed_freeze: 225
failed_prepare: 0
failed_suspend: 1601
failed_suspend_late: 0
failed_suspend_noirq: 0
failed_resume: 0
failed_resume_early: 0
failed_resume_noirq: 0
failures:
  last_failed_dev:      alarmtimer
                        alarmtimer
  last_failed_errno:    -16
                        -16
  last_failed_step:     suspend
                        freeze

I see your point about the CPU usage display, but I think if the suspend graph is still there then the sleep is being reported well enough. Seeing a high sleep number and low CPU usage graph would possible even let people think their tablet battery is lasting longer :)

Andy Branson ( 2016-05-16 13:09:45 +0200 )edit

1

Re suspend debug: Indeed, this is identical to what we have on ports. Do you have SFOS Programming API installed? Maybe you could help to debug the issue? The code in question is at

https://github.com/rinigus/harbour-systemmonitor/blob/master/daemon/datasourcesuspend.cpp

We could also move the discussion regarding this SysMon bug to github, under https://github.com/rinigus/harbour-systemmonitor/ by opening corresponding issue.

rinigus ( 2016-05-16 13:17:12 +0200 )edit

1

Yes, I've uncommented your debug lines to test. So far I see 'Suspend DATA' in the log, but no 'suspend: ' line after. Will look deeper and create an issue.

Andy Branson ( 2016-05-16 13:38:29 +0200 )edit

Answer 3

2

answered 2016-02-24 17:15:53 +0200

cgfrost
41 ●1 ●1 ●6

updated 2016-02-24 17:32:18 +0200

Got mine (in the UK) via yoybuy yesterday, they tried to use EMS first and it got sent back at customs in China but then used DHL and no other problems. Came with SF pre-installed and all the original packaging but it's been opened at some point. I also feel bad for going rouge but it's better than not having one. To answer your questions.

0.9.0.21
Less than 24 hours but it was getting heavy use.
Not sure yet, need to use it more.
92% at idle, plugging the charger in makes it 1.8%. (Using the same monitoring tool as you).

Something isn't right with the CPU usage me thinks.

edit flag offensive delete publish link

Comments

4

one likely reason for the high CPU usage is that it is scaled back during idle and your monitoring software counts utilisation based on current speed, not maximal speed of the CPU; while on charger, the CPU probably goes full-speed at all times

mornfall ( 2016-02-24 17:37:14 +0200 )edit

Well regarding the CPU usage on the tablet app 'System Monitor' - it was just compiled and uploaded by not the original developer, so there is still the chance that this is a Bug in System Monitor and not optimized at all for the Tablet.

Alex ( 2016-02-24 19:13:22 +0200 )edit

@mornfall, that is actually a very good point. The first rule in doing science is to actually understand how your equipment works...

Moshroom ( 2016-02-24 23:01:50 +0200 )edit

I think you are correct. After a few more days of use it is still consistently showing very high CPU usage when not plugged in but the rate the battery is dropping I expect to get almost a week out of it with only light usage. That would be so if the CPU was maxed out all the time, I haven't noticed it getting hot either.

cgfrost ( 2016-02-26 13:45:15 +0200 )edit

Answer 4

1

answered 2016-03-02 09:45:30 +0200

lupastro

4100 ●57 ●76 ●76

updated 2016-03-02 09:46:37 +0200

Well, this is how the last 8 hours usage report looks like. The tablet was on idle status, because I was sleeping...

image description

That's the performance of the Tablet (in idle mode) between 22:56 yesterday and 6:56 today. I find the WLAN peaks interesting....as well as the constant 93,5% CPU usage.

The battery dropped about 6% in the meantime.

Android Support is not installed. The only apps I installed are "Sysmon", "QuasarMX" and "Sailfish Utilities". Any of them was running during the night.

edit flag offensive delete publish link

Answer 5

1

answered 2016-03-08 21:52:04 +0200

jaine

2185 ●23 ●34 ●41

https://www.facebook.com/...

0.9.0.9
Not sure yet
Not sure yet, but seems like 1 hour idle -> 1% battery drop
87.7%

The CPU usage is really interesting...

edit flag offensive delete publish link

Answer 6

1

answered 2016-03-16 04:27:37 +0200

wkevink
11 ●1 ●1 ●3

hey, i got mine via yoybuy to canada last week. it took about two weeks to ship here.

my answers:

0.9.0.21
sometimes everyday if i use realy often
48 h
85 %

edit flag offensive delete publish link

Answer 7

1

answered 2016-05-13 23:02:36 +0200

jolladiho
2932 ●45 ●74 ●67

original jolla tablet

0.9.0.25
every day if using it, to often for a tablet
48h (drains with a rate of round about 1% / h)
98%

have a test running with some mods it still needs more 12h to run. At the moment the drain is 4% in the first 12h if tablet is idle. More tablet like, but not perfekt and only a workaround.

Let's hope Jolla can fix something with the next update.

edit flag offensive delete publish link

Comments

which mode have you done? ;-)

minojolla ( 2018-10-27 11:22:54 +0200 )edit

I wrote a script that puts the last 2 cpu's offline if the screen turns off and sets the powersave mode via mcetool. That slows down the consumption a little bit.

echo "0" > /sys/devices/system/cpu/cpu2/online
echo "0" > /sys/devices/system/cpu/cpu3/online
mcetool -F enabled

jolladiho ( 2018-10-27 20:23:32 +0200 )edit

Seem a brillant idea. I can try applying theese 3 lines via Terminal? I am testing a Jolla 'Aigo' Tablet, but I noticed this big issue. Is it possible that Jolla forget about a similar issue, in a tablet?

minojolla ( 2018-10-27 21:31:43 +0200 )edit

The idea was better then the power management of the tablet.

Yes, that are terminal commands. You need root permissions, you do it by your own risk and you know what you are doing.

Nobody cares that tablet any more, because it and its adaptation are a products of a failed project.

jolladiho ( 2018-10-28 12:28:56 +0200 )edit

Ok so basically It became devel-su password echo "0" > /sys/devices/system/cpu/cpu2/online echo "0" > /sys/devices/system/cpu/cpu3/online mcetool -F enabled

But where the instruction states the 'display off' condition? And, there is an undo command line?

minojolla ( 2018-10-28 14:45:10 +0200 )edit

Answer 8

1

answered 2016-05-15 20:38:46 +0200

shfit
3003 ●30 ●38 ●41

updated 2016-05-16 01:24:36 +0200

original jolla tablet

1) 0.9.0.25 2) every day 3) from 25% to 20% in about 4h when idle in airplane mode with screen off and system monitor running 4) about 6% when connected to charger, about 75% when disconnected.

there is something wierd going on. some system process seems to be hogging cpu while the tablet is idle and disconnected from the carger. i inspected it briefly with top and all i could see was that a kworker and lipstick were using cpu, but i havent investigated it n detail.

edit flag offensive delete publish link

Questions for Jolla Tablet owners (about battery and CPU usage)

Comments

8 Answers

Comments

Comments

Comments

Comments

Question tools

Stats

Related questions

Questions for Jolla Tablet owners (about battery and CPU usage)

Comments

8 Answers

Comments

Comments

Comments

Comments

Question tools

Public thread

Stats

Related questions