RK3229 playback

[tomsag]

There is a strange behavior in RK3229 (cheap) android TV boxes, so I wanted to start a new topic.
Rockchip is very powerful and gets 4K playback really good, if the firmware is right.
For example, I can reproduce a 75Mbit/s stream without any problems in video/audio once booted on my MXQ 4K Rockchip box.
But, after some time (day or more) the system tends to choke while playing videos.
This could be due to temperature regulations on the processor (which slows down in excessive temp) or some other android processes (or bugs in the OS), which I'm trying to look for.
I'm sure you will have the same experience but, no-one has pointed it as far as I know to the subject!

Any experience is welcome!

[mo123]

There is a strange behavior in RK3229 (cheap) android TV boxes, so I wanted to start a new topic.
Rockchip is very powerful and gets 4K playback really good, if the firmware is right.
For example, I can reproduce a 75Mbit/s stream without any problems in video/audio once booted on my MXQ 4K Rockchip box.
But, after some time (day or more) the system tends to choke while playing videos.
This could be due to temperature regulations on the processor (which slows down in excessive temp) or some other android processes (or bugs in the OS), which I'm trying to look for.
I'm sure you will have the same experience but, no-one has pointed it as far as I know to the subject!

Any experience is welcome!

The MXQ 4K RK3229 is known for overheating issues at 1.4Ghz.
The CPU and system throttles down and then you get bad performance.
More expensive RK3229 devices don't have this problem and have better firmware support.

[scooter2014]

Yes agree. Cheap unsupported may never be fixed because locked boot and other issues. D5 and Nagrace two devices i know work and ui fast so proves the point.

[tomsag]

The MXQ 4K RK3229 is known for overheating issues at 1.4Ghz.
The CPU and system throttles down and then you get bad performance.
More expensive RK3229 devices don't have this problem and have better firmware support.

Yes agree. Cheap unsupported may never be fixed because locked boot and other issues. D5 and Nagrace two devices i know work and ui fast so proves the point.

I can agree with you both, however overeating can be solved by better cooling. If it's about cooling, let us be honest and solve it. Perhaps, my (much bigger) CPU cooler is not
appropriate, or the 3M cooler tape is not right to put it on RK3229. Will try to put the whole thing on very cool place and see whether it happens, but there could be some processes
which are not needed on these boxes and have to be disabled.
Or, there might be bugs in the ROM which let to overheating, or I do not know?

But a lot of users can benefit from this shared experience.

[scooter2014]

Easy test if it has vents slap a 3 inch usb case fan to it. That will blow more air then it would ever need. Also cheap fix. 😃 The other devices dont have this issue so being honest if its fix with fan not firmware. Its hardware then you can mod as you need to resolve

[tomsag]

It's NOT the temperature, since the box plays now continuous (over 4 hours a 4K HDR video with over 75Mbit/s) without stuttering.
However it was booted today, and some background programs are not active anymore - deleted some stuff which is not needed.
Load monitoring/testing via SSH droid. 30% load is Kodi.

BTW, has anyone an idea how to get CPU temperature from Rockchip via Linux shell?

[shomari_]

There is a strange behavior in RK3229 (cheap) android TV boxes, so I wanted to start a new topic.
Rockchip is very powerful and gets 4K playback really good, if the firmware is right.
For example, I can reproduce a 75Mbit/s stream without any problems in video/audio once booted on my MXQ 4K Rockchip box.
But, after some time (day or more) the system tends to choke while playing videos.
This could be due to temperature regulations on the processor (which slows down in excessive temp) or some other android processes (or bugs in the OS), which I'm trying to look for.
I'm sure you will have the same experience but, no-one has pointed it as far as I know to the subject!

Any experience is welcome!

Please upload the full contents of dmesg as follows to help diagnose the issue:

1. Take dmesg at boot.
2. Take dmesg after playing multimedia
3. Take dmesg after slow down symptoms occur.

Upload those dmesg output contents here and I'll help you determine what may be the issue. If throttling is suspected, I can disable the RK specific throttling mechanisms to test your theory. First I suggest looking at the dmesg output.

EDIT: both from your observations and my experience, what's usually attributed to heat-related throttling on RK devices is typically a symptom of something other than heat. I've seen the same thing occur with the RK3368, where (after careful software modification) ALL inbuilt RK throttling can be removed with no risk or adverse performance affects on the CPU. I've done just about all I can to my test devices for the last 6 months, and performance stays level - and anywhere from 5K-10K point difference in system benchmarks over devices with throttling turned on (all RK devices).

[tomsag]

Since I've uninstalled some bloatware, no throttling was detected.
Attached is dmesg from fresh boot, perhaps you might find out something...
BTW, how to disable RK Cpu throttling?
dmesg_boot.txt

[shomari_]

Since I've uninstalled some bloatware, no throttling was detected.
Attached is dmesg from fresh boot, perhaps you might find out something...
BTW, how to disable RK Cpu throttling?
[ATTACH]n611418[/ATTACH]

Thanks for the log. To disable throttling at the software level you have to modify resource and boot files to disable the values in the device tree at node level and in init script parameters.

[shomari_]

I'm curious what the dmesg shows after multimedia playback, based off those log contents. Also, do you know what the actual ddr/gpu/cpu operating points are?

You have a very old firmware, it seems, by the way.

[tomsag]

Agter heavy load, I've got some interesting readings on dmesg (temperature was 105 C) and sound vanished, but it recuperated on standby play...
Hre some paste from recent dmesg:
[ 8746.992970] [<c035437c>] (update_timer+0xe4/0x148) from [<c005c508>] (kthread+0xa0/0xac)
[ 8746.992992] [<c005c508>] (kthread+0xa0/0xac) from [<c000dba0>] (ret_from_fork+0x14/0x34)
[ 8863.163670] vcodec_service: 20030000.rkvdec: power off...
[ 8863.163737] vcodec_service: done
[ 8866.241840] vcodec_service: 20030000.rkvdec: power on
[ 8931.167908] vcodec_service: 20030000.rkvdec: power off...
[ 8931.167978] vcodec_service: done
[ 8937.375911] vcodec_service: 20030000.rkvdec: power on
[ 8942.515736] lowmemorykiller: Killing 'd.apps.sshdroid' (1783), adj 1000,
[ 8942.515736] to free 29300kB on behalf of 'kswapd0' (64) because
[ 8942.515736] cache 183940kB is below limit 184320kB for oom_score_adj 1000
[ 8942.515736] Free memory is 4kB above reserved
[ 8942.526481] init: untracked pid 1858 killed by signal 9
[ 8944.803617] type=1400 audit(1479925053.875:52): avc: denied { call } for pid=3742 comm="main" scontext=u:r:zygote:s0 tcontext=u:r:servicemanager:s0 tclass=binder permissive=1
[ 8944.805643] type=1400 audit(1479925053.875:53): avc: denied { search } for pid=134 comm="servicemanager" name="3742" dev="proc" ino=37854 scontext=u:r:servicemanager:s0 tcontext=u:r:zygote:s0 tclass=dir permissive=1
[ 8944.805886] type=1400 audit(1479925053.875:54): avc: denied { read } for pid=134 comm="servicemanager" name="current" dev="proc" ino=40068 scontext=u:r:servicemanager:s0 tcontext=u:r:zygote:s0 tclass=file permissive=1
[ 8944.806062] type=1400 audit(1479925053.875:55): avc: denied { open } for pid=134 comm="servicemanager" path="/proc/3742/attr/current" dev="proc" ino=40068 scontext=u:r:servicemanager:s0 tcontext=u:r:zygote:s0 tclass=file permissive=1
[ 8944.806287] type=1400 audit(1479925053.875:56): avc: denied { getattr } for pid=134 comm="servicemanager" scontext=u:r:servicemanager:s0 tcontext=u:r:zygote:s0 tclass=process permissive=1
[ 8944.810888] type=1400 audit(1479925053.885:57): avc: denied { call } for pid=3742 comm="main" scontext=u:r:zygote:s0 tcontext=u:r:system_server:s0 tclass=binder permissive=1
[ 8944.811129] type=1400 audit(1479925053.885:58): avc: denied { transfer } for pid=3742 comm="main" scontext=u:r:zygote:s0 tcontext=u:r:system_server:s0 tclass=binder permissive=1
[ 8944.839916] type=1400 audit(1479925053.915:59): avc: denied { call } for pid=2568 comm="Binder_B" scontext=u:r:system_server:s0 tcontext=u:r:zygote:s0 tclass=binder permissive=1
[ 8944.842978] init: untracked pid 3739 exited with status 132
[ 8945.064050] init: untracked pid 3742 exited with status 0
[ 8947.696119] type=1400 audit(1479925056.815:60): avc: denied { write } for pid=2141 comm="dropbeard" path="socket:[16065]" dev="sockfs" ino=16065 scontext=u:r:init:s0 tcontext=u:r:init:s0 tclass=tcp_socket permissive=1
[ 8947.704950] type=1400 audit(1479925056.825:61): avc: denied { read write } for pid=3421 comm="DecoderVideo" name="iep" dev="tmpfs" ino=1303 scontext=u:r:untrusted_app:s0 tcontext=ubject_r:video_device:s0 tclass=chr_file permissive=1
[ 8955.164787] lowmemorykiller: Killing 'ogle.android.gm' (1665), adj 1000,
[ 8955.164787] to free 38076kB on behalf of 'kswapd0' (64) because
[ 8955.164787] cache 182608kB is below limit 184320kB for oom_score_adj 1000
[ 8955.164787] Free memory is -1092kB above reserved
[ 8957.658874] type=1400 audit(1479925066.945:62): avc: denied { read } for pid=2141 comm="dropbeard" path="socket:[16065]" dev="sockfs" ino=16065 scontext=u:r:init:s0 tcontext=u:r:init:s0 tclass=tcp_socket permissive=1
[ 8989.804480] vcodec_service: 20030000.rkvdec: power off...
[ 8989.804562] vcodec_service: done
[ 8992.110579] vcodec_service: 20030000.rkvdec: power on
[ 9012.230626] vcodec_service: 20030000.rkvdec: power off...
[ 9012.230714] vcodec_service: done
[ 9012.487039] vcodec_service: 20030000.rkvdec: power on
[ 9023.546643] vcodec_service: 20030000.rkvdec: power off...
[ 9023.546713] vcodec_service: done
[ 9093.746187] vcodec_service: 20030000.rkvdec: power on

[shomari_]

... there should be thermal data in the (full) dmesg output. Your chip's set to throttle the system once it reaches 95°. &#194;&#160;Does the device feel hot?
&#194;&#160;

Re my suggestion to put fan as quick test before screwing with set and thermal. Removing it always risks frying chip at some point. May not even be the issue.

[tomsag]

No, it's only warm - actually measured (from IR thermometer) is not more than 43C.
BTW, how to read CPU temperature of Rockchip from SSH/Linux? My only information is from Kodi and that might be wrong.

[shomari_]

... the correct path actually varies by firmware, usually in /sys ...

[tomsag]

yes I know, but a lot of different data in /sys, what to look for?