Linux Distributions for Orange PI H3 boards

bronco

GreyWolf replied at 2015-7-30 21:36
CPU0:
scaling_max_freq 1536000
scaling_cur_freq 480000

So you reported thermal values when your system was totally idle. BTW: I don't know how it's on H3 but on others sunxi SoCs you only have to query the values of CPU0 since cpufreq settings always apply to all cores.

Would be interesting to "apt-get install stress" and then put significant load on the system: "stress -t 900 -c 4 -m 2 -i 2" (if it survives the 15 min. then your heatsink helps)

loboris

bronco replied at 2015-7-30 17:41
So you reported thermal values when your system was totally idle. BTW: I don't know how it's on H3 ...

with heatsink and fan it can survive stress indefinitely...

GreyWolf

bronco replied at 2015-7-30 19:41
So you reported thermal values when your system was totally idle. BTW: I don't know how it's on H3 ...

Thanks for the idea !

Here is the result:


stress: info: [844] dispatching hogs: 4 cpu, 2 io, 2 vm, 0 hdd
stress: FAIL: [844] (416) <-- worker 850 got signal 9
stress: WARN: [844] (418) now reaping child worker processes
stress: FAIL: [844] (452) failed run completed in 2s

and
/sys/devices/system/cpu/online
3

bronco

GreyWolf replied at 2015-7-31 01:07
Thanks for the idea !

Here is the result:

In other words: When starting stress on all 4 cores within only 2 seconds the SoC's internal temperature sensor reported above 70°C and the driver shut the first CPU core down. In this case, stress aborts. But if you would for example compile the linux kernel mit "make -j4" you might realize that your board shuts down after a few seconds.

@loboris: I'm aware that using an effective cooling solution might help. The question is: What's effective? And measuring an idle system without testing worst case scenarios isn't a good idea when you want know to whether the specific solutions survives a few moments under full load.

BTW: I recently continued to adopt RPi-Monitor to be useable on sunxi devices. Since I'm always interested in thermal numbers under high load (then RPi-Monitor isn't able to query data sources on its own and fails to record values in its RRD databases to be analyzed later) I wrote a small daemon for this purpose: http://forum.armbian.com/index.p ... ts-for-rpi-monitor/

Just stresstested an A20 device (comparable to Orange Pi/Mini):

1. root@lamobo:/mnt/sda# stress -t 900 -c 2 -m 2 -i 2 -d 2
   
2. stress: info: [3209] dispatching hogs: 2 cpu, 2 io, 2 vm, 2 hdd
   
3. stress: info: [3209] successful run completed in 904s

Copy the Code

Looks like this:


In case anyone's interested to start support for the H3 (I'm currently only after AXP209 equipped devices with A10/A13/A20 SoC) this would be a good starting point: https://github.com/XavierBerger/RPi-Monitor/pull/110

Especially the sunxi specific config to get an idea whether there are interesting values exposed via sysfs: https://github.com/ThomasKaiser/ ... 9_cpu_pmu_temp.conf

BTW: Does anyone here has a clue how the PMU/PMIC the H3 uses is called?

loboris

Stress test, OPI-2

1. orangepi@OrangePI:~$ sudo stress -t 900 -c 4 -m 2 -i 2 --vm-bytes 64M
   
2. stress: info: [3016] dispatching hogs: 4 cpu, 2 io, 2 vm, 0 hdd
   
3. stress: info: [3016] successful run completed in 900s
   
4. orangepi@OrangePI:~$

Copy the Code

Max temperature during test 41 C, 1 minute after test 28 C

bronco

loboris replied at 2015-7-31 03:22
Stress test, OPI-2
Max temperature during test 41 C, 1 minute after test 28 C

Cool! :-)

And scaling_cur_freq stayed at 1.5 GHz all the time? Which thermal sensor did you query? temp0 or temp1? And which one of them reports what?

loboris

bronco replied at 2015-7-30 22:06
Cool! :-)

And scaling_cur_freq stayed at 1.5 GHz all the time? Which thermal sensor did you query ...

Freq. was constant max during the test. Both temp readings were equal during the test (and practically constant, reading every 10 sec).
When idle, temp is 28-29C and temp1 27-28C (room temperature 25C).
OPI-2 without case.
Just finished 1 hour stresstest, results are the same.

bronco

loboris replied at 2015-7-31 05:26
Freq. was constant max during the test. Both temp readings were equal during the test (and practica ...

So still none of you H3 hackers as well as the manufacturer of the board has neither a clue which temp value is what (one is clearly from the Thermal Sensor Controller inside the SoC but what's the other?) nor which PMU/PMIC is used on the board?

Had anybody a look into sources which sensors were exposed via sysfs this way?

magicse

Got Somebody to out an signal on the analog TV via OrangePi2?  

blindpet

loboris replied at 2015-7-31 00:47
with heatsink and fan it can survive stress indefinitely...

Can you link to the heatsink and fan unit you bought and a quick how to on how to install it? which pins etc?