OrangePi 3 H6 occasional reset

[zeljko]

install to SDCard:
Armbian_5.89.190626_Orangepi3_Debian_stretch_dev_5.1.12.img

update, upgrade

install image to EMMC 8GB

install lxde (desktop)

I rarely have a reset, 
I think the processor is overheating.

Can I reduce the speed without using the fan?
(settings debian)

 

for experiments, psensors would be good

https://askubuntu.com/questions/41794/how-to-monitor-log-server-hardware-temperatures-load

https://oss.oetiker.ch/rrdtool/download.en.html

[Igor]

58 minutes ago, zeljko said:

for experiments, psensors would be good

 

Type armbianmonitor into the console - we develop this to get the best out of those board without using much resources.

 

59 minutes ago, zeljko said:

Can I reduce the speed without using the fan?

 

Sure. armbian-config -> system IMO fan is not needed. Just place a large enough heat sink.

 

59 minutes ago, zeljko said:

I think the processor is overheating.

CPU voltage settings https://github.com/armbian/build/pull/1434 are not well tuned which is normal since this board is still under development.

 

[dziobak]

 

13 hours ago, zeljko said:

 

install to SDCard:
Armbian_5.89.190626_Orangepi3_Debian_stretch_dev_5.1.12.img

 

 

Fixed voltage was released 4 days later. Look for release 5.90.190630 and next.

[data]

On the Lite2, I am experiencing a complete freeze every few (1-3) days. This is happening since the past two or three weeks

The system  stops responding and I have to power off and on again. There is nothing in any of the logs. It's as if one would 

pull the plug.

 

Can this be related to the modified voltages?

[dziobak]

Use StabilityTester

Put correct values in stabilityTester.sh. For OPi3 I used this:

MINFREQUENCY=1000000 #Only test frequencies from this point.
MAXFREQUENCY=2000000 #Only test frequencies upto this point.
COOLDOWNTEMP=55000 #Cool down after a test to mC degrees
COOLDOWNFREQ=720000 # Set to this speed when cooling down

CPUFREQ_HANDLER="/sys/devices/system/cpu/cpu0/cpufreq";
SCALINGAVAILABLEFREQUENCIES="scaling_available_frequencies";
SCALINGMINFREQUENCY="scaling_min_freq";
SCALINGMAXFREQUENCY="scaling_max_freq";

SOCTEMPCMD="/sys/class/thermal/thermal_zone0/temp"

REGULATOR_HANDLER="/sys/class/regulator/regulator.4"
REGULATOR_MICROVOLT="microvolts"

 

[data]

There was no load on the Lite2. I just keep it on to see how stable it is.

[dziobak]

If this is no load, then CPU work with lowest frequency with voltage like original (no undervoltage reason for freeze). Problem can be when freq > 1GHz, where we try use lowest possible, stable voltage (less voltage = less heat = no CPU throttling). My stable values may not work for all CPUs, so StabilityTester is best tool for check this. But not in case, when freeze is when CPU work with lowest frequencies (no system load).

[data]

Since I had no better idea, I tried stabilitytester as suggested:

 

The lite2 freezes always when stabilitytester reaches this:

TEST Freq:  816 MHz    Soc temp:  40.92 ℃    CPU Freq:  816 MHz    CPU Core:  960 mV
TEST Freq:  888 MHz    Soc temp:  40.59 ℃    CPU Freq:  888 MHz    CPU Core:  960 mV
TEST Freq: 1080 MHz    Soc temp:  42.13 ℃    CPU Freq: 1080 MHz    CPU Core:  960 mV
TEST Freq: 1320 MHz    Soc temp:  52.48 ℃    CPU Freq: 1320 MHz    CPU Core:  960 mV

 

I added additional cooling and now, it reboots instead of freeze

 

So is this a voltage related issue?

 

 

[dziobak]

I think so. First, change values in StabilityTester.sh to shows correct voltage values, try this (if not, try other regulators):

 

REGULATOR_HANDLER="/sys/class/regulator/regulator.4"

and set min freq value to 1000000 (there is no need to test lowest frequencies).

 

Next, get current DTB file (/boot/dtb/allwinner/sun50i-h6-orangepi-lite2.dtb for Lite 2) and convert this binary to source, by:

dtc -I dtb -O dts -o sun50i-h6-orangepi-lite2.dts sun50i-h6-orangepi-lite2.dtb

open source file for editing, change voltage value for problematic frequency (after decoding numbers are hex, but you can put decimal inside) +10mV (first value of 3), write then generate DTB:

dtc -I dts -O dtb -o sun50i-h6-orangepi-lite2.dtb sun50i-h6-orangepi-lite2.dts

put new binary in /boot/dtb/allwinner/

reboot, use StabilityTester

[data]

@dziobak

Thank you very much for your reply. Prior to running the script, I changed to regulator.4 as suggested in your post above.

Once I've tried different voltages I'll let you know the results.

 

[dziobak]

Check name of every regulator:

cat /sys/class/regulator/regulator.*/name

I've used vdd-cpu, which is regulator.4 in OPi3.

 

In last post here are my stable values for OPi3 also.