Orangepi 3 h6 allwiner chip

[krachlatte]

Hi guys,

 

are ther any plans to upgrade u-boot to 2019.10 from u-boot v2019.04, because libreelec is allready using this and they have a working device tree for the H6 beelink device

 

many thanks

 

 

[martinayotte]

10 minutes ago, krachlatte said:

are ther any plans to upgrade u-boot to 2019.10 from u-boot v2019.04

It is planned, I'm already using it for my local builds ...

[Igor]

13 minutes ago, martinayotte said:

It is planned, I'm already using it for my local builds ...

You already sort out patches? Then we can perhaps squeeze this in by the end of the month?

[martinayotte]

11 minutes ago, Igor said:

You already sort out patches? Then we can perhaps squeeze this in by the end of the month?

I've sorted out patches that were preventing builds, but there still some patches where chunks failed without disturbing builds, let says, for OPi3/OPiLite2/OPiOnePlus/PineH64 as well as H3/H5 ones .

Yes, maybe end of the month is a good target ...

[Igor]

10 minutes ago, martinayotte said:

I've sorted out patches that were preventing builds, but there still some patches where chunks failed without disturbing builds, let says, for OPi3/OPiLite2/OPiOnePlus/PineH64 as well as H3/H5 ones .

Yes, maybe end of the month is a good target ...

IHMO u-boot bumping is not that critical and can wait for next major release. Time for proper testing, not just for adjusting patches, is missing component  I would like to start following this if there will be no major oppositions to the plan ... I am probably travelling in the early December. In any case its busy month for all.

[kexec]

does anyone know where is the trick, older image generates less heat than newer one? is this related with cpu voltage or uboot version ?

 

for example older image:

Welcome to Ubuntu Bionic with Armbian Linux 5.1.12-sunxi64
root@orangepi3:~# cat /sys/devices/virtual/thermal/thermal_zone[01]/temp
38168
39109
root@orangepi3:~# cat /sys/devices/virtual/thermal/thermal_zone[01]/type
cpu_thermal
gpu_thermal

 

newer  stays around 48 celsius in same environment (no load on device, room temperature same,small passive cooler on cpu)

 

Welcome to Ubuntu Bionic with Armbian Linux 5.3.9-sunxi64

root@orangepi3:~# cat /sys/devices/virtual/thermal/thermal_zone[01]/temp
47375
48517
root@orangepi3:~# cat /sys/devices/virtual/thermal/thermal_zone[01]/type
cpu-thermal
gpu-thermal

 

 

[megi]

Did you measure the heatsink/package temperature? It should be a few (~5°C, depending on the difference between room/die temps) centigrades lower than the die temperature that's reported by the driver.

For example for me, I see 43°C idle (all cores at 1.8GHz) reported by the driver, and 37-38°C measured by the IR thermometer. (no heatsink, but I use a fan) Without a fan it would be close to 50-55°C idle.

Thermal drivers and OPP tables maz be different between those two releases.

[dolphs]

On 11/8/2019 at 5:54 PM, martinayotte said:

I've sorted out patches that were preventing builds, but there still some patches where chunks failed without disturbing builds, let says, for OPi3/OPiLite2/OPiOnePlus/PineH64 as well as H3/H5 ones .

Yes, maybe end of the month is a good target ...

should be excellent considering kernel 5.4 will be stable by then, will stay tuned chaps. thanks for your efforts!

[kexec]

On 11/9/2019 at 4:13 PM, megi said:

Did you measure the heatsink/package temperature? It should be a few (~5°C, depending on the difference between room/die temps) centigrades lower than the die temperature that's reported by the driver.

For example for me, I see 43°C idle (all cores at 1.8GHz) reported by the driver, and 37-38°C measured by the IR thermometer. (no heatsink, but I use a fan) Without a fan it would be close to 50-55°C idle.

Thermal drivers and OPP tables maz be different between those two releases.

no, i havent measured outside temps as have no IR thermometer. but once I am changing sd card I can feel that little difference.

I guess more scientific way would be to measure power consumption. will try to do that and update

 

BSP image from opi page generates highest temp, then goes armbian latest and  then older version (for example 5.1.12) which acts like i would expect when there is no action  happening on cpu

[Werner]

There was  a - let's say - "controversial" discussed patchset that optimizes the opp table for the H6 SoC and at least in my eyes it turned out that these boards are not really identical. Some even crash with the new table, others (like mines) running decent with it. Their components seem to have quite wide tolerances. A too high voltage for a certain chip could lead to a higher temperature.

Check this: https://github.com/armbian/build/pull/1547

[JORGETECH]

17 hours ago, Werner said:

There was  a - let's say - "controversial" discussed patchset that optimizes the opp table for the H6 SoC and at least in my eyes it turned out that these boards are not really identical. Some even crash with the new table, others (like mines) running decent with it. Their components seem to have quite wide tolerances. A too high voltage for a certain chip could lead to a higher temperature.

Check this: https://github.com/armbian/build/pull/1547

Was this change introduced in the nightlies? Because I'm running nightly and my board boots just fine (I'm writing from it right now).

[Werner]

I think the "problems" started somewhen around this point: https://github.com/armbian/build/pull/1407

[JORGETECH]

12 hours ago, Werner said:

I think the "problems" started somewhen around this point: https://github.com/armbian/build/pull/1407

Can I run an stability test or does it need to be verified by other means?

[Werner]

I never dug deeper into this situation as it worked for me nicely out of the box. I just followed the discussion more or less.

[kexec]

i have checked power consumption when device boots up and nothing is connected to it

 

Ubuntu	opi BSP 4.9 kernel	490ma
ubuntu	armbian 5.1.12	330ma
ubuntu/debian	armbian 5.3.9	360ma
LibreELEC	9.80-nightly-20191108-27a5673	330ma
android	7 (from opi site)	490ma

 

 

[Localhost]

There is a chance that higher power consumption is coming from graphics unit being used (probably)

[jernej]

18 minutes ago, Localhost said:

There is a chance that higher power consumption is coming from graphics unit being used (probably)

LibreELEC uses it too, but admittedly at low frequency, 270 MHz IIRC, because you don't need higher speed for GUI rendering.

[JORGETECH]

Is there any way to change a dts file when compiling Armbian? The GPU is disabled in the dts file for H6 and I wanted to enable it just for curiosity.

[martinayotte]

14 hours ago, JORGETECH said:

The GPU is disabled in the dts file for H6

Are you sure ? Which image are you using ?

What "cat /proc/device-tree/soc/gpu@1800000/status" is reporting ?

[JORGETECH]

23 hours ago, martinayotte said:

Are you sure ? Which image are you using ?

What "cat /proc/device-tree/soc/gpu@1800000/status" is reporting ?

I guess the GPU is active since the output from that command is "okay".

 

Is there an Armbian patch that enables it? I can see it's disabled in the kernel code (display engine disabled too): https://megous.com/git/linux/tree/arch/arm64/boot/dts/allwinner/sun50i-h6.dtsi?h=opi3-5.3

 

It's a pity that Mesa has support for this GPU disabled (or not implemented?) I would like to test even though it could be unstable.

[martinayotte]

15 minutes ago, JORGETECH said:

I guess the GPU is active since the output from that command is "okay".

Right !

15 minutes ago, JORGETECH said:

I can see it's disabled in the kernel code (display engine disabled too)

They are disabled in DTSI, because it common practice, but enabled in Main DTS of OPi3 : https://megous.com/git/linux/tree/arch/arm64/boot/dts/allwinner/sun50i-h6-orangepi-3.dts?h=opi3-5.3

This mean it is not a kernel issue, but libraries and applications ...

[JORGETECH]

4 minutes ago, martinayotte said:

They are disabled in DTSI, because it common practice, but enabled in Main DTS of OPi3 : https://megous.com/git/linux/tree/arch/arm64/boot/dts/allwinner/sun50i-h6-orangepi-3.dts?h=opi3-5.3

I was just going to point that out.

6 minutes ago, martinayotte said:

This mean it is not a kernel issue, but libraries and applications ...

I could ask in Mesa IRC if I can enable it for the Mali T720 MP2 for testing or if it's not implemented yet.

[jernej]

1 minute ago, JORGETECH said:

I could ask in Mesa IRC if I can enable it for the Mali T720 MP2 for testing or if it's not implemented yet.

For T720 you need at least kernel 5.4 and mesa master along with this patch. I'll start testing mesa support for T720 in LibreELEC soon, but according to developer who made T720 mesa PR to LE, it is already quiet good. Hopefully by the time of mesa 20, major bugs will be fixed.

[JORGETECH]

5 hours ago, jernej said:

For T720 you need at least kernel 5.4 and mesa master along with this patch. I'll start testing mesa support for T720 in LibreELEC soon, but according to developer who made T720 mesa PR to LE, it is already quiet good. Hopefully by the time of mesa 20, major bugs will be fixed.

Just what I was looking for! In that case I'll wait to kernel 5.4 and test.

[martinayotte]

3 minutes ago, JORGETECH said:

In that case I'll wait to kernel 5.4 and test.

For sunxi-dev, it is planned to switch to 5.4.y by the end-of-the-month, maybe next week ...

[AZ8]

I have 2  questions:

1. how far Armbian on Opi3 from stable point?

2. did someone compare performance Kodi on Armbian over LibreELEC? Is LibreELEC significantly faster?

[jcaron]

24 minutes ago, AZ8 said:

1. how far Armbian on Opi3 from stable point?

 

This probably depends on your specific use case and the specific features you use. In my case I find it very stable, but I don't use any of the video acceleration features for instance, or sound, or cameras, and I haven't tested WiFi much. The only thing I'm waiting for is 3D GPU support.

[GeorgeP]

2 hours ago, AZ8 said:

1. how far Armbian on Opi3 from stable point?

2. did someone compare performance Kodi on Armbian over LibreELEC? Is LibreELEC significantly faster?

 

1:      I'm running an OPi3 as a headless server and I have no issues whatsoever....

Welcome to Ubuntu Bionic with Armbian Linux 5.3.0-sunxi64
System load:   0.02 0.07 0.03  	Up time:       55 days		
Memory usage:  36 % of 1993MB 	Zram usage:    5 % of 996Mb  	IP: 192.168.x.x
CPU temp:      54°C           	
Usage of /:    52% of 7.1G

2: I have no idea. If I wanted a media centre I'd just buy a TV-box

[AZ8]

Hi guys!

 

Is it possible to include this patch into Armbian tree?

 

patch: https://lists.freedesktop.org/archives/pulseaudio-discuss/2019-June/031168.html

explanation: https://habr.com/ru/post/456476/

 

[Igor]

2 hours ago, AZ8 said:

Is it possible to include this patch into Armbian tree?

If you prepare it and maintain when needed, why not. And for all different kernel options of course: https://github.com/armbian/build/tree/master/config/kernel