New OPi Zero - Yet another high temperature issue...

[bozden]

1 hour ago, kfishy said:

BTW on mainline kernel it idles around a reasonable 45C or so at room temp. 

 

I don't know anything on mainline kernel, never used it. Do you know what the difference as far as DVFS is concerned?

 

[kfishy]

9 minutes ago, bozden said:

 

I don't know anything on mainline kernel, never used it. Do you know what the difference as far as DVFS is concerned?

 

No idea, cpufrequtils don't even work properly on mainline so I can't even tell if DVFS is on or not. 

[Atgeek]

The opi zero  need an active + passive cooling system, in idle the cpu temp with just 2 core enabled is 55/60 C

[kfishy]

8 hours ago, Atgeek said:

The opi zero  need an active + passive cooling system, in idle the cpu temp with just 2 core enabled is 55/60 C

I realize it's not the coolest chip ever made but what I don't get is why the temperature is substantially lower under mainline (by at least 10C). I'd get well over 60C when idling under 3.4 but only 50C under mainline. 

[chwe]

12 hours ago, Atgeek said:

The opi zero  need an active + passive cooling system, in idle the cpu temp with just 2 core enabled is 55/60 C

IMO a small shitty passive copper cooler is enough (idles @55°C, rev. 1.4, 3.4.113-sun8i Kernel). Only if you stress it with minerd --benchmark it starts themal throttling after 2-3min (4cores active 912 --> 768MHz max 74°C).  I'm not happy with the 'improvement' of rev. 1.4 but it's not often that I need full cpu power for my applications. So it depends on your use case if active cooling is needed or not. I think, it doesn't make sense to clock it @1200 MHz cause it heats up way to fast before themal throttling starts.

[bozden]

In all of uses I decided to add a small 5V 30x30 fan to the design. Yesterday in one of the installations (OPi One based push-button video player) the cable got disconnected during installation and today it shows bad symptoms... It is in a controlled museum air conditioning. Adding that fan keeps the temperature below 40 C at all times (OPi one).

 

Yes it depends on the use case but for any "professional" application I would not gamble... In addition to v1.4 gamble

 

 

[kfishy]

So I've done some more tests and it seems on mainline the OPi Zero is stuck at around 480MHz constantly. However on 3.4 it supposedly idles around 240MHz according to CPUFreq so it really shouldn't be heating up to this extent.

[ykchavan]

I got OPi zero v1.4 yesterday and it has same problem.

I used 2A power, tronsmart usb cable, hdmi/gpu/usb/ethernet are all off. It still remains 61C for 240MHz. Cpuminer takes it to 82-84C at 480MHz for long time.

 

[Igor]

1 hour ago, ykchavan said:

It still remains 61C for 240MHz.

Can you measure with some external meter to see if it matches reported by sensor?

[ykchavan]

I do not have external meter to check temperature. But I can not touch it for more than 3-4 seconds.

[Igor]

8 hours ago, ykchavan said:

I do not have external meter to check temperature. But I can not touch it for more than 3-4 seconds.

Understand, thank you. Chinese are claiming its only a false reading and that's why we need numbers  to understand how far from reality this is.

[ykchavan]

This is reply I got from Shenzhen Xunlong Software CO.,Limited aliexpress store

It is normal at 45-80degrees. Cause the Zero is a tiny baord.
And we have set the dynamic frequency configuration on our offical images. This can help decrease the temp.
It will be lower under linux image than the Android image.

[chwe]

On 25.7.2017 at 10:45 PM, Igor said:

Can you measure with some external meter to see if it matches reported by sensor?

 

Temperature measured with a DS18b20 on the SoC is 56-57°C (internal temp 69-70°C, during stress --cpu 4). Idle @480MHz 45°C (internal temp 52°C). There'll be some more 'insights' here:

 

[ykchavan]

I tested using DS18B20 connected to arduino. It was quick and very dirty setup. Good to get rough idea. I have ordered infrared thermometer on aliexpress.

It is important to note temperature depends on area of package where sensor is in contact. Some contact area gives high temperature. Some gives comparatively lower.

I got similar results. On OPi zero v1.4(dram 428, usb off and wifi on)

43C @ 240MHz. 'armbianmonitor -m' gives 60C

65C @ 648/480MHz with 'minerd --benchmark'. armbianmonitor gives 79/80. Lots of cooling states seen on opi-monitor.

 

on OPi PC(dram 624 and ethernet connected)

41C @ 480MHz. 'armbianmonitor -m' gives 38C

71C @ 1200/1296MHz with 'minerd --benchmark'. armbianmonitor gives 75/78. Few cooling states.

 

Edited July 29, 2017 by ykchavan
few corrections

[ykchavan]

I should also note that OPi zero is dissipating heat to nearby components. dram, sdcard holder, wifi chip all are hot.

OPi zero dram temperature is 39C at idle conditions.

OPi PC dram is 35C at idle conditions.

 

[ykchavan]

OPi zero v1.4 idle conditions TestPoints

1V2S  1.21

3_3V   3.41

1_5V  1.36

with minerd --benchmark

1V2S  1.21

3_3V   3.41

1_5V  1.36

 

Yes, there is no difference in both conditions. On OPi PC I can see clearly rise in voltage when cpu mined.

 

[Igor]

Thank you! 

https://github.com/armbian/documentation/commit/77128469627c56914ac85270359d68dfdb05015d

[chocomega]

Hi guys,

I have an Opi zero v1.4 with the latest ubuntu legacy image and armbianmonitor also reports high temp (~65 - 70 without the case, up to ~80 with the case and expansion board)

 

By the way, when I shutdown the system (halt command) the Opi still generates heat (not hot but warm). if we assume the power supply is good enough, is it an expected behavior when shutdown?

If not, I guess I'll have to try another power supply.

Thanks

[martinayotte]

There is no real PMIC on OPiZero, this means the power supply still plugged on the SoC, even after shutdown command.

 

[chocomega]

3 minutes ago, martinayotte said:

There is no real PMIC on OPiZero, this means the power supply still plugged on the SoC, even after shutdown command.

 

Ok, I guess I should not worry about it then.

Thanks for the information.

[jim st]

 I have an Orange Pi zero 512+ v1.4 board, ordered with the audio daughter board, and the cute black box.

 

in the black box, it runs nearly 80 C, so I pulled it out and am running it in the open. 

 

I am running the latest download i could find, Armbian 5.30 3.4.113 with the heat situation.
 

I have a Radio Shack IR thermometer, and did some measuring between the armbianmonitor -m numbers and the RS device.

 

With the system quiet, the armbianmonitor-m (amm) reading is 63C.  The radio shack when positioned so the IR detector views the H2 chip, is seeing about 50.4C.

 

I have some canned air, here, and cool the chip down to 34C on the Radioshack device.  The reading on the amm reading dipped to 48C.  So there is about a

12 or 13C delta that holds up on both ends of the scale for my test.

 

I also had to fix a problem with the dhcp on the wired connection killing the default gateway (another thread topic) but worth mentioning here that I set to a static IP to make sure the device didn't go off the grid, causing me to think it was in thermal lockup.

 

I'm now going to get some various cooling gizmos (heatsinks) from my pi cooling work and see if I can get it down 10 or 15C inside the enclosure.

 

thanks for all the info here and all the thermal thread work, appreciate all the different insights and things to look at.

 

Quote

Thumb test please: let 'armbianmonitor -m' run, then press your thumb for 15 seconds on H2+ and provide output from OPiZ-1 and OPiZ-3 (for both also 'armbianmonitor -u' output, at least '### cpuinfo' section!)

 

[pzw]

Infrared thermometers need to be setup correctly for the material they need to measure.  (Emissivity setting in the thermometer..) See http://www.thermoworks.com/learning/emissivity_table for some real life numbers. Another issue is that they are not measuring a spot, but they are measuring in a cone. Therefore if you are too far away from the surface, then the measurement might be completely wrong since it is not just measuring the point (the laser dot) but also the surrounding area..

 

The best way to measure temperatures of surfaces is just to use a contact thermometer (thermocouple or PT100 probe)..

[grg]

Just adding this for a bit of historical context... This was running v1.1 of the OPi Zero board. I had a little script running to create some load. Also, this was inside a pretty tight enclosure with no heat spreader or heatsink.

[grg]

Same microSD card, v1.4 board, pushing the CPU a little harder (maybe even twice as hard).
 

Nothing else was done to help it run cooler.

 

[guidol]

On 26.7.2017 at 8:33 AM, ykchavan said:

This is reply I got from Shenzhen Xunlong Software CO.,Limited aliexpress store

It is normal at 45-80degrees. Cause the Zero is a tiny baord.
And we have set the dynamic frequency configuration on our offical images. This can help decrease the temp.
It will be lower under linux image than the Android image.

 

Today I did get my Orange Pi Zero (V1.4?) with Audio/USB-PCB and case.
My first try was armbian  with legacy kernel ( https://dl.armbian.com/orangepizero/Ubuntu_xenial_default.7z ) but the OPi Zero keeps getting very warm.
So I did try a "official" image from http://www.orangepi.org/downloadresources/ :
Raspbian Server - updated: 2017-05-05 
which links to http://www.orangepi.org/downloadresources/orangepizero/2017-05-05/orangepizero_e7c74a532b47c34968b5098.html
Original release date was 2016-11-24 with version 0.8.0 ( Raspbian_server_For_zero_H2+_V0_1 )

 

Measured with my fingers the OPi Zero runs cooler at they stated for "official Images".

 

Does the mainline kernel include these dynamic frequency configuration? (legacy doesnt seems to or not for v1.4?)
I used the legacy kernel version because of the label stable.

 

 

[chwe]

3 hours ago, guidol said:

Measured with my fingers the OPi Zero runs cooler at they stated for "official Images".

 

 

Just tested your 'official image'. Kernel is horrible outdated (uname -r : 3.4.39), I didn't had the feeling that it's colder so I looked into the two thermal zones:

cat /sys/devices/virtual/thermal/thermal_zone1/temp
cat /sys/devices/virtual/thermal/thermal_zone0/temp

Temps are around 67-64°C idle (armbian idle: 58-60°C), since sunxi-tools aren't installed (bin2fex) and apt-get install git failed, I lost the interest to investigate this image any further. 

[guidol]

16 hours ago, chwe said:

 

Just tested your 'official image'. Kernel is horrible outdated (uname -r : 3.4.39)

Temps are around 67-64°C idle (armbian idle: 58-60°C)

when you say armbian - do you use the unstable mainline version or the stable legacy?

mainline has - as I read - no wifi-support for the onboard chip. any other drawback?

as I use ethernet no-wifi wouldnt be a problem.

 

[EDIT] I installed the mainline to another MicroSD Card, but as written in the follwing thread

 

there is no audio-driver available, because normally the OPi Zero has no audio-expansion board
So I will try mainline later again - and will now use the horrible outdated kernel, because I want audio-support.

[guidol]

Today I checked the legacy kernel with a known good Sandisk Ultra Class 10 card, but legacy seems to be hotter (dont know why):

 

Linux orangepizero 3.4.113-sun8i #18 SMP PREEMPT Thu Jun 15 02:16:06 CEST 2017 armv7l armv7l armv7l GNU/Linux

using ethernet (not wifi)

 

x 00:33:29:  240MHz  0.08   4%   1%   2%   0%   0%   0%   71°  x
x 00:33:35:  240MHz  0.07   4%   1%   2%   0%   0%   0%   70°  x
x 00:33:40:  240MHz  0.07   4%   1%   2%   0%   0%   0%   71°  x
x 00:33:45:  240MHz  0.06   4%   1%   2%   0%   0%   0%   70°  x
x 00:33:50:  240MHz  0.06   4%   1%   2%   0%   0%   0%   71°  x
x 00:33:55:  240MHz  0.05   4%   1%   2%   0%   0%   0%   70°  x
x 00:34:01:  240MHz  0.05   4%   1%   2%   0%   0%   0%   71°  x
x 00:34:06:  240MHz  0.04   4%   1%   2%   0%   0%   0%   70°  x
x 00:34:11:  240MHz  0.04   4%   1%   2%   0%   0%   0%   70°  x
x 00:34:16:  240MHz  0.04   4%   1%   2%   0%   0%   0%   71°  x
x Time        CPU    load %cpu %sys %usr %nice %io %irq   CPU  x
x 00:34:22:  240MHz  0.03   4%   1%   2%   0%   0%   0%   71°  x
x 00:34:27:  648MHz  0.03   4%   1%   2%   0%   0%   0%   71°  x

 

root@orangepizero:~# cat /sys/devices/virtual/thermal/thermal_zone0/temp
70
root@orangepizero:~# cat /sys/devices/virtual/thermal/thermal_zone1/temp
70
 

and with a small metal cooler stand under the OPi Zero:

root@orangepizero:~# cat /sys/devices/virtual/thermal/thermal_zone0/temp
65
root@orangepizero:~# cat /sys/devices/virtual/thermal/thermal_zone1/temp
65
 

 

[bozden]

These figures are consistent with my results in the original post. Even 3-4 degrees hotter, but that may depend on the environment...

[chwe]

15 hours ago, guidol said:

when you say armbian - do you use the unstable mainline version or the stable legacy?

legacy

 

For everything inside a case I can't give you an answer, my OPi laying somewhere around. But as bozen mentioned, this boards aren't running stable inside a case and I think he tested it more than once. At the moment I wouldn't recommend this board for high load use cases. Seems that rev. 1.4 is somehow a design failure. I have one running since ~20d stable (without case, last reboot was due to updates not crash, legacy) with node-red on it over ETH without problems. You can try with a heat sink and set max_freq with h3consumption to 912MHz (this should aviod feeding the CPU with 1.3V) but I'm not sure if this is enough to avoid overheating.