CPU Temp with Heatsink and then fan.

phreon

When I first fired up my OP PC I noticed the cpu felt very hot, too hot to touch. I found in my junk box a heatsink from an old desktop motherboard gpu and was happy to see that it would fit easily over the cpu. I cleaned the old thermal paste off of it and added some new thermal paste and it holds the cpu temp at about 38 deg C. I also had a small 12v cpu fan in the junk box that ran silently at 5 volts and dropped it on top of the heatsink. I wired the fan to ground and 5v of the OP PC header. Under load, with Chrome and xterm running the cpu holds at 30C to 31 C.   I just leave it running 24/7 now.

To check the temp I use:

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

bronco

So you think, you're safe now? Could you please give cpufreq-ljt-stress-test a try as outlined here: http://linux-sunxi.org/Hardware_Reliability_Tests#CPU

And if that works reliably and without errors, please open up 3 other terminals and fire up 3 instances of cpuburn-a7 as well? And then see whether cpufreq-ljt-stress-test shows errors or not?

Thx in advance

phreon

bronco replied at 2015-11-19 22:33
So you think, you're safe now? Could you please give cpufreq-ljt-stress-test a try as outlined here: ...

Safe from what? It has been running 24/7 for 6 days now.  It holds around between 29C and 31C.
I am not sure what the value of an exteme stress test would be. I doubt that there is anything I will run that would ever stress the OP PC to the same level as the test you suggest. If you do my set up and run the tests let us know the results.
Thanks
P

bronco

phreon replied at 2015-11-19 23:16
I am not sure what the value of an exteme stress test would be.

What do you fear? That your overclocked system has stability issues and corrupts data under load?

I hope you live in the Northern Hemisphere? Then your setup (don't put relevant load on your SBC while clocking it insanely high) would make sense a bit: Using an SBC as additional heat source during winter months. But I believe that's not worth both wasting energy and risking data integrity and stability problems.

BTW: We use a couple of ARM boards as servers 24/7 since months/years. Without any problem and without fan. You can solve many problems without throwing additional hardware (in this case a heatsink + fan) on it.

reb100

Just as a reference.  Running a Orange pi PC at default clock with two python scripts running - reading DS18B20 temperature sensors and a remote access session.  Zone 0 read 64 degrees C and zone 1 reads 62 degrees C.  No heat sinks or fans.  Run time about 10 days.

bronco

reb100 replied at 2015-11-22 22:37
reading DS18B20 temperature sensors and a remote access session.  Zone 0 read 64 degrees C and zone 1 reads 62 degrees C

Insane high values. And yes, you name the problem. A SoC doing nothing heating up like hell. That's an indication that something's really going wrong

Sitheek

Did you connect fan to gpio pins?
I am thinking about somting like Cool&Quiet - PWM for cooling fan but have no idea whether there are timers, hardware interupts available in H3 or not ??
I must study documentation for H3 ...

phreon

Sitheek replied at 2015-11-23 01:42
Did you connect fan to gpio pins?
I am thinking about somting like Cool&Quiet - PWM for cooling fan ...

I connected to the board header at 5v and GND. Fan runs fine and CPU stays cool.

bronco

Good news: There's no need for a fan. Simply stop overclocking/overvolting and you're done: http://www.orangepi.org/orangepi ... 6&fromuid=29411

bronco

reb100 replied at 2015-11-22 22:37
Zone 0 read 64 degrees C and zone 1 reads 62 degrees C.  No heat sinks or fans.

Could be 15°C less if you stop overclocking/overvolting: http://www.orangepi.org/orangepi ... =6966&fromuid=29411