Discussion

Easiest way to compile 3.17/3.19 kernel

87 19404
Thanks.  Igor, I see your scipt spports the orange.  Does it work for the orange pi plus?  

igorpec  
Not yet.

Orange Plus is different hardware (Allwinner H3) - other u-boot and kernel (sources). The SDK was just released & not much thing work at the moment.

I am working on it.

tkaiser  
silentcreek replied at Tue May 12, 2015 09:57
OK, there are a couple of issues to reply to:

1) @tkaiser:

Thx for pointing that out. So it seems that most of the stuff outlined here in the forums regarding these hardware related topics gets for the birds within a few months

tkaiser  
Edited by tkaiser at Wed May 13, 2015 02:56
silentcreek replied at Tue May 12, 2015 09:57
This is also the temperature that the cpufreq driver (introduced in kernel 4.0) uses for thermal throttling.

For the second temperature sensor (in the PMU) there is a patch by Correntin Labbe, available on his site:
http://sunxi.montjoie.ovh/ (scroll to the bottom)


BTW: Have you tried out his patch? The values he reported as example ("SoC temperature: +40.6°C", PMU temperature +23.5°C) are weird. When I did some extensive testing half a year ago I noticed two simple facts:

  • internal chip temperatures are a few degrees above ambient temperature (sounds reasonable?)
  • when the board is idle the temperatures of both chips do not differ that much (exception: when they are at different locations, eg. one on the top and the other one on the bottom side of the PCB)


Given the values above his Cubieboard is in a room where the ambient temperature is a) below 18°C and b) the thermal situation of the AXP209 and the A20 4 centimeters away differs dramatically. And while it makes sense that the SoC's temperature is lower than the PMU's (heatsinks for the 19x19 mm A20 BGA are widely available while you have to take special precautions to apply a 6x6 mm heatsink on the PMU) the opposite is quite unlikely in this specific temperature range (exception: When the ambient temperature at that time was below 5° and his Cubieboard both utilised CPU and GPU at maximum -- then the values reported might be possible)

It seems with mainline kernel it happens again: thermal drivers/tools spit out numbers without meaning and people trust these numbers blindly. Please keep in mind the setting of the CHOP_TEMP_EN register in his first A20 temp readout and the wrong results produced by that: http://forum.lemaker.org/forum.php?mod=viewthread&tid=8137. Seems the very same stuff happens again right now with 4.0: The values reported have nothing to do with reality.

I will stay away from this crap for a few months and then have a look how things evolved.

I haven't tried this patch and I had my suspicions why this isn't mainlined. To be fair, most people won't need that anyway. And the temperature from the touchscreen controller that you can read in mainline seems reasonable.

One note about the PMU temperature though: From my experiments this depends on the actual power draw of the device. So, the more devices you have attached to your board, the higher the PMU temperature is even when there is no CPU load.

tkaiser  
Edited by tkaiser at Wed May 13, 2015 03:45
silentcreek replied at Wed May 13, 2015 03:10
And the temperature from the touchscreen controller that you can read in mainline seems reasonable.


To be confirmed

Since reading out correct values depends on the register contents of TP_CTRL1 and the current code is based on Corentin's with the wrong register settings I doubt that it will work correctly. Will give it a try in a few months and compare directly to previous results.

Regarding PMU's temp your observation is right: The more power the PMU has to supply the hotter it gets. And since the PMU contains overheating protection (shuts off immediately -- been there, experienced that) in my eyes it's more important to monitor this temperature than the SoC's (with 960 MHz now with kernel 4.0 max. you won't run into thermal troubles except when you use any of the thermally broken cases for Banana Pi/Pro -- I still wonder why Lenovator or any other case maker completely ignores the fact that everything that gets hot is on the lower side of the PCB -- therefore I will never buy a Pro but will get a BPi M1+ instead where some airflow is possible even in a small case)

BTW: Monitoring the PMU's temp is especially important on the Lamobo R1 (router board) since the A20 is on the upper side but the AXP209 on the lower side of the PCB and due to the BCM bridge chip and the RTL8192CU the device draws more power and the PMU gets way hotter. Thermal throttling based on the SoC's temperature won't help and the available enclosures are crap (thermally broken by design) as usual.

Thanks Igor.  I would like to follow your progress on the orange plus.  Is there a thread on the orange forum I should watch for updates?

Mark.

Edited by silentcreek at Wed May 13, 2015 13:57
tkaiser replied at Wed May 13, 2015 10:39
To be confirmed  

Since reading out correct values depends on the register contents of TP_CTRL1 ...


I tried it yesterday and it showed 40.1°C (mostly idling) which was at least in the range (+/-2°C) of the results I got with the older 3.4 kernels, so nothing that made me suspicious.

You have to log in before you can reply Login | Sign Up

Points Rules