Other

Lubuntu\'s Epiphany performances ?

35 1736
tkaiser  
hnsncxy replied at Mon Nov 2, 2015 19:21
We are optimizing the Linux Software for the temperature, and it may reduce around 8-10 degrees, a ...

Haha, where and how does this happen? I can't see anything on https://github.com/LeMaker/linux-actions/commits/s500-master
And what do you mean by 'optimizing' the 'Linux Software'? You define temperature tresholds to let the S500 throttle earlier? You prevent it from overheating with lower clock speeds? You change hardware related kernel code to make use of ARM sleep states (which only helps on idle systems or systems with moderate load but not in 'full load' situations when temperatures really matter)? Which sort of optimazation are we speaking of?

Also nice try to let people think, the S500 kernel stuff will ever be ready for mainline. You know that this will never happen because it would require a totally different development style and months/years of work. And the target market's quantity for SBCs is just a small fraction compared to the Android market where Actions Semi comes from and where releasing sources doesn't matter at all.

This sums it up regarding Actions Semi and 'Open Source' in the past: https://groups.google.com/d/msg/ ... diWSp4/aoqSBm0wju0J (and things won't change overnight)

Ok, thanks for your reply.
So I'll test sound and be able to publish my review.

Bye

Laurent

hnsncxy  
Edited by hnsncxy at Tue Nov 3, 2015 04:55
tkaiser replied at Tue Nov 3, 2015 01:49
Haha, where and how does this happen? I can't see anything on https://github.com/LeMaker/linux-act ...

We are optimizing, don't release the optimization code at present, and please wait in patient. we will release the newer version in this month.  If you are interested in the Android system, and you can test it and compare with the Linux system. ( the Android will have a lower temperature).  The test shows that Lemaker Guitar running Linux system will consume more current than Android system at no-load.  we will release more information when we have tested.

Comments

This sounds interesting, my DualCore A20 gets in idle, when standing, not flat on the ground 41° (21°C ambient). How low does Guitar go and your ambient Temp.?  Post time Nov 03, 2015 06:33

tkaiser  
Edited by tkaiser at Tue Nov 3, 2015 06:18
hnsncxy replied at Tue Nov 3, 2015 04:47
We are optimizing, don't release the optimization code at present, and please wait in patient. we w ...

Ok, based on what you wrote you're not 'optimising' but instead trying to get a clue what's going on.

My tests with the Guitar a while ago showed that the S500 unlike other SoCs I tested so far has a much higher consumption when clocked at higher rates when being idle than with lower (maybe this difference is due to the other SoCs being Cortex-A7 and the S500 Cortex-A9). On an Allwinner A20 for example it makes nearly no difference on an idle system whether the SoC's clock speed is at 960 MHz or 144 MHz. Both consumption and temperature remain nearly the same as long as the CPU is really idle: http://linux-sunxi.org/Cpufreq#The_.22performance.22_governor

With the S500 it makes a huge difference whether the system is idle at 1104 MHz or at 408 MHz. In my tests the SoC's temperature differed by 4°C and the PMU's by 3°C in such an idle scenario solely depending on the CPU clock speed.

Ambient temperature always directly influences the IC's internal temperatures. As well as the position and surrounding environment of the board does. Temperatures differ a few degress depending on board position and airflow possible. You can read all the details here: http://forum.armbian.com/index.p ... of-lemakers-guitar/

So given you changed the position of the board, measured Android in the morning and Linux around noon and used your weird cpufreq settings in Linux but not in Android (what's the purpose of using the ondemand governor and setting scaling_min_freq and scaling_max_freq to the same value as outlined here: http://wiki.lemaker.org/LeMaker_ ... cking_CPU_frequency -- seems you don't understand that for such cases the performance governor would be the better choice since ondemand is considered broken by design in the meantime: http://linux-sunxi.org/Cpufreq#Possible_adjustments ) it's obvious where 8°-10° C difference might originate from.

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

Points Rules