TK1 mainline power saving

Technical Support
, ,
1

Hi,
We have migrated our custom image from downstream to mainline kernel. Now we noticed that the system gets a lot hotter then before. The CPU governor is set to “ondemand”.
Are there any other possibilities to reduce the power consumption?
For example we do not use the GPU, so can we disable it?
Regards

2

Hi @qojote

Thanks for writing to the Toradex Community!

Could you share your .kernel config?

What is the Cpu’s frequency in the mainline kernel? What is the temperature?
How were the frequency’s and temperature’s values for your application using downstream kernel?

Best regards,
Jaski

3

Hi. The CPU frequency is autoamtically set to minimum as there is not much system load. How to measure the two different temperature zones in the mainline kernel? As you can see in the code below i just can readout hwmon0. Please find my Kernel config attached.

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq 
204000
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor 
ondemand
cat /sys/class/hwmon/hwmon0/temp1_input 
48125
cat /sys/class/hwmon/hwmon1/temp1_input 
cat: /sys/class/hwmon/hwmon1/temp1_input: Operation not supported
4

PS: I thought the maximum frequency for the TK1 was 2070000 Hz but according to “/sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq” it is 2116500 Hz?

5

Ultimately, it really depends on each chips silicon quality grading aka speedo.

6

How can i reduce the power consumption in respect to the heat development. I am wondering why the module is getting hotter in idle mode (lowest frequency, CPU usage < 5%) as the downstream kernel running my application in performance mode.

7

I believe that has to do with the GPU which does not support any automatic voltage and frequency scaling as of that mainline 4.14 based version.

8

I have already set the GPU frequency to the lowest:

cat /sys/kernel/debug/dri/129/pstate 
01: core 72 MHz *
02: core 108 MHz
03: core 180 MHz
04: core 252 MHz
05: core 324 MHz
06: core 396 MHz
07: core 468 MHz
08: core 540 MHz
09: core 612 MHz
0a: core 648 MHz
0b: core 684 MHz
0c: core 708 MHz
0d: core 756 MHz
0e: core 804 MHz
0f: core 852 MHz
DC: core 72 MHz
9

HI @qojote

If you already set the GPU to the lowest frequency, then nothing more can be done.

Are all CPU cores running at low frequency?

Best regards,
Jaski

10

Yes, all cores are at low frequency (ondemand) and the cpu load is marginal.

11

Could you check with Bsp 3.0b3 please? Thanks.

12

I have check the behaviour with BSP3.0b3:

root@apalis-tk1-mainline:~# uname -a
Linux apalis-tk1-mainline 4.14.160-3.0.3 #1 SMP PREEMPT Tue Dec 31 18:08:08 UTC 2019 armv7l GNU/Linux
root@apalis-tk1-mainline:~# cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq
204000
root@apalis-tk1-mainline:~# cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
ondemand
root@apalis-tk1-mainline:~# cat /sys/kernel/debug/dri/129/pstate
01: core 72 MHz *
02: core 108 MHz
03: core 180 MHz
04: core 252 MHz
05: core 324 MHz
06: core 396 MHz
07: core 468 MHz
08: core 540 MHz
09: core 612 MHz
0a: core 648 MHz
0b: core 684 MHz
0c: core 708 MHz
0d: core 756 MHz
0e: core 804 MHz
0f: core 852 MHz
DC: core 72 MHz
root@apalis-tk1-mainline:~# cat /sys/class/hwmon/hwmon0/temp1_input
43500
root@apalis-tk1-mainline:~#
13

I think you can mark this question as resolved. We will install an additional fan and i guess this will keep the whole device cool enough.

14

Perfect that the issue is solved. Thanks for the feedback.

Best regards,
Jaski