I'm investigating the thermal impact to the qrb5165 SOC when the chip is operating under stress. My rb5 right now has LE installed. I built the standard Linux tool stress-ng to load the chip. What I found is under heavy load presented by the stress-ng tool, the operating frequency of the chip, as obtained from /sys/devices/system/cpu/cpu*/cpufreq/{scaling_cur_frequency, cpuinfo_cur_freq} was not changing at all. This sounds too good to be true since I also obtained the power data from /sys/class/power_supply/usb/input_current_now and /sys/class/power_supply/usb/voltage_now, and I Do see the power scales down to 60%-80% of the idle power, when the stress tool was running.
What I did to the settings is that I set the scaling_governor to "performance" which will guide the chips to work at the maximum frequency. But even with this setting, I do expect the frequency to scale down by the chip itself as a protection to the hardware, which was also suggested by the power scale down.
So my question is, is the scaling_cur_frequency or cpuinfo_cur_freq reporting the frequency of the cores correctly? What am I missing here? If anyone can shed some light on this it will be greatly appreciated!
hello customer,
"scaling_cur_frequency" or "cpuinfo_cur_freq" represents the CPU frequency block set by the software, which can represent the current frequency of the cpu.
let us know if you need more information
Regards