Kill-A-Watt measurements
#1
Power measurements of kernel builds on X cores with make -jX. [Note, the base system is booted off an SD card with the kernel source code located on the eMMC card. Eventually, I will redo the measurements when booted straight from eMMC].

-j1: 10.0W +/- 0.5W (0 ... 07.0W with screen disabled)
-j2: 12.7W +/- 0.5W (0 ... 09.6W with screen disabled)
-j3: 13.7W +/- 0.5W (0 ... 10.2W with screen disabled)
-j4: 14.3W +/- 0.5W (0 ... 11.2W with screen disabled)
-j5: 15.0W /- 0.5W (0 ... 11.3W with screen disabled)
-j6: 15.0W /- 0.5W (0. .. 11.4W with screen disabled)

When the screen is turned on, power consumption is quite constant with a variance of +/- 0.5W. It's puzzling that occasionally no power is consumed for a few seconds when the screen is turned off. Maybe the battery is taking over during these short intervals?

The external power supply can provide up to 15W. Utilizing 4 cores, the system requires from time to time almost all that power and while utilizing 5 or 6 core, the system requires all the time all that power and probably occasionally even more. This helps explaining the sudden drops in battery charge from 100% to 91% on systems under load (there are are a few threads here on this board).


I also noticed weird things going on which are barely explainable [the actual timestamps are rough guesses. Numbers for the sake of numbers]:

t=00min. I stress the system with make -j5. Power consumption 15W
t=10min. I continue stressing the system with make -j5: Power consumption goes to 0W.
t=11min. I stop the make -j5. Power consumption remains at 0W.
t=20min. Power consumption goes to 13W.
t=40min. Over the last 20 min, the power consumption went slowly down from 13W eventually reaching 8.4W. 8.4W also the power consumption of a freshly booted system.


I gain the impression that at t=10min, the internal battery may have taken over and then from t=20 to 40, the power supply charged back the internal battery. I did not see any capacity drops on /sys/class/power_supply/cw2015-battery/charge, though. Maybe the charge dropped by less than 0.5% keeping the rounded value of 100?


Let's focus on display off (without load): When the display is off, there is an oscillation of 0W for 2 seconds and 5.4W for two seconds. Do here really main power and battery flip flop charging the Pinebook Pro?

Any comments? Can you guys do further power measurements?
  Reply
#2
I also have a Kill-a-watt, and a DrOk.
  Reply
#3
Quote:When the screen is turned on, power consumption is quite constant with a variance of +/- 0.5W. It's puzzling that occasionally no power is consumed for a few seconds when the screen is turned off. Maybe the battery is taking over during these short intervals?


Yeah, what happens here is, that the battery is charged, the battery voltage cutoff (~4.35V) is reached, the charger switches off, the battery gets discharged until it's under the hysteresis switching point and then the charger will draw current again. 

If you want to do real reliable power measurements I recommend opening up your PBP, unplugging the battery (very important!), and then plugging the small loose jumper cable together, which will allow you to power the Pinebook Pro from the wall supply alone. 

You can then measure the power consumption pretty reliably. 

I have used a clip on ampere meter around both wires (don't get fooled by the wire colors, both red and black are connected in paralell) to measure the current draw directly. With the kill-a-watt and power supply you'll probably add another 10-15%, but that's okay.  

About the power draw in the longer test: Have you ensured that the PBP is properly ventilated? The underside can get pretty hot and the RK3399 would limit it's maximum TDP (and thus power draw) to prevent overheating.

Here's a photo of my measurement setup: 
[Image: jbdfnsvmizx.jpg] 
You can see that the PBP is drawing ~ 1.18A @ 5V, running Manjaro with full screen brightness at idle CPU.
  Reply
#4
Cool setup. I have not used my ampere meter for these measurements. Right here I'm not concerned about precision and I prefer the convenience of a kill-a-watt. I can easily use it the same time I use the laptop.

(02-28-2020, 08:20 AM)manawyrm Wrote: Yeah, what happens here is, that the battery is charged, the battery voltage cutoff (~4.35V) is reached, the charger switches off, the battery gets discharged until it's under the hysteresis switching point and then the charger will draw current again. 

Why does the charger switch off? It's supposed to a) charge the battery as long as it's not fully charged while b) powering the laptop circuitry all the time.

(02-28-2020, 08:20 AM)manawyrm Wrote: About the power draw in the longer test: Have you ensured that the PBP is properly ventilated? The underside can get pretty hot and the RK3399 would limit it's maximum TDP (and thus power draw) to prevent overheating.

I see this behavior even with 1 cm of free space under the laptop. TDP does not explain the observed behavior, does it?


I see a lot of weird behavior once the system is running for a couple of hours ... preferably after the system got stressed for some time in the past.



Half off topic. I was grep'ing my way through the kernel source and learned on how to cap the max temperature to, say 55C.

Code:
manjaro: # echo 55000 > /sys/devices/virtual/thermal/thermal_zone0/trip_point_0_temp

The kernel is then throttling the cpu frequencies to reach that temperature. The default trip point comes from the device tree. Rockchip set it to 70C and someone from Pine64 re-set it to whopping 80C.

Here are current frequencies under load when temperature is caped to 55C

Code:
manjaro: # cat /sys/devices/system/cpu/cpu?/cpufreq/cpuinfo_cur_freq
1416000
1416000
1416000
1416000
1800000
1800000
  Reply


Possibly Related Threads…
Thread Author Replies Views Last Post
  Power measurements jpakkane 25 8,264 11-23-2019, 03:29 AM
Last Post: pfeerick

Forum Jump:


Users browsing this thread: 1 Guest(s)