1. Test procedure
  2. Research hypothesis and bias
  3. Radio tests
    1. 100% brightness, no modules
      1. no radios
      2. radios on, +0.03W min, +-0W average, +0.01W stdev
    2. minimum brightness, no modules
      1. no radios
      2. radios: +0.29W min, +0.25W average, +0.25W stdev
    3. screen off, no modules
      1. no radios
      2. radios: +0.10W min, +0.25W average, +0.04W stdev
    4. screen off, 1 USB-C
      1. no radios
      2. radios: +0.18W min, +0.30W average, +0.34W stdev
    5. screen off, 2 USB-C
      1. no radios
      2. radios: +0.15W min, +0.44W average, +0.63W stdev
    6. screen off, 1 USB-A
      1. no radios
      2. radios: +0.13W min, +0.14W average, +0.05W stdev
    7. Conclusion: radios take 100mW+
  4. no radios tests
    1. 100% brightness, no modules
    2. minimum brightness, no modules
    3. screen off
      1. no modules
      2. conclusion: screen power usage between 70mW to 2.55W
      3. 1 USB-C
      4. 2 USB-C
      5. conclusion: USB-C cards take little to no power
      6. 1 USB-A
      7. 2 USB-A
      8. 3 USB-A
      9. conclusion: USB-A cards may use 10mW or negligible
      10. setting baseline at 2.10W
      11. 1TB SSD: 1.5-2W
      12. 1 MicroSD: 1.6W-6W!!
      13. 1 HDMI: ~500mW
      14. 2 HDMI
      15. 1 DisplayPort: 300mW
      16. no module again

Test procedure

  1. raise DPMS timeout to avoid screen saver skewing results:

    xset 600 3
    
  2. prepare the test (disable/enable radios, plug in module, etc)

  3. set desired brightness, run test, reset brightness

    sudo true; xbacklight -set 0 ; sudo powerstat ; xbacklight -set 100
    

We run sudo before changing the backlight to make sure the sudo token is still recorded and the second sudo won't prompt us while the screen is blank.

Note that those tests were performed in full multi-user mode, in a graphical Xorg session. Further tests performed in single-user mode showed that single user mode saves about 300-400mW of power from the full session.

Other folks have different test procedures. Hrishi Olickel worked on a method based on GNOME battery bench which is interesting for real-life, actual work loads. The tests here are only idle workloads, which have limited applications for real life situations, but do show a good baseline for power consumption.

See also the powerstat-306 page for results after the 3.06 BIOS beta release and powerstat-wayland page for results done on wayland (instead of Xorg, here).

Research hypothesis and bias

My bias is to interpret results based on the "minimum": that's the ground, base noise that a new module (for example) would introduce.

The powerstat tool is used here. It does, by default, 30 samples every 10 seconds (so for 5 minutes) with a 3 minute "settle down" period before. That settle-down period ensures the machine spins back down to idle so there's less be variation based on CPU usage or frequency scaling.

If there's a result that's really out of whack, it will be rerun to see if we can reproduce.

I originally forgot to turn off radios. I was assuming radios would take power and introduce variability, so all tests were reran with radios turned off, basically with the equivalent of rfkill block all, but done with the handy fn-F10 keybinding which puts the laptop in "airplane mode".

The hypothesis is that the screen, WiFi, and some modules take power. Specifically, I heard that USB-C and DisplayPort modules do not take power while all the others take some power, with the SSD and MicroSD cards taking up to 1 to 5W.

Radio tests

Those tests were specifically done to see if turning on Bluetooth (not connected to anything) and WiFi (connected to a WPA access point) were taking power.

The "radios on" tests here were actually done as part of the normal batch (below) and were retroactively added back here.

100% brightness, no modules

no radios

This is the same test result as the "100% brightness, no radios, no modules" test below.

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   0.0   0.1  99.7   0.0  1.0 1102.5  288.9 85.9 21.7 86.4   4.70
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1098.1  287.8 83.5 19.2 83.9   4.70
  StdDev   0.0   0.0   0.0   0.1   0.0  0.2   99.9   26.8 25.5 15.2 26.6   0.06
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.3   0.0  1.0  919.7  250.9 73.0 16.0 73.0   4.61
 Maximum   0.2   0.2   0.2  99.8   0.1  2.0 1307.9  366.7 196.0 87.0 195.0   4.87
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   4.70 Watts on average with standard deviation 0.06

radios on, +0.03W min, +-0W average, +0.01W stdev

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   0.0   0.1  99.7   0.0  1.1 1073.7  276.0 83.3 20.5 84.3   4.70
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1067.2  275.0 81.8 18.9 82.7   4.70
  StdDev   0.0   0.0   0.0   0.1   0.0  0.4  117.1   23.2 20.3 12.3 20.9   0.07
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.5   0.0  1.0  863.2  233.8 72.0 16.0 72.0   4.58
 Maximum   0.2   0.2   0.2  99.8   0.1  3.0 1261.5  314.4 186.0 80.0 185.0   4.88
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   4.70 Watts on average with standard deviation 0.07

minimum brightness, no modules

no radios

This is the same test result as the "minimum brightness, no radios, no modules" test below.

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   0.0   0.1  99.7   0.0  1.1 1036.1  275.0 82.5 20.5 82.6   2.24
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1031.4  274.0 81.0 18.8 81.0   2.24
  StdDev   0.0   0.0   0.0   0.1   0.0  0.4  100.4   23.2 20.6 12.7 20.5   0.05
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.5   0.0  1.0  875.4  241.7 72.0 16.0 72.0   2.13
 Maximum   0.2   0.2   0.2  99.8   0.1  3.0 1317.4  342.2 186.0 80.0 186.0   2.35
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.24 Watts on average with standard deviation 0.05

radios: +0.29W min, +0.25W average, +0.25W stdev

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.2   0.0   0.1  99.6   0.0  1.0 1540.7  361.6 82.8 18.6 83.1   2.49
 GeoMean   0.2   0.0   0.1  99.6   0.0  1.0 1527.3  351.4 82.0 17.8 82.3   2.48
  StdDev   0.0   0.1   0.0   0.1   0.0  0.2  220.0  103.1 14.4  7.4 13.9   0.30
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.2   0.0  1.0 1303.9  282.0 75.0 16.0 76.0   2.30
 Maximum   0.3   0.3   0.3  99.7   0.1  2.0 2219.0  742.5 158.0 56.0 156.0   3.71
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.49 Watts on average with standard deviation 0.30

screen off, no modules

no radios

This is the same as the "screen off, no radios, no modules" test below.

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.2   0.0   0.1  99.7   0.0  1.0 1116.0  331.7 82.9 19.7 82.7   2.24
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1100.8  305.4 81.8 18.5 81.7   2.22
  StdDev   0.1   0.0   0.0   0.1   0.0  0.0  208.8  225.2 15.4  9.3 15.1   0.33
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.1   0.0  1.0  849.8  231.8 72.0 16.0 72.0   2.06
 Maximum   0.6   0.2   0.2  99.8   0.1  1.0 2038.4 1535.4 151.0 56.0 151.0   3.81
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.24 Watts on average with standard deviation 0.33

radios: +0.10W min, +0.25W average, +0.04W stdev

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.2   0.0   0.1  99.6   0.1  1.0 1577.7  377.9 83.5 19.6 83.9   2.47
 GeoMean   0.2   0.0   0.1  99.6   0.0  1.0 1561.8  367.3 82.5 18.5 82.9   2.44
  StdDev   0.0   0.1   0.0   0.1   0.0  0.2  235.6   99.5 15.6  8.9 15.5   0.37
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.2   0.0  1.0 1308.0  287.6 73.0 16.0 74.0   2.16
 Maximum   0.3   0.3   0.3  99.8   0.1  2.0 2201.1  681.6 157.0 56.0 156.0   3.56
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.47 Watts on average with standard deviation 0.37

screen off, 1 USB-C

no radios

This is the same test result as the "screen off, no radios, 1 USB-C" test below.

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   0.0   0.1  99.7   0.0  1.0 1044.4  271.7 81.2 19.6 81.6   2.09
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1037.3  270.8 80.2 18.4 80.6   2.09
  StdDev   0.0   0.0   0.0   0.1   0.0  0.0  126.1   22.7 15.5  9.2 15.4   0.07
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.4   0.0  1.0  813.5  239.0 70.0 16.0 71.0   1.96
 Maximum   0.2   0.2   0.2  99.8   0.1  1.0 1460.2  349.1 153.0 56.0 153.0   2.23
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.09 Watts on average with standard deviation 0.07

radios: +0.18W min, +0.30W average, +0.34W stdev

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.2   0.0   0.1  99.6   0.0  1.0 1504.4  356.2 80.9 18.4 81.0   2.39
 GeoMean   0.2   0.0   0.1  99.6   0.0  1.0 1485.4  344.7 80.5 17.7 80.6   2.36
  StdDev   0.0   0.1   0.1   0.2   0.0  0.2  273.3  113.8  8.9  6.9  9.0   0.41
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  98.9   0.0  1.0 1237.9  276.5 75.0 16.0 75.0   2.14
 Maximum   0.2   0.4   0.4  99.8   0.2  2.0 2549.2  797.7 125.0 53.0 126.0   3.80
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.39 Watts on average with standard deviation 0.41

screen off, 2 USB-C

no radios

This is the same test result as the "screen off, no radios, 2 USB-C" test below.

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   0.0   0.1  99.7   0.0  1.1 1100.6  282.5 86.9 21.7 86.8   2.09
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.1 1096.0  281.3 84.7 19.6 84.7   2.09
  StdDev   0.0   0.0   0.0   0.1   0.0  0.4  102.6   25.9 24.1 13.5 23.9   0.05
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.5   0.0  1.0  900.8  246.1 73.0 16.0 73.0   1.99
 Maximum   0.2   0.1   0.2  99.8   0.1  3.0 1335.8  365.8 188.0 80.0 187.0   2.19
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.09 Watts on average with standard deviation 0.05

radios: +0.15W min, +0.44W average, +0.63W stdev

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.2   0.0   0.1  99.5   0.0  1.1 1608.6  388.2 87.4 20.5 87.0   2.53
 GeoMean   0.2   0.0   0.1  99.5   0.0  1.0 1568.0  364.4 84.5 18.8 84.3   2.46
  StdDev   0.4   0.1   0.1   0.5   0.0  0.2  421.7  176.0 32.5 12.6 30.5   0.68
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  96.9   0.0  1.0 1127.3  265.6 72.0 16.0 72.0   2.14
 Maximum   2.2   0.4   0.6  99.8   0.1  2.0 3366.0 1150.2 258.0 83.0 247.0   4.56
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.53 Watts on average with standard deviation 0.68

screen off, 1 USB-A

no radios

This is the same test result as the "screen off, no radios, 1 USB-A" test below.

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   0.0   0.1  99.7   0.0  1.0 1052.7  279.1 81.6 19.6 81.6   2.11
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1049.1  278.3 80.5 18.4 80.6   2.11
  StdDev   0.0   0.0   0.0   0.1   0.0  0.0   88.1   22.5 16.0  9.2 15.6   0.05
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.5   0.0  1.0  881.0  250.4 72.0 16.0 72.0   2.00
 Maximum   0.2   0.2   0.2  99.8   0.1  1.0 1290.1  351.0 158.0 56.0 156.0   2.23
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.11 Watts on average with standard deviation 0.05

radios: +0.13W min, +0.14W average, +0.05W stdev

Conclusion: radios take 100mW+

Overall, radios seem to add a minimum of 0.10W (or 100mW) of power usage to the tests. That's minimal, of course, but it does significantly increase standard deviation, by about half a watt.

One theory is that an active network connection would trigger network traffic which in itself means more CPU usage, but it also means graphical updates as the status bar shows traffic measurements.

So it's good that we redid those tests.

no radios tests

100% brightness, no modules

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   0.0   0.1  99.7   0.0  1.0 1102.5  288.9 85.9 21.7 86.4   4.70
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1098.1  287.8 83.5 19.2 83.9   4.70
  StdDev   0.0   0.0   0.0   0.1   0.0  0.2   99.9   26.8 25.5 15.2 26.6   0.06
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.3   0.0  1.0  919.7  250.9 73.0 16.0 73.0   4.61
 Maximum   0.2   0.2   0.2  99.8   0.1  2.0 1307.9  366.7 196.0 87.0 195.0   4.87
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   4.70 Watts on average with standard deviation 0.06

minimum brightness, no modules

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   0.0   0.1  99.7   0.0  1.1 1036.1  275.0 82.5 20.5 82.6   2.24
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1031.4  274.0 81.0 18.8 81.0   2.24
  StdDev   0.0   0.0   0.0   0.1   0.0  0.4  100.4   23.2 20.6 12.7 20.5   0.05
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.5   0.0  1.0  875.4  241.7 72.0 16.0 72.0   2.13
 Maximum   0.2   0.2   0.2  99.8   0.1  3.0 1317.4  342.2 186.0 80.0 186.0   2.35
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.24 Watts on average with standard deviation 0.05

screen off

no modules

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.2   0.0   0.1  99.7   0.0  1.0 1116.0  331.7 82.9 19.7 82.7   2.24
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1100.8  305.4 81.8 18.5 81.7   2.22
  StdDev   0.1   0.0   0.0   0.1   0.0  0.0  208.8  225.2 15.4  9.3 15.1   0.33
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.1   0.0  1.0  849.8  231.8 72.0 16.0 72.0   2.06
 Maximum   0.6   0.2   0.2  99.8   0.1  1.0 2038.4 1535.4 151.0 56.0 151.0   3.81
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.24 Watts on average with standard deviation 0.33

conclusion: screen power usage between 70mW to 2.55W

The monitor seems surprisingly effectively, especially at the minimum brightness, which is almost indistinguishable from the "screen off" power usage.

We need to look at the "minimum" stats to see a different: there we see that the minimum brightness might take 70mW (2.13W-2.06W), while the full brightness takes a more solid 2.55W.

One question that remains is: why is the above result lower than the USB-C result below? Normally, it should be the same or lower... Note, however, that we have higher standard deviation above, so it could be outliers messing with the numbers.

1 USB-C

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   0.0   0.1  99.7   0.0  1.0 1044.4  271.7 81.2 19.6 81.6   2.09
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1037.3  270.8 80.2 18.4 80.6   2.09
  StdDev   0.0   0.0   0.0   0.1   0.0  0.0  126.1   22.7 15.5  9.2 15.4   0.07
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.4   0.0  1.0  813.5  239.0 70.0 16.0 71.0   1.96
 Maximum   0.2   0.2   0.2  99.8   0.1  1.0 1460.2  349.1 153.0 56.0 153.0   2.23
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.09 Watts on average with standard deviation 0.07

2 USB-C

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   0.0   0.1  99.7   0.0  1.1 1100.6  282.5 86.9 21.7 86.8   2.09
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.1 1096.0  281.3 84.7 19.6 84.7   2.09
  StdDev   0.0   0.0   0.0   0.1   0.0  0.4  102.6   25.9 24.1 13.5 23.9   0.05
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.5   0.0  1.0  900.8  246.1 73.0 16.0 73.0   1.99
 Maximum   0.2   0.1   0.2  99.8   0.1  3.0 1335.8  365.8 188.0 80.0 187.0   2.19
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.09 Watts on average with standard deviation 0.05

conclusion: USB-C cards take little to no power

It looks like the theory that USB-C cards are passive is valid. From the above, we may have a 30mW power drain between one and two USB-C cards, but that is within the standard deviation (50mW).

1 USB-A

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   0.0   0.1  99.7   0.0  1.0 1052.7  279.1 81.6 19.6 81.6   2.11
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1049.1  278.3 80.5 18.4 80.6   2.11
  StdDev   0.0   0.0   0.0   0.1   0.0  0.0   88.1   22.5 16.0  9.2 15.6   0.05
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.5   0.0  1.0  881.0  250.4 72.0 16.0 72.0   2.00
 Maximum   0.2   0.2   0.2  99.8   0.1  1.0 1290.1  351.0 158.0 56.0 156.0   2.23
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.11 Watts on average with standard deviation 0.05

2 USB-A

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   0.0   0.1  99.7   0.0  1.0 1073.5  289.0 80.1 18.7 80.0   2.12
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1069.3  288.1 79.8 18.0 79.6   2.12
  StdDev   0.0   0.0   0.0   0.1   0.0  0.2   95.4   23.6  7.9  6.4  8.0   0.05
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.4   0.0  1.0  869.8  246.3 72.0 16.0 72.0   2.01
 Maximum   0.2   0.2   0.2  99.8   0.1  2.0 1270.1  356.5 116.0 49.0 117.0   2.25
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.12 Watts on average with standard deviation 0.05

3 USB-A

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   0.0   0.1  99.7   0.0  1.0 1074.0  286.1 81.7 19.8 82.0   2.13
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1067.8  284.9 80.6 18.6 80.9   2.13
  StdDev   0.0   0.0   0.0   0.1   0.0  0.2  119.1   26.5 15.8  9.2 15.8   0.06
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.5   0.0  1.0  910.8  240.9 73.0 16.0 73.0   2.03
 Maximum   0.2   0.2   0.2  99.8   0.1  2.0 1338.3  347.9 156.0 56.0 156.0   2.27
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.13 Watts on average with standard deviation 0.06

conclusion: USB-A cards may use 10mW or negligible

This looks like a fairly reliable pattern: adding a USB-A card adds 0.01W of power usage, both in minimum and average power consumption. That is within the standard deviation however, which doesn't really change by adding more modules, so it's likely the USB-A ports use very little power as well.

setting baseline at 2.10W

It's hard to establish a baseline power usage for those tests, because they actually fluctuate a lot, even between similar test environments.

For example, the first few tests in this "screen off" section seem to show USB-C cards use less power than no cards at all. And further powerstat runs are having trouble reproducing the 1.96W minimum encountered above.

Let's set a 2.10W baseline and call it that.

1TB SSD: 1.5-2W

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   0.0   0.1  99.7   0.0  1.0 1127.2  283.4 87.5 21.7 87.9   3.89
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1123.0  282.6 85.4 19.6 85.8   3.89
  StdDev   0.0   0.0   0.0   0.1   0.0  0.2  100.6   22.0 24.0 13.5 23.9   0.18
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.5   0.0  1.0  944.1  255.1 75.0 16.0 75.0   3.75
 Maximum   0.2   0.1   0.2  99.8   0.1  2.0 1423.3  356.3 188.0 80.0 188.0   4.49
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   3.89 Watts on average with standard deviation 0.18

This is at least 1.65W above the baseline, with 1.79W on average and a max 2W power usage.

This is just with the SSD plugged in, detected by the kernel, but not mounted or active in any way.

1 MicroSD: 1.6W-6W!!

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   1.3   0.1  98.4   0.0  1.3 1169.5  375.8 85.3 19.5 86.0   5.00
 GeoMean   0.1   0.0   0.1  98.4   0.0  1.2 1165.5  351.3 84.0 18.1 84.6   4.69
  StdDev   0.0   2.5   0.0   2.5   0.0  0.5  102.5  155.2 19.4 11.6 19.5   1.99
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  93.5   0.0  1.0 1076.2  264.7 78.0 16.0 79.0   3.65
 Maximum   0.2   6.3   0.1  99.8   0.1  3.0 1570.5  719.4 188.0 80.0 188.0   8.71
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   5.00 Watts on average with standard deviation 1.99

There is also a 1.6W power drain minimum here, but look at the maximum: that's over 6 watts used! On average we have a 3 watt power drain here, with a very large standard deviation (1.99)!

This is the most power-hungry module that I have tested, by far.

1 HDMI: ~500mW

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.2   0.0   0.1  99.7   0.0  1.0 1151.7  287.7 86.2 20.5 86.4   2.86
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1148.1  286.2 84.6 18.6 84.8   2.86
  StdDev   0.0   0.0   0.0   0.1   0.0  0.0   93.0   29.5 21.6 13.8 21.3   0.10
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.4   0.0  1.0 1036.3  249.6 78.0 16.0 78.0   2.74
 Maximum   0.2   0.2   0.2  99.8   0.1  1.0 1376.2  365.8 195.0 87.0 194.0   3.18
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.86 Watts on average with standard deviation 0.10

This seems to confirm the hypothesis that the HDMI modules are not passive. But let's try with two to be certain.

2 HDMI

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.2   0.0   0.1  99.6   0.0  1.0 1161.9  306.0 89.6 21.5 89.5   3.30
 GeoMean   0.2   0.0   0.1  99.6   0.0  1.0 1141.3  295.9 84.1 18.3 84.3   3.29
  StdDev   0.4   0.0   0.1   0.5   0.0  0.0  279.7  111.4 51.4 23.4 49.6   0.29
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  97.2   0.0  1.0  963.5  248.3 75.0 16.0 75.0   3.12
 Maximum   2.3   0.0   0.5  99.8   0.1  1.0 2602.5  892.2 366.0 147.0 356.0   4.60
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   3.30 Watts on average with standard deviation 0.29

Here we add another 440mW on average, 380mW on minimum, and an extra 190mW of standard deviation.

Clearly the HDMI cards do use some power.

1 DisplayPort: 300mW

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   0.0   0.1  99.7   0.0  1.0 1120.7  282.4 84.0 19.6 84.7   2.49
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1117.2  281.6 83.1 18.4 83.7   2.49
  StdDev   0.0   0.0   0.0   0.1   0.0  0.0   91.0   22.8 15.0  9.2 15.3   0.06
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.4   0.0  1.0 1018.2  260.2 77.0 16.0 77.0   2.41
 Maximum   0.2   0.2   0.2  99.8   0.1  1.0 1354.3  355.8 154.0 56.0 156.0   2.70
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.49 Watts on average with standard deviation 0.06

Here, the result is a little more ambiguous. We're pretty close to the baseline here, 300mW above the 2.1W baseline we declared above. Still, it's non-negligible so I'd tend to call this card non-passive because of that power usage.

no module again

  Time    User  Nice   Sys  Idle    IO  Run Ctxt/s  IRQ/s Fork Exec Exit  Watts
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Average   0.1   0.0   0.1  99.7   0.0  1.0 1122.2  282.4 83.7 19.6 83.8   2.16
 GeoMean   0.1   0.0   0.1  99.7   0.0  1.0 1117.8  281.5 82.7 18.4 82.8   2.16
  StdDev   0.0   0.0   0.0   0.1   0.0  0.2  101.3   23.3 15.7  9.2 15.4   0.06
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
 Minimum   0.1   0.0   0.1  99.4   0.0  1.0  969.0  248.4 74.0 16.0 74.0   2.06
 Maximum   0.2   0.2   0.2  99.8   0.1  2.0 1308.8  342.4 158.0 56.0 156.0   2.32
-------- ----- ----- ----- ----- ----- ---- ------ ------ ---- ---- ---- ------
Summary:
System:   2.16 Watts on average with standard deviation 0.06

This should normally be the same as the first test in this section. But the average differs, by about 80mW. This is pretty close to the standard deviation so it could be considered part of the margin of error for those measurements.

What is confusing is this measurement and the related one above is that they are both higher than the USB-C tests, which both have a lower average (2.09W exactly) and minimums (1.96W and 1.99W).

So there are clearly some uncontrolled variables in my tests.

Created . Edited .