Run absolute minimal processes ( 5-55%? over 100% time )
Some devices set low frequency scaling ( with upper performance hit )
Put simple, running the absolute minimum services and lower scaling might get you to ( very roughly ) perma 35-55% savings maximum...
Which is kinda significant over two years... ( 100-200AUD$ / 50-120$US worth? ) but in the context of a whole household it's not that much...
Specifically, as you note: some devices either by poor design, power supply inefficiency or other... just chew a whole lot more than others ( thanks for posting )... but often these big consumers offer better baseline performance...
With these cases... do as you did, and;
compare with a power meter
eliminate/swap the power supply
try the steps above and try to minimise / identify the hogs
In 30%+ of cases it will be power supply... most of the rest old/inefficient design ( not much you can do about that )... and a handful chew more for a reason... but you would see / be able to tell the performance difference fairly easily...
Diluted to the base questions;
Is this design / power supply inefficient / efficient?
Does this devices efficiencies come at a performance hit?
As a travel router i'd love a MT300A or an all in one SoC... but for my home, the more dedicated chips ( more power ) within the router the better IMHO... a little more pricey... alot more robust...
In general, the higher end your router is, the more power it will require - be it for providing power to USB devices, or to drive 802.11ac or 802.11ax wireless (especially at full performance). Top end routers can easily chug ~20 watts idle.
Acknowledged. But the difference between the rpi4b and the wrt1900acs is remarkable. The rpi4b is a quad core scaling between 600 and 1500MHz, the wrt1900acs a duo core with a fixed frequency of 1600 MHz.
The power consumption of a wrt1900acs pays in 2 years the price of a NanoPi R1, which I ordered today.
The costs of electric power will definitely increase in the future. Therefore I pay more attention to the power consumption of my routers. They are running 24/7 for VoIP and mostly idling.
You're comparing apples with oranges here. A single connected (doesn't even need to actively transmit data) 1 GBit/s ethernet port (100 MBit/s needs less) requires between 1.0 and 1.2 watts, your WRT1900AC has five of those, the RPi4 a single one (and by using a switch, you're merely externalizing the power consumption). While wireless needs less power than ethernet beyond 1 GBit/s, it's also a considerable part of the calculation.
Adding a related thought: power consumption=heat. I am less concerned with electricity cost, and more concerned with whether I can fry an egg on the router.
My home has a small enclosed metal telecom junction box for the home ethernet lines with an electrical outlet built into it. The box is recessed into a closet wall. I sleep better at night knowing the EdgeRouter X enclosed in this box generates very little waste heat.
Modems are another matter. I find they are either based on yesterday's technology and run hot, or they are part of a giant "ISP Gateway" device that is too big for my enclosure.
The 9.5 W of the wrt1900acs result from 2 gigabit interfaces connected (1 wan, 1 lan). The one gigabit interface of the rpi4b combines 4 vlans connected via an 8-port switch (1.5 W). The single rpi4b interface had more traffic than the 2 interfaces of the wrt1900acs.
If you compare the archer C7 and the wrt1900acs: both have 5 gigabit interfaces, but the archer uses only 3.5 W (with all interfaces connected to active devices).
If your numbers are correct, why does the archer not use more than 3.5 W with all interfaces connected to active devices?
Why to not buy the Nanopi R2S it's nearly the same price.
No wifi, but dual gigabit (941 Mbps both in/out).
FriendlyWrt with docker.
Annd following this 2 video full load it consume betwen 2.6 & 3w.
it will cost you in the worst case (full load 24H/7) 7,9 euro year