> I read on many different sites that AMD’s current CPUs beat Intel’s CPUs in terms of performance per watt. We can better achieve goals 2 and 3 (low noise and

> I read on many different sites that AMD’s current CPUs beat Intel’s CPUs in terms of performance per watt. We can better achieve goals 2 and 3 (low noise and low power usage) by using fewer watts, so we’ll pick an AMD CPU and mainboard for this build. Unfortunately many reviews are very misleading here. Zen 2/3 CPUs have good performance per Watt, that's true. But for a machine like this, which will be mostly idle, this is not the interesting metric and Zen 2/3 systems show that you can combine good perf/W with poor idle power consumption (which is not true for their monolithic APU brethren, which are used in laptops). One of the biggest idle power hogs for these is the IO die, so make sure that XMP is disabled and the memory uses one of the slow JEDEC timings. This should be fine for a router. Check that the SoC/NB voltage is set to 1 V or less. Some boards set this higher. In the AMD CBS section of the board firmware there should be an item "SoC OC Mode" somewhere. Disable it. Some boards allow you to set a new PPT (package power target), but it's worth pointing out that values which are too low will make the CPU very slow because it essentially forces all cores to very low power states in order to meet the PPT since the CPU can't influence the baseline power (due to fabric and I/O die). The upside of using a reasonable PPT of e.g. 50-60 W is that you reduce power consumption if some errant task hogs the CPU. These settings make a big difference, but only if the CPU is really idle. Even fairly light loads (e.g. on a desktop, moving the mouse on the background) has everything rev up. In deep idle (nothing running at all, no user interaction on a desktop) you might get a Zen 2/3 CPU down to around 20 Watts, but as soon as anything is happening at all we're straight back to the 40-70 W region. Using an Intel system for this would have likely saved 10-20 W.