Bumping this. I have been testing z370 platform as well, asus prime z370-a ii, with a i5-8400. I bought it specifically because it has tons of slots since I'm obsessed with PCI-e passthrough. (btw, its annoying how AMD has more lanes but I can't find any motherboards with as many slots as this board. I include m.2 slots in pcie mode since they can be converted.)
With no drives using iGPU booting unraid off flash drive it idled at 27-28watts at the wall. This is also with 4sticks of ram 1.35v/3200C16. I should try disabling the onboard sound, I can't pass it through anyway. There's also an asmedia onboard USB controller that I'm not sure I can get to work in passthrough anyway (quick test last night was not promising) that I might try disabling. I honestly haven't played with the power settings much, I left them at defaults except getting the ram up to XMP.
One thing I noticed is it seems the iGPU is unbeatable on this for idle power.
Idle with dedicated card and iGPU enabled - Idle with IGPU only ~= Idle with dedicated and iGPU disabled. There's at most 1watt which is in the noise.
This doesn't gel with my ivybridge setup, where I somehow got better idle power by installing a PCI card. My guess is its a driver downclocking thing and either the version of unraid I'm running has better drivers or coffeelake handles downclocking in a different manner. Its also possible I just made some configuration screw up when I was setting that up.
I have a x1700x+x470 motherboard platform. I near as I can tell that is about 10-12watts worse idle than coffeelake with similar load out. And if doesn't have an iGPU so you're stuck burning 5 more watts if you want to see something on the screen. It kind of stinks but I believe so much of Ryzen idle power is spent keeping the infinity fabric going that Intel has an advantage at idle usage. Now, I also have to set Power Supply Current to Typical to keep it from locking up in unraid so that might result in an extra hit on Ryzen.
Ryzen is harder to get good idle numbers on btw, it bounces around in the 3-4w range for usage. This even goes back to that old FM2+ platform I used to own, that was also really noisy. My Intel tests also bounce in a much tighter band. Not really a problem (average idle should be all that matters) but makes quick tests harder.