It would have been nice to see power and performance comparison with Intel CPUs in the same price range, i.e. an Intel Celeron and a Pentium model from the Haswell generation.
5150/asrock am1h-itx dc in, ssd... idles at 16 watts. (the 5350 should idle at the same amount, but i only have the 5150.)
haswell g3240/msi h81i, 19 watts.
it seems that all of the haswell cpus at or below the i5-4440 level, idle at the same rate.
using a bronze level 300 tfx psu (included with the inwin 631t case, oddly enough uses the same or 2 watts less compared to the seasonic ss300-tgw gold tfx psu. at 20 watts.)
of course the gold level psu is more efficient at 20% power level...
which is why IMO Silentpcreview should measure the efficiency at 20 watts for all psus not just at 10% and higher, because with todays processors and a ssd's it is easy to get a system that idles at 20 watts ... although it becomes Moot point when you get pragmatic about power consumption.
My mythtv backend server with 2HDD (1 WD red 2TB , av-gp 3TB, corsair neutron SSD, i3-4330, 8GB) idles at 29 watts.
(seasonic ss300 sfx, again uses the same power at 28 watrs idle compared to the silverstone 450 gold sfx PSU)
At these low power levels, PSU efficiency starts, as you rightfully said, becoming a huge factor, as well as motherboard features and for Haswell CPUs supported sleep states.
It'd be good to use a universal super-efficient power brick + DC-DC (when no onboard DC-DC is available) to test power consumption for these boards. For me, a dedicated PSU stops making much sense at sub-60W load; a good external power brick + DC-DC is often more efficient, and 100% silent.
The Celeron (and Pentium) Haswells don't support all the deep sleep states (C7 ?), so I expect an i3 and up to have slightly lower power consumption than the cheaper versions.
It'd also be very interesting to see a comparison of different mini-itx and micro-atx motherboards for Haswell, where only power consumption is tested, while keeping CPU/storage/RAM/PSU identical between different boards.