Strangely enough I just finished reading that article in its entirety twenty minutes before clicking on this thread.
The power consumption figures surprised me, especially at idle; seems that the Haswell based G3220 (25W) actually idles lower than the Bay Trail-D (27W) and Kabini (27-28W) based solutions.
Disappointing to be honest.
There are a few AM1 boards with four, although they're generally a little more expensive.
I don't think I've seen numbers for bay trail vs low power Haswell's but bay trail and kabini use WAY less power than most of those systems there.
The issue is reviewer stupidity mostly, psu's. Benchmark a range of systems with a 500 or 1000W psu and at the bottom end the psu is so inefficient below the 5% capacity mark that you've got the psu effecting the power usage.
I think it was pcper, /looks it up
http://www.pcper.com/news/Editorial/AMD-AM1-Retested-60-Watt-Power-Supply
Their first review showed "meh" numbers, the retest with a psu designed for much lower power usage showed completely different numbers.
12W idle with a 60w power adaptor and 23W idle using a 500W atx psu.
29W load with the 60W and 37W load with the atx psu.
The problem with testing is that most websites use a standard setup.
It could also be argued if some of the higher end cpu's wouldn't also show better idle numbers with a better psu, and that is likely the case, but you can't really get around needing a psu that serves the maximum power you'd be using very well and efficiency at idle(with a low enough idle system) will be effected.
With a Kabini/bay trail system you'd seriously want to think about how you power it to get real efficiency out of it.
Kabini seems in general a little better power wise than Baytrail, at least in home systems.
I think it's Toms that had the 25W Kabini using 21W average during a 3dmark run while the 10W(lol) Intel comparable bay trail chip used 19W average in the same test.
Bay trail categorically uses way beyond it's rated numbers when it's powered up and Kabini uses way way below what you think it would idle given it's tdp rating. Which chip actually uses less power is a fairly difficult question.
Really poor reviewing in general. Would like to see an absolute minimal desktop system on a say 150W psu and get some idle numbers from the lowest power Haswell/Kaveri, and idle numbers on maybe a 40W power adaptor for Kabini/bay trail and see how they all come out. That no one has done it yet is pretty stupid.
I'd certainly lean towards one of the mobo's that comes with a power adaptor connector instead/as well as a atx psu connector.