From the origonal paper www.cs.wisc.edu/vertical/papers/2013/hpca13-isa-power-struggles.pdf (which ExtremeTech does not link to):

Technology scaling and projections:
Since the i7 processor is 32nm and the Cortex-A8 is 65nm, we use technology node characteristics from the 2007 ITRS tables to normalize to the 45nm technology node in two results where we factor out tech-
nology; we do not account for device type (LOP, HP, LSTP).
For our 45nm projections, the A8â(TM)s power is scaled by 0.8Ã-- and
the i7â(TM)s power by 1.3Ã--. In some results, we scale frequency
to 1 GHz, accounting for DVFS impact on voltage using the
mappings disclosed for Intel SCC [5]. When frequency scal-
ing, we assume that 20% of the i7â(TM)s power is static and does
not scale with frequency; all other cores are assumed to have
negligible static power. When frequency scaling, A8â(TM)s power is
scaled by 1.2Ã--, Atomâ(TM)s power by 0.8Ã--, and i7â(TM)s power by 0.6Ã--.
We acknowledge that this scaling introduces some error to our
technology-scaled power comparison, but feel it is a reasonable
strategy and doesnâ(TM)t affect our primary findings (see Table 4).

If you look at the graph "raw average energy normalised" you see that the ARM A9 core has the lowest energy score -> that clearly shows ARM being the most efficient and hence the conclusion is completely wrong.
Still the test is very interesting. I would like to see it updated with latest CPUs

CPU performance is worse than Haswell and IGP is better - that's what I said. Motherboards are about the same price

They don't anymore. Kaveri is about same price as i3 Core/pentium Haswell, but with more power draw and less performance. Where AMD win is the IGP, which always has been better

I'm glad to see AMD is using their development budget wisely and not wasting it on other stuff, like it making their x86 cores competitive versus Intel

Not really, because the tech is improving all the time. At 20nm the have high-k metal gates and at 16nm FinFETs.

Each semiconductor node shrink is faster and more power effiecient than the previous. For instance, TSMC 20nm process is 30% higher speed, or 25% less power than 28nm. Likewise, 16nm will provide 60% power saving than 20nm.

He claims that Unreal engine works well on iOS, but this is not really true. Basic features like dynamic shadows are not supported at all and performance is pretty poor - you really need a A6 based device to run it well.

If you're running a game you will typically have the GPU and CPU maxed out, so basically all the clever power gating and duty cycle stuff is switched off. Basically, the battery isn't going to last much longer than prev gen CPUs.

There are no HSA applications (just a few tech demos) and it doesn't look like they are going to appear any time soon.
HSA will only work on kaveri CPUs, which are maybe 0.001% of the market. Maybe if Intel adoped HSA then there would be some more motivation for developers to support it.

Reminds me of this dilbert cartoon: http://www.dilbert.com/fast/20...

An interesting article here discribes the horrendiously difficult challenges that face EUV:
https://www.semiwiki.com/forum...

You mean they *had* plans to it in breeder reactors.
Japan's fast breeder plan has been more or less mothballed after the failure of Monju and Fukushima out cry:
http://asia.nikkei.com/Politic...

I don't think there are any rare earth materials used in polysilicon cells. In fact, the solar city cells are cheaper to manufacture because they are using copper instead of (standard) silver electrodes

So, only the benchmarks that you say are important count. That's pretty much the dictionary definition of cherry picking