Knc produced about the most power inefficient 28nm rig on the market, so I suspect their 20nm product will offer a similar power efficiency as other companies 28nm products.
Proof? How power efficient are CT and HashFail?
I wasn't even talking about CT or HashFail because they both suck.
November Jupiters use about 1.3W/GH at the wall, and the Neptune is supposed to be ~30% more efficient which puts it at ~1W/GH. Several competitors have already matched that including the Coincraft rigs, Dragon knockoffs, BitFury rigs, the Bitmain S2, and Spondoolies-Tech.
Several of these competitors (BitFury and Bitmain) have produced a product that uses 1W/GH using older 55 nm tech.
Also, the next generation ASICMiner chip is supposed to use well under 1W/GH.
So only undervolted CoinCraft rigs matched KnC. They are the only ones using 28nm. The rest aren't. You said "the most power inefficient 28nm rig" The other rigs aren't 28nm so you can't compare them.
Big deal. The point is, 20 and 28 nm were chosen for increased power efficiencies. It's a well known fact that in general, smaller die = less power. However, KNC failed in that regard when there is a list of hardware that is more efficient.
If AMD came out with a 55nm GPU that was more efficient than Nvidia's 28nm GPU, would you still brag that Nvidia was the most power efficient 28nm GPU? Of course not.
