Semiconductor process is only one variable. KNC's prior 28nm product was exceeded in power efficiency by some 55nm devices. The actual power performance is matters, not the lithography.
With all due respect, at the wall I'm not too sure they did. I tried to get to the bottom of that at the time of launch so there was a fair comparison and received a mixed bag of variable answers and little evidence to back up claims of sub 1W/Gh/s. The November boxes eat for sure, but the original October boxes were 1W/Gh/s at the wall. I even toyed with one at 0.85, they can be down clocked. I expect to see other manufacturers doing the same, it makes sense over the long term.gmaxwell is right. Bitfury systems regularly pull .9 at the wall for complete systems. And that's a 55nm chip!!!
Reported from their threads, a Bitfury 16 board rigs @ 500 GH and a power draw of 475watts, plus danger of overloading the pcie connectors if you go over that. Not sure if there is a linear correlation between overclocking efficiency and W/Gh, would be interested in seeing what an October Jupiter pulls @ 500 and 400Gh/s to compare.
