KnCMiner has annouced that it will launch deployment of its "Solar" line of chips in Q1 2015.
Based in Sweden, KnCMiner has indicated that its new chip will have a 16 nm process, which represents an improvement over the 20 nm process which was used in the Neptune.
KnCMiner was first to market with a 20nm process with the Neptune, which achieved an efficiency of 0.7 W per Gh/s.
If a 20 nm process can achieve an efficiency of 0.7 W per Gh/s, then what do you think the Solar line of chips can achieve? Well, if sources at KNC are to be believed, a ten-fold improvement is to be had at 0.07 W per Gh/s.
In other words, a 1.4 KW miner which could run from your 15 amp home outlet could hash at a rate of 20 TH/s. Yikes!
The technology which enables this leap in efficiency is called FinFET which allows chip designers to scupt transistors in 3D, instead of the standard planar modeling.
The Solar ASIC, which features 5000 cores, has yet to go into full production. Whether the Solar manufacturing rollout goes smoothly or whether it will be plagued by manufacturing issues like those which sunk manufacturer Hashfast remains open question. Time will tell.
Source:
http://www.asicspace.com/blog/knc-miner-solar-chipsW
The key line here is "if sources at KNC are to be believed", and if you feel that they can be trusted then that's a personal decision, since they don't plan on selling any then you'll never know whether this is horseshit or not. Might frighten their competition though, but going on KNC's past record for efficiency I wouldn't worry too much if I was them, Spondoolie's 28nm design easily outperforms their 20nm one. Cointerra's figures are much more believable, although they are rather on the high side.
What is fact is that moving to 16nm Finfet from 20nm does NOT give you a 10 fold decrease in power consumption, even with 3D transistors. What is fact is that there are physical solutions in 28nm that can better their figure with good 'ol 2D transistors - translating to 0.034 W / (Gh/sec) in 16nm.
So there's something for the "sources at KNC" to consider.
Incidentally, why the 5000 cores guys? Got something against working in powers of 2 or is the yield expected to be so bad that 20% of your cores are unusable?
