I think you should stop spreading misinformation
http://www.eda-stds.org/edps/Papers/4-4 FINAL for Tom Quan.pdf
refer to slide 19.
16FF/28HPM 16FF/20SoC
Speed @ same total power 38% 20%
Total power saving @ same speed 54% 35%
Gate density 2X 1.1X
Since 16FF is 20% faster at same total power compared to 20SOC and 38% faster compared to 28HPM and if we keep speed with 28HPM as 1x and speed with 16FF as 1.38x then speed with 20nm is 1.38x / 1.2 = 1.15x
So there is a 15% speed improvement at same power from TSMC 28 HPM to TSMC 20 SOC. remember 28 HPM is the best 28nm high k process at TSMC and superior to TSMC's 28HP and 28HPL.
the key with 20nm is there is a close to doubling of transistor density at 1.9x. so this allows much bigger chips to be produced. they can be clocked at lower speeds/voltage so that within the same power budget you can still get a significant performance improvement. look at the Apple Cyclone core on Samsung's 28nm process.
http://www.anandtech.com/show/7355/chipworks-provides-first-apple-a7-die-shot
http://www.anandtech.com/show/7335/the-iphone-5s-review/5
Cyclone clocked at 1.3 Ghz is competing with Baytrail boosting to 2.4 Ghz. Even in multithreaded benchmarks like cinebench r11.5 Baytrail is running at close to 2.4 Ghz. you can derive that from the scaling from single thread to multi thread performance.
TSMC, Samsung and GF are well funded to pursue these multi billion dollar efforts. Most importantly these foundries have customers who sell hundreds of millions of chips every year like Qualcomm, Apple, Nvidia, AMD and Samsung themselves. Their chips power desktop, mobile, console and PC gaming, HPC/servers and other markets like embedded. So there is no shortage of customers for these foundries.
If anything I have to say further node progresses are getting difficult and out of Intel's control as lithography becomes the primary constraint and not just the transistor device. There is a very good chance that Intel will not get to 7nm before the end of the decade. the 2 year node progression for Intel is going to be very difficult going forward. Intel 14nm requires double patterning immersion litho, 10nm requires quadruple pattering immersion litho and 7nm is mostly not possible without EUV. thats the reason Intel, TSMC and Samsung have invested in ASML for EUV development. still EUV seems to be ready only by the end of the decade.
http://www.intc.com/releasedetail.cfm?ReleaseID=690165
http://www.asml.com/asml/show.do?ctx=5869&rid=46974
http://www.asml.com/asml/show.do?ctx=5869&rid=46903
http://www.forbes.com/sites/jimhandy/2012/08/27/whys-everyone-investing-in-asml/
this is utter crap. Why isn't Intel able to destroy Nvidia in HPC where Intel is competing with Knights Corner. Nvidia and AMD's architectures are vastly superior to Intel's for GPU compute and HPC. Nvidia dominates the HPC market with their Teslas. :biggrin: