Intel has officially announced that the eighth generation of Core processors will continue to be produced on a 14 nm process. Intel prefers to position this generation as “14 nm +”, emphasizing technical improvements :

• improved channel strain;
• improved profile (fin profile);
• integration of design and production.

Thanks to the improvement of the technical process, we managed to achieve a significant increase in productivity, which will be more than 15% according to the SysMark test. Thus, this year the performance of Core i7 processors will grow more than in the past. This is shown in the slide from the presentation above under the heading “Moore's Advance of 14 Nm”.

A new generation of processors on an improved platform of 14 nm is scheduled for release in the second half of 2017. They will be designated as the Core i7 / i5 / i3-8000 family and will replace the existing 7th generation family.
')
At the presentation for investors, Intel did not say anything about plans to release the Cannonlake family (formerly called Skymont) - microprocessors on the 10-nm process technology. It is assumed that they should be released at the end of 2017, and a working sample of Cannonlake at 10 nm was shown recently at CES. The Cannonlake family was previously positioned as the 8th generation of processor architecture, which will replace Skylake in the framework of the tick-to-tack strategy. Now there is another family that has nothing to do with Cannonlake. Perhaps this is an attempt to sell the old product in a new package.

According to available information , Intel's plans for the release of Kaby Lake-X, Skylake-X and Cannonlake in the second half of 2017 remain unchanged, despite the appearance of the aforementioned new Core i7 / i5 / i3-8000 family. While it is not entirely clear which of the planned product lines should be positioned a new family of processors for desktops.



Perhaps the new family is a kind of transitional stage before the release of Coffee Lake, which is currently planned for I quarter. 2018 year. Some experts suggest that by this step Intel could push the actions of AMD, which is going in March 2017 to bring to the market a line of Ryzen processors .

Cancel Tick-Tack

Intel has consistently adhered to the tick-to-do strategy since 2006. Since then, every two years, she has released processors using a new process technology, significantly increasing the number of transistors on a chip. Each transition to a new technical process was designated as “tick”, and the subsequent improvement of the micro-architecture with the same technical process - “so”. The giant semiconductor industry for ten years worked like a clock, producing new architectures without fail.

Year	Microarchitecture code name	Technical process	"Tick" or "so"
2006	65 nm	P6, NetBurst	"teak"
2006	65 nm	Core	"So"
2008	45 nm	Penryn	"teak"
2009	45 nm	Nehalem	"So"
2010	32 nm	Westmere	"teak"
2011	32 nm	Sandy bridge	"So"
2012	22 nm	Ivy bridge	"teak"
2013	22 nm	Haswell	"So"
2014	14 nm	Broadwell	"teak"
2015	14 nm	Skylake	"So"
2016	14 nm	Kaby lake	"So"

It seems that in 2016, the Intel “clock” short-circuited by 14 nm, and the company announced it was abandoning this strategy .

In principle, nothing wrong with that. Repeat, this year the performance growth of chips (more than 15%) will be even greater than in the past (15%), Intel said. Maybe it is really better to squeeze the entire reserve from the existing process technology, optimizing it, and only then move on. We cannot criticize Intel for withdrawing from a strategy that it voluntarily set itself.

One way or another, but now the tick-to-do strategy has been modified to a different kind.



Instead of a measured metronome, a new procedure has now been implemented with more emphasis on optimization. Perhaps the new architecture will not come out every two years, as it was before.

Why Intel does not force the transition to 10 nm? She doesn’t need to do this, because she believes that she has been so strongly torn off in her technological superiority from competitors in the semiconductor industry (Samsung, TSMC and others). The company estimates this gap at about three years.



This stock allows you to feel quite confident.

New plant for 7 nm

The bright future of Moore's law is to provide the new plant Intel Fab 42 , which will be able to ensure production by the 7 nm process technology.



Construction and equipment will take another three to four years and will require significant investment. The factory in Chandler (Arizona) will reduce the number of local unemployed by about 3,000 people (+ another 10,000 jobs will be added indirectly).

The construction of the plant in Chandler began in 2011. It should be the most advanced and innovative semiconductor enterprise in the world. The building itself was finished in 2013, but instead of installing equipment at 14 nm in early 2014, Intel decided to postpone the launch of the pipeline. At the moment, the plant is ready: air conditioning systems, heating and others - everything functions, it remains only to install and adjust the equipment. Intel does not plan to use this factory for the production of 10 nm process technology, so after a few years, it is quite likely that they will master the production at the following rate of 7 nm.

According to Intel, the equipment will cost about $ 7 billion. This is the cost of a modern industrial enterprise. It is not yet known what specific equipment will be needed. Perhaps Intel will start using photolithography in the deep ultraviolet (EUV) there.

In the early 2000s, Intel hoped that by 2005, the frequency of the processors would increase to 10 GHz, and they would work under voltage below a volt. As we know, this did not happen. About a decade ago, the Dennard scaling law stopped working, stating that with decreasing transistor sizes, the voltage applied to the gate can be reduced and the switching speed increased. Since then, rarely which processor gets a nominal operating frequency above 4 GHz, but there are more cores, the north bridge migrated to the chip from the motherboard, other optimizations and accelerations appeared. Now Moore's law is also slowing down, an empirical observation that indicates a constant increase in the number of transistors on a chip due to a decrease in their size.