Microprocessor giant Intel (INTC -1.79%) is currently in mass production on a chip manufacturing technology that it refers to as 14-nanometer+, an improved version of the 14-nanometer technology that it debuted in the second half of 2014.
The company's next generation technology, known as 10-nanometer, is expected to go into production during the second half of 2017. Over the next couple of years following the introduction of the first 10-nanometer technology, the company says it will introduce both 10-nanometer+ and 10-nanometer++ technologies to improve upon the performance of the initial 10-nanometer technology.
Beyond the various 10-nanometer flavors, though, Intel is working to develop a 7-nanometer technology. Unlike 10-nanometer+ and 10-nanometer++, which are both expected to deliver performance enhancements but not reductions in chip area (known as a "shrink"), the company's 7-nanometer technology is expected to bring performance and power improvements as well as a shrink.
Although much about this upcoming technology remains a mystery, I have a key detail about the technology to share.
Refining a familiar technology
In late 2011, Intel went into production on its 22-nanometer technology. This technology was unique because it was the first implementation of a transistor structure, known as a FinFET, to go into high volume manufacturing. In a nutshell, FinFETs (which Intel marketed as "Tri-Gate") enable substantially better power/performance than prior generation "planar" transistors.
A FinFET. Image source: Intel.
With Intel's 14-nanometer technology, the company refined the same FinFET transistor structure used in its 22-nanometer technology to deliver better performance and power characteristics. With 14-nanometer+ the company further continued to refine the technology.
At 10-nanometer, Intel has already publicly said that it will continue to refine the FinFET transistor structure, and since Intel is planning three iterations of 10-nanometer, the FinFET is likely to be Intel's transistor of choice for several years to come.
There had been some speculation that Intel could move away from the FinFET technology when it transitions to its 7-nanometer technology (after all, the FinFET will be nearly a decade old by the time this technology ramps into production), but it would appear that this won't be the case.
According to a LinkedIn profile of an Intel engineer, Intel's 7-nanometer technology will, indeed, be a further refinement of the FinFET structure, as you can see below:
Image source: LinkedIn.
If we assume that Intel spends three generations on 7-nanometer before transitioning to 5-nanometer (as the company is on track to do with 14-nanometer and plans to do with 10-nanometer), then it might not be until 2023/2024 before Intel transitions to a post-FinFET transistor structure.
What else do we know about 7-nanometers?
At a recent presentation, Intel also provided some insight into the transistor density/cost metrics that it aims to deliver with its 7-nanometer technology.
Image source: Intel.
Indeed, based on the above image, it looks like Intel is promising both area and cost-per-transistor reductions, although it seems as though the company hasn't quite nailed its targets down yet.
The upshot, though, of lower area and ultimately cost per transistor is that chips with more transistors can be made more cheaply on a newer, denser technology than on an older, sparcer technology. The one caveat, though, is that these cost-per-transistor calculations assume equivalent manufacturing yields, something that -- at least with Intel's 14-nanometer technology -- isn't the case when a new technology goes into production.
