Last week, I wrote about the first 20nm application processors, which are slated to be shipping in products early next year. But if the chip-manufacturing companies are a little later than I would have expected for 20nm, they are planning to shift quickly to the next node, the 14nm and 16nm chips. It won’t surprise me if we see very few 20nm chips, and instead see a lot of designs skip that generation and go directly from the 28nm processes standard on most leading-edge chips today to the 14 or 16nm generation.

Of course, Intel is on its own cadence, having started shipping 22nm chips two years ago, with 14nm chips slated for mass availability in the second half of this year. Instead, I’m talking about chips from the fabless semiconductor companies–everyone from Apple and Qualcomm to Nvidia and AMD–which use manufacturing firms known as foundries–such as TSMC, Samsung, and Globalfoundries–to actually produce the chip. All the major foundries are using traditional planar transistors at 20nm, while planning to introduce 3-D or FinFET designs at the next step, which TSMC calls 16nm and Samsung and Globalfoundries call 14nm. In both cases, this would involve changing and shrinking the transistors themselves while leaving the back end at the same design as for 20nm, so it’s something like a “half-node,” instead of a full-generation shrink. (I discussed the difficulties facing chip scaling earlier this month.)

Last week’s big announcement in this vein came from Samsung and Globalfoundries, which announced plans to collaborate on 14nm production, so that chip-design companies could theoretically manufacture the same designs in factories from either company.

Effectively, this seems to mean that Samsung is licensing its 14nm FinFET process to Globalfoundries, which will allow a broader number of factories to use that process, creating a stronger competitor to TSMC, which is the leading foundry. The two groups often vie for leading-edge customers, such as Apple. TSMC and Samsung showed early test chips produced on their 16 and 14nm processes at the ISSCC show a few weeks ago.

Samsung is prototyping 14nm at its factory in GiHeung, South Korea and will be offering manufacturing at its plants in Hwaseong, South Korea and in Austin, Texas, while Globalfoundries will be offering it in its plant near Saratoga, NY.

In the announcement, the two companies said this process will enable chips that are up to 20 percent higher speed using the same power, or could run at the same speed and use 35 percent less power. (Note when any chip manufacturer talks about speed or power, they are talking on a transistor level; finished products often are quite different.) They also said this process provides 15 percent area scaling over industry 20nm planar technology, a nice increase for a half-node. Samsung has already begun prototyping and said it plans to begin mass production by the end of 2014. (Again, note there is usually a lag of several months between when a foundry starts mass production and the chips appear in consumer products.)

This first generation will be on the Low Power Enhanced (LPE) process, with a Low Power Plus (LPP) process providing a performance boost available in 2015. Globalfoundries would be ramping LPE production in early 2015. This is later than its original roadmap but at least the gap between it and 20nm hasn’t gotten any longer.

Both companies say they have their 20nm process working for test products now, and expect production to ramp later this year, though we haven’t heard any specific products announced yet. Globalfoundries says its 20nm technology provides up to 40 percent performance improvement and twice the gate density of its 28nm products, while Samsung has previously said that its 20nm process is 30 percent faster than its 28nm one.

TSMC says it has started full production of 20nm and will be ramping 20nm SoC production in the second half of the year. TSMC claimed its 20nm process can provide 30 percent higher speed or 25 percent less power than its 28nm technology, with 1.9 times the density. Moving to 16nm, TSMC is planning 16-FinFET and 16-FinFET Plus processes, and has said the first version will offer a 30 percent improvement in speed at the same power. More recently, the company has said the Plus version will offer an additional 15 percent speed improvement or a 30 percent power reduction compared with first version (for a total of a 40 percent speed improvement and 55 percent reduction in power over 20nm). This will be followed by a 10nm version, slated to start “risk production” (early prototypes) at the end of 2015, with a 25 percent speed improvement or 45 percent power reduction, compared to the 16-FinFET Plus version, along with a 2.2 X improvement in density.

So far only Qualcomm has announced a major 20nm product, with its first 20nm modem made by TSMC due out in products in the second half of this year, and its first 20nm application processor – the Snapdragon 810 – aimed at products shipping in the first half of 2015. But remember it always takes some time between when the foundries say they are in mass production until real consumer products show up in volume.

The collaboration between Samsung and Globalfoundries is interesting as both have been members of the Common Platform Alliance, which was based around chip-manufacturing processes from IBM. Common Platform apparently covered technologies from 65nm to 28nm, so it seems like this is really the two big manufacturing companies coming together on Samsung’s process without IBM’s involvement. But both Samsung and Globalfoundries are still working with IBM through an R&D group in Albany, NY that is exploring options for 10nm and beyond.

If the companies can actually deliver on their promises, we should see leading-edge consumer products using 28nm most of this year, 20nm next year, 14 or 16 nm in 2016, and maybe 10nm in 2017. Meanwhile, Intel says it is manufacturing 14nm in volume now, and we should see it in many products in the second half of this year, with 10nm following two years behind. This could make the next few years quite interesting, as we may see improvements in power and energy efficiency in our products every year.