Thursday, September 20, GlobalFoundries announced a 14-nanometer (nm) XM PDK (eXtreme Mobility Process Development Kit) featuring FinFET three-dimensional transistors. Customer tape out is expected in early 2013 and volume production in 2014 at GlobalFoundries’ Fab 8 in Malta, New York. This surprise announcement is just one year after the foundry is scheduled to put its 20nm process in production.

Introducing the 14nm node called "14XM" just one year after 20nm
Introducing the "14XM" just one year after 20nm in a bid to reach parity with Intel, which is scheduled to introduce the 14nm process "Broadwell" during 2014 as well

The14nm XM successfully combines 14nm FinFET features onto the company?s existing 20nm offering. Gregg Bartlett, GlobalFoundries? CTO (Chief Technology Officer) said: "We have more than a decade of FinFET R&D to build on as we prepare to bring this technology to production. We are confident this foundation will enable us to lead the foundry volume ramp of FinFETs, just as we did with High-K Metal Gate (HKMG)." GlobalFoundries is a key foundry member of the Common Platform Alliance along with IBM and Samsung.

Evolution of FinFET
Evolution of FinFET in Common Platform road map

GlobalFoundries said their 14nm FinFET will have a 48nm pitch, the same as analysts expect Intel to have on their 14nm tri-gate process. They claim other feature size measurements are similar or identical to what Intel will have with their 14nm offering. This should make the 2014 smartphone SoC (System on a Chip) benchmarking a lot of fun.

Mike Noonan, executive vice president of worldwide marketing and sales at GlobalFoundries, said: "The goal is to give our customers the power and performance to compete with Intel," and their 14nm tri-gate process.
GlobalFoundries? transistor is a 14nm FinFET but the rest of the PDK is unchanged from their 20nm-LPM process. It is basically 20nm process with 14nm transistors (at 20nm spacing). This process has many of the advantages of 14nm.

20-55% higher performance depending on operating voltages.
20-55% higher performance depending on operating voltages.

40-60% increased battery life]]
40-60% increased battery life. 

Last month ARM and GlobalFoundries announced an agreement to develop 20nm process and FinFET. ARM agreed to develop a platform of Artisan physical IP including standard cell libraries, memory compilers and processor-optimization packages (POPs) for the GlobalFoundries processes. Simon Segars, executive vice president and general manager of the processor and physical IP divisions at ARM, then said: "By proactively working together to enable next-generation 20nm-LPM and FinFET process technologies, our mutual customers can be assured a range of implementation options that will enable two more generations of advanced semiconductor devices."

Clearly GlobalFoundries’ announcement is putting pressure on competition with a technology lead over the other two dedicated foundry suppliers. Taiwan Semiconductor Manufacturing Company (TSMC) and United Microelectronics Corporation (UMC) have said they plan to integrate FinFETs in their 20nm processes, with UMC set to put them in production in the second half of 2014 and TSMC will likely follow some time after.

Since all major players such as Apple, AMD, Qualcomm and Nvidia are known to be having problems with TSMC’s 28nm process not having adequate volume, GlobalFoundries’ 14nm XM PDK announcement will be a wake up call in San Diego, Santa Clara, Sunnyvale and Cupertino.

However, an old hand in the semiconductor industry we know said this about Intel’s tri-gate and GlobalFoundries FinFET process technology: "if you now need two masking steps instead of one masking step, it adds another bottleneck to an already overburdened part of the process." Therefore, everyone will be checking out GlobalFoundries? 14nm XM test chips and watching closely the first tape outs.

BSN Take
Today’s announcement is much more important than at first look. For better part of the past decade, we’ve seen Intel showing off process advantage, overtaken by TSMC on just copuple occasions with half-node efforts (55nm before Intel’s 45nm, 28nm before 22nm). GlobalFoundries announcing 14nm to go in place during the same year when Intel is expected to introduce its desktop/mobile/server processor "Broadwell" (Haswell die-shrink) effectively opens the door for AMD to get back into the game, process-node wise. AMD, NVIDIA, Qualcomm would probably wish if we could teleport into 2014. Battle call looms…