J38701 CadenceTECHTALK Automotive Design Banner 800x100 (1)
WP_Term Object
(
    [term_id] => 31
    [name] => GlobalFoundries
    [slug] => globalfoundries
    [term_group] => 0
    [term_taxonomy_id] => 31
    [taxonomy] => category
    [description] => 
    [parent] => 158
    [count] => 240
    [filter] => raw
    [cat_ID] => 31
    [category_count] => 240
    [category_description] => 
    [cat_name] => GlobalFoundries
    [category_nicename] => globalfoundries
    [category_parent] => 158
)

GLOBALFOUNDRIES Extends the FDSOI Roadmap

GLOBALFOUNDRIES Extends the FDSOI Roadmap
by Scotten Jones on 09-19-2016 at 12:00 pm

 On September 8, 2016 GLOBALFOUNDRIES (GF) announced their 12nm FDSOI technology node. On September 12th I had a chance to interview Greg Bartlett, GF Senior Vice President for the CMOS Business Unit (as a side note, GF has: RF SOI, ASIC and CMOS business units).

GF currently has 22FDX – a 22nm FDSOI technology in development with risk production scheduled to begin in Q4 2016 at their Dresden fab. As has been previously disclosed 22FDX has a 14nm front end, a Middle Of Line (MOL) with two double patterned layers and comes in 4 variants.

You can read my write up on 22FDX here.

GF – FDSOI (FDX) offering is being positioned as a simpler, more cost effective alternative to FinFETs for low power and embedded applications. If your design goal is to maximize absolute performance and density FinFETs offer the best solution but for many other applications FDSOI is a better solution. FinFETs are currently in production with 16nm/14nm processes available from three foundries and 10nm and 7nm processes expected over the next two years, TSMC has even disclosed plans for a 5nm FinFET technology in 2019. 12FDX is now being announced as a roadmap extension to show a path beyond 22FDX for FDSOI.

The idea behind 22FDX and 12FDX is that 22FDX provides a path to scale down 28nm designs with lower cost than 16nm/14nm FinFETs and 12FDX will offer a lower cost alternative to 10nm FinFETs.

  • 22FDX requires 40% fewer masks than a 16nm/14nm FinFET process used for mobile application processors. 12FDX has a target of 40% fewer masks than a 10nm process. The lower metal layers for the FDX processes reflects the target applications with simpler designs.
  • 12FDX is designed to avoid triple and quadruple patterning required for 10nm FinFET processes.
  • 22FDX can run at 0.4 volts and 12FDX is expected to run at 50% power savings versus 16nm/14nm FinFETs and 22FDX offers a 70% active power reduction versus 28nm planar.
  • 22FDX and 12FDX provide good analog performance and allow designers to size transistors as needed whereas in FinFET designs sizing is in increments of Fins. Planar devices such as the FDX devices also have lower parasitic capacitance.
  • 22FDX has excellent RF performance with an f[SUB]MAX[/SUB] >300Ghz, 12FDX should be even higher, FinFETs have f[SUB]MAX[/SUB] of ~150GHz. f[SUB]MAX[/SUB] needs to be approximately 6x the application frequency and 300Ghz supports 5G and millimeter wave requirements.
  • Design costs for a 16nm/14nm FinFET process are estimated to be >2x the design costs of a 28nm planar design (~120 million dollars for an SOC). 22FDX offers costs similar to 28nm planar (~$60 million dollars for an SOC). Further significant increases in design costs are expected for 10nm and 7nm FinFET processes.
  • 12FDX should offer a >20% performance improvement versus 22FDX.
  • GF is working with Everspin on MRAM embedded memory for 22FDX and they will carry it over to 12FDX, embedded Flash is problematic below 28nm.
  • Both FDX technologies offer body biasing to tune performance and power, this is a unique feature that FinFETs don’t support to allow “performance on demand”.
  • FDSOI has inherently good soft error performance and is ideal for harsh environments such as automotive.
  • 22FDX risk production will begin in Q1 2017. 12FDX tapes outs are expected in early 2019 with wafers out late 2019.

A lot of observers talk about FinFETs and FDSOI as if it is one technology versus the other whereas I believe GF views them as complimentary technologies. FinFETs are targeted at very large – high performance designs that need the highest possible performance and density. The FDX – FDSOI technologies are designed and targeted to address low power, smaller, lower performance designs. The low power operation, good performance on-demand, and analog and RF capabilities of the FDX technologies make them ideal for a wide variety of applications such as are envisioned for the Internet Of Things (IOT), automotive and mobile communications.

If you think about a “typical” IOT design you might have the following requirements:

  • A sensor.
  • An analog interface to connect with the sensor.
  • A processor for control and data processing.
  • Memory for code and data storage. The memory needs to be nonvolatile to retain the code and data in case of power loss.
  • RF capability to transmit the data.
  • A battery to power the device.

The FDX platform can integrate the processor, memory, analog interface and RF all on one highly power efficient device to maximize battery life.

Applications in automotive and mobile communications can also take advantage of the unique FDX capabilities. The excellent RF performance of the FDX platform will be particularly important for 5G communications. There has been some speculation that 12FDX could be used for cell phone application processors. While I think there may be a play at the low end the high end processors are still driving towards ever higher performance for on board video processors and I would expect that would continue to be served by leading edge FinFET technologies.

FDSOI has been a technology with a lot of promise for several years. With the introduction of 22FDX and 12FDX by GF along with the commitment to offer high volume manufacturing capacity FDSOI is finally positioned to take off in the market place.

Share this post via:

Comments

0 Replies to “GLOBALFOUNDRIES Extends the FDSOI Roadmap”

You must register or log in to view/post comments.