GlobalFoundries introduces 22nm FD-SOI process technologies

[monstercameron]

The GlobalFoundries 22FDX family of process technologies consists of four options:
22FD-ulp: Designed for inexpensive system-on-chips used for mobile applications. The process uses body-biasing, and delivers greater than 70 per cent power reduction compared to 0.9V 28nm HKMG. Moreover, select chips made using 22FD-ulp will be able to operate at 0.4V.
22FD-uhp: Designed for networking applications with analog integration. The manufacturing tech offers forward body-bias, application optimized metal stacks, and support for 0.95V overdrive.
22FD-ull: The process tech is aimed at Internet-of-Things devices and features leakage of as low as 1pa/um. The manufacturing tech sports flexible body-biasing and other capabilities designed to reduce power consumption.
22FD-rfa: The process tech developed specifically for radio frequency and analog applications. The technology will be used to make high-volume RF applications such as LTE-A cellular transceivers, high order MIMO WiFi combo chips, and millimeter wave radar.
GlobalFoundries’ 22FDX process technologies are expected to be used by multiple designers of chips and are supported by various IP providers. Among supporters of the manufacturing processes are ARM, Imagination Technologies, Freescale, IBS, STMicroelectronics, VeriSilicon and other. Design starter kits and early versions of process design kits (PDKs) are available now with risk production starting in the second half of 2016. Mass production of 22FDX chips will start in the second half of 2017.

source http://www.kitguru.net/components/a...-introduces-22nm-fd-soi-process-technologies/

@IDC what do you think?

 

[Idontcare]

@IDC what do you think?

GF needs process tech that differentiates it from Samsung and TSMC.

I think it is great news and one can only hope the leaders in fabless companies are looking to get wearables and ultra-mobiles produced on it asap.

 

[monstercameron]

GF needs process tech that differentiates it from Samsung and TSMC.

I think it is great news and one can only hope the leaders in fabless companies are looking to get wearables and ultra-mobiles produced on it asap.

Thanks. Also you were the who told me about minimum transistor switching voltage, and here glofo claims equivalent design ported from 28nm running at 0.9v and operate as low as 0.4v on the new process, would this significantly lowered voltage lead to substantial gain in perf/w metrics?

 

[Phynaz]

GF needs process tech that differentiates it from Samsung and TSMC.

I think it is great news and one can only hope the leaders in fabless companies are looking to get wearables and ultra-mobiles produced on it asap.

Press release says volume production 2H2017. Add the couple of quarters of slippage that is now the norm in this industry and we're looking at a planer process in 2018.

Where do you think the other foundries will be at with their low cost options in that time frame? Obviously finfet isn't for the masses.

 

[dark zero]

Amd would love this... At last 22nm FX chips

 

[Vesku]

This isn't for high performance CPUs/GPU/APUs, this is GF offering a path for the majority of fabless companies to move past low power 28nm to a new node that is cheaper both in design cost and per transister cost (GF's target goals for this process) than 16/14FF. Guess this is GFs play for all those TSMC customers that reliably sit on nodes for years.

 

[dark zero]

This isn't for high performance CPUs/GPU/APUs, this is GF offering a path for the majority of fabless companies to move past low power 28nm to a new node that is cheaper both in design cost and per transister cost (GF's target goals for this process) than 16/14FF. Guess this is GFs play for all those TSMC customers that reliably sit on nodes for years.

So... Here we go 22nm Amd Mullins

 

[CHADBOGA]

Amd would love this... At last 22nm FX chips

Too late now.

 

[dark zero]

Too late now.

Not at all.. It could give some air before Zen... On Excavator mode at 22nm would become decent.

In the worst case Mullins and Carrizo-L at 22nm could be interesting to see

 

[CHADBOGA]

Not at all.. It could give some air before Zen... On Excavator mode at 22nm would become decent.

In the worst case Mullins and Carrizo-L at 22nm could be interesting to see

None of those products will come out on GF's 22nm FD-SOI process.

 

[dark zero]

None of those products will come out on GF's 22nm FD-SOI process.

I know... I was joking Anyways, seems that the IoT gets the benefits of that.

 

[el etro]

I know... I was joking Anyways, seems that the IoT gets the benefits of that.

22FD-SOI main aimming is low and mid-end smartphones. Can be used on bigger xPUs, but not on cpu's and gpu's.

 

[DrMrLordX]

22nm FD-SOI as announced would be suitable for something like Nolan or Amur, if they choose to resurrect those parts. But really they should be on 14nm for Basilisk by 2017.

 

[Fjodor2001]

What customers can we expect on 22 nm FD-SOI, and why?

Isn't everyone heading for 14/16 nm?

 

[mrmt]

I think they promised 14nm for 2015 and it seems globalfoundries missed this target.

 

[dark zero]

One thing is Finfet and another different is SOI. So it isn't dead at all. And that is not bad since Finfet cost way more than SOI.

And AMD could use that on Mullins or Nolan.

 

[Idontcare]

Thanks. Also you were the who told me about minimum transistor switching voltage, and here glofo claims equivalent design ported from 28nm running at 0.9v and operate as low as 0.4v on the new process, would this significantly lowered voltage lead to substantial gain in perf/w metrics?

Exactly, the lower voltage reduces dynamic power consumption, and the 1pA/um leakage reduces static power consumption.

Precisely the thing we want and need so our wearables and ultra-mobiles don't need to be recharged every other day or have a bulky power pack built in.

Press release says volume production 2H2017. Add the couple of quarters of slippage that is now the norm in this industry and we're looking at a planer process in 2018.

Where do you think the other foundries will be at with their low cost options in that time frame? Obviously finfet isn't for the masses.

The main strategy of the foundries is to get EUV going for cost reduction. I don't see anyone stepping back from Finfet at this point. FDSOI is an interesting and helpful development, but much like FRAM it isn't something that is about to create a disruptive technology gap over that of evolutionary improvements in traditional (finfet) cmos.

 

[desura]

Intel's fab lead is so incredible sometimes.

 

[erunion]

What customers can we expect on 22 nm FD-SOI, and why?

Isn't everyone heading for 14/16 nm?

No, only high performance, leading edge customers are. TSMC projects lots of 28nm demand for years to come.

From my industry outsider perspective, I think GF is looking to appeal to those 28nm customers with this node.

 

[NostaSeronx]

Should probably be pointed out that... 22nm {GF} FDX is 14nm {STM} FEOL and 28nm {STM} BEOL.

The process is 14nm with a density of 28nm.

 

[dark zero]

Also... isn't SOI superior to FinFet in some aspects?
And also... isn't ULV even more superior than the 2 latter ones?

 

[Azuma Hazuki]

Where's NostaSeronx when you need him? Every time I see "FD-SoI" my mind's eye conjures up that human-faced cat avatar of his...

*looks in mirror in dark room*
NostaSeronx
NostaSeronx
NostaSeronx

 

[Azuma Hazuki]

Holy CRAP it worked.

 

[Fjodor2001]

No, only high performance, leading edge customers are. TSMC projects lots of 28nm demand for years to come.

Yes, that I know. But I meant those that wants to create a chip on one of the latest nodes. This new 22 nm FD-SOI process will be out quite late compared to 14/16 nm. So is it worth it to design a new chip for this node, or migrate an existing one to it?

 

[beginner99]

Yes, that I know. But I meant those that wants to create a chip on one of the latest nodes. This new 22 nm FD-SOI process will be out quite late compared to 14/16 nm. So is it worth it to design a new chip for this node, or migrate an existing one to it?

Because it will be cheaper than 28 nm and especially cheaper than 14/16 nm (design and production cost). 14/16nm costs more per transistor than 28 nm!!! So only good for products that depend on low power usage, eg. smartphones and performance must increase.