ARM is stepping up its effort to support the FD-SOI ecosystem. “Yes, we’re back,” confirmed Ron Moore, VP of ARM’s physical design group. This and much more good news came out of the recent FD-SOI Symposium organized in Silicon Valley by the SOI Consortium.
The full-day Symposium played to a packed room, and was followed the next day by a full-day design tutorial. Though it was a Silicon Valley event, people flew in from all over the world to be there. (BTW, these symposia and tutorials will also be offered in Japan in June, and Shanghai in the fall). I’ll cover the Silicon Valley FD-SOI design tutorial (which was excellent, btw) in a separate post.
Most of the presentations are now posted on the SOI Consortium website. Here in this ASN post, I’ll touch on some of the highlights of the day. Then in upcoming posts I’ll cover the presentations from Samsung and GlobalFoundries.
If you’re designing in FD-SOI, we’ll help: that was the key message from ARM’s Ron Moore during the panel discussion at the end of the day. Earlier that morning, he’d given an excellent presentation entitled Low-Power IP: Essential Ingredients for IoT Opportunities.
CAGR for most IoT units is roughly 50%, he said, counting home (1.6B units by 2020), city (1.8B), industrial (0.6B) and automotive (1.1B). Compare that to the 2.8B smart phones – which he sees as a remote control and display device. The key differentiator for IoT is that 90% of the time the chip is idle, so you really don’t want leakage.
FD-SOI, he said, gives you a silicon platform that’s highly controllable, enables ultra-low power devices, and is really good with RF. ARM’s worked with Samsung’s 28FDS FD-SOI offering comparing libraries on bulk and FDSOI, for example, and came up with some impressive figures (see the picture below).
The foundry partners and wafer providers are in place. So now ARM is asking about which subsystems are needed to fuel FD-SOI adoption. Ron recognizes that the ARM IP portal doesn’t yet have anything posted for FD-SOI, but they know they need to do it. He called on the SOI Consortium to help with IoT reference designs and silicon proof points.
In the Q&A, audience member John Chen (VP of Technology and Foundry Management at NVIDIA) asked about FD-SOI and low-cost manufacturing of IoT chips. Moore replied that we should be integrating functionality and charging a premium for IoT chips – this is not about your 25-cent chip, he quipped.
Geoff Lees, SVP & GM of NXP’s Microcontroller business gave a terrific talk on their new i.MX 7 and 8 chips on 28nm FD-SOI. (And Rick Merritt gave it great coverage in EETimes – see NXP Shows First FD-SOI Chips.)
NXP’s been sampling the i.MX 7 ULP to customers over the last six months, the i.MX 8QM is ramping, and the i.MX 8QXP, 8Q and 8DX are enroute. Each of these chips is optimized for specific applications using biasing. A majority of the design of each chip is hard re-use, and the subsystems can be lifted and dropped right into the next chip in the series. Power consumption and leakage are a tiny fraction of what they’d had been in previous generations. Ultra low power (aka ULP) is heading to new levels, he says.
With FD-SOI, it’s easy to optimize at multiple points: in the chip design phase, in the production phase and in the use phase. They can meet a wide range of use cases, precisely targeting for power usage. FD-SOI makes it a win-win: it’s a very cost effective way to work for NXP, plus their customers today need that broader range of functionality from each chip.
Geoff tipped his hat to contributions made here by Professor Boris Murmann of Stanford, who’s driving mixed signal and RF into new areas, enabling high-performance analog and RF integration. (Folks attending the FD-SOI tutorial the next day had the good fortune to learn directly from Professor Murmann.)
Finally, he cited something recently pointed out by Soitec (they’re the SOI wafer folks) Chief Scientist Bich-Yen Nguyen: if half your chip is analog and/or RF, she’s observed, the future is very bright indeed for FD-SOI.
Briefly, here are some more highlights.
Synopsys: John Koeter, VP of the Marketing Solutions group showed slides of what they’ve done in terms of IP for Samsung and GlobalFoundries’ FD-SOI offerings. But there’s a lot they’ve done with partners he couldn’t show because it’s not public. In terms of tools and flows, it’s all straightforward.
Dreamchip: Designing their new chip in 22nm FD-SOI was 2.5x less expensive than designing it in FinFET would have been, said COO Jens Benndoorf in his presentation, New Computer Vision Processor Chip Design for Automotive ADAS CNN Applications in 22nm FDSOI. One application for these chips (which taped out in January) will be “digital mirroring”: replacing sideview mirrors with screens. Why hasn’t this been done before? Because LED flickering really messes with sensor readings – but they’ve mastered that with algorithms. The chip will also be used for 360o top view cameras and pedestrian detection. They’re using Arteris IP for the onchip networking, and implemented forward body bias (FBB). The reference platform they created for licensing has generated lots of interest in the automotive supply chain, he said.
Greenwaves: CEO Loic Lietar talked about the high performance, ultra-low power IoT applications processor they’re porting from bulk to FDSOI with a budget of just three million euros. The RISC-V chip leverages an open source architecture (which he says customers love) and targets smart city, smart factory, security and safety applications. As such, it needs to wake up very fast using just microwatts of power – a perfect match for body biasing in FD-SOI.
Leti: In her talk about roadmaps, CEO Marie-Noelle Semeria said the main two drivers they’re seeing in the move to FD-SOI are #1: low power (a customer making chips for hearing aids can cut power by 8x using body biasing, for example) and #2: RF (with Ft and Fmax performance that “…will be hard for FinFET to achieve”). Leti knows how to pull in all kinds of boosters, and is finding that RF performance is still excellent at the 10/7nm node. They’ve developed a low-power IoT platform with IP available for licensing. Other recent FD-SOI breakthroughs by Leti include: demonstration of a 5G mmW 60GHz transceiver developed with ST; the first 300mm Qbit, opening the door to quantum computing; a photodiode opening the door to a light-controlled SRAM; and a new 3D memory architecture leveraging their CoolCubeTM that they’re working on with Stanford.
IBS: CEO Handel Jones predicts that there “will be war in the year to come” at the 22nm node, as all the big foundries take aim. FD-SOI is the best technology for RF, ULP and AMS, and there’s a huge market for it. He also said China made the right decision to support FD-SOI, and will come out ahead in 5G.
The day ended with a lively panel discussion (moderated by yours truly) featuring experts from ARM, GF, Invecas, Soitec, Synopsys, Verisilicon and Sankalp. IP availability was a big theme, but generally there was agreement that while some gaps still exist, they’re being filled: lack of IP is no longer an issue. Soitec VP Christophe Maleville confirmed that the wafers for FD-SOI are readily available and that they’re seeing excellent yields.
All in all, it was another really good day for FD-SOI in Silicon Valley.
If you’re headed to DAC (June 5-9 in Austin,TX) and are interested in learning more about FD-SOI, there will be lots of opportunities. Here’s a quick rundown.
Synopsys (stands 149 & 361) and GlobalFoundries are hosting a dinner on Tuesday evening (7 June) at the Austin Hilton around the theme, What’s Important for IoT—Power, Performance or Integration… or All of the Above? They’ll be talking about how FD-SOI addresses these challenges. Panel members will discuss design techniques to push the envelope on low power, low leakage, burst performance and optimal cost to enable the design of innovative IoT-based products. Attendance is free, but registration is required and seating is limited. Click here to go to the registration site.
Samsung Foundry (stands 607 and 706) and partners will be doing a number of presentations on Samsung’s 28nm FD-SOI offering, 28FDS. They’ll be showcasing 28FDS wafers, offering multiple presentations by Samsung Foundry’s experts, and sharing solutions built on the 28FDS technology by their Foundry Ecosystem partners. As noted in ASN coverage of the recent SOI Consortium event in San Jose (read it here), Samsung is now in commercial production of 28FDS. They have a strong 28nm FD-SOI tape-out pipeline for 2016, and interest is rising fast.
IP Track: Minimizing SOC Power Consumption: A Top Down Design Methodology or Bottoms Up Starting With the Process Selection Problem? Panelists include Carlos Mazure (of the SOI Industry Consortium & Soitec) and Ron Martino (of NXP) Monday, June 6th from 4:00pm – 5:00pm in Ballroom G.
Variation-Aware Design at Advanced and Low-Power Processes. Panelists include Azeez Bhavnagarwala (ARM), Glen Wiedemeier (IBM), John Barth (Invecas) and Jeff Dyck (Solido). Monday, June 6th from 10:30am – 11:30am, Room: 9BC.
Presentation 9.1 Impact of Leakage & biasing on Power in 22FDX Process. By Krishnan Subramanian et al (Invecas) and Sankar Ramachandran – (Apache Design). Monday, June 6th, 3:30pm – 4:00pm, Ballroom G.
Presentation 50.4 Leveraging FDSOI through Body Bias Domain Partitioning and Bias Search. By Johannes M. Kuehn et al (Eberhard Karls Univ. Tubingen & Keio Univ.) Wednesday, June 8th, 1:30pm – 3:00pm, Room: 17AB. This presentation will be given at 2:15. (You can also get the paper from the ACM site here.)
101.12 Parametric Exploration for Energy Management Strategy Choice in 28nm UTBB FDSOI Technology. By Jorge Rodas et al (CEA-Leti Minatec & Univ. Grenoble Alpes) Work-in-Progress (WIP) poster session, Wednesday, June 8th, 6:00pm – 7:00pm, Room: Trinity St. Foyer
Stands & More
Cadence Theater (stand 43 – full schedule here)
Tuesday, June 7th
Wednesday, June 8th
Leti (stand 1818) – a driving force behind all things SOI, stop by to learn more about Silicon Impulse®, their FD-SOI platform for IoT & ultra-low-power (ULP) apps that helps start-ups, SMEs and large companies evaluate, design, prototype & move to volume (more here).
And finally, the opening keynote on Monday morning (at 9:15 in Ballroom A) will be given by NXP’s Lars Reger, CTO of their Automotive Business Unit. The topic is Revolution Ahead – What It Takes to Enable Securely Connected, Self-Driving Cars. When it comes to automotive, NXP is the original SOI pioneer, dating to back to 1999. NXP’s sold billions of SOI-based chips for high-voltage automotive applications – they’re used by virtually every carmaker on the planet (read about the early history here and here).
And now with the Freescale acquisition, NXP is full speed ahead with FD-SOI applications processors. If you missed it, you’ve got to read the recent ASN series by Ron Martino (NXP’s VP for i.MX Applications Processor and Advanced Technology Adoption). He explains why they chose 28nm FD-SOI, and exactly what it does for the i.MX 7 series (32-bit ARM v7-A core, targeting the general embedded, e-reader, medical, wearable and IoT markets) and i.MX 8 series (64-bit ARM v8-A series, targeting automotive applications, especially driver information systems, as well as high-performance general embedded and advanced graphics applications) Click here to read it now. NXP gave a demo of the I.MX 8 at FTF 2016 a few weeks ago – check out the video they posted on Twitter here.
If you go to DAC and you have a Twitter account, be sure to tweet #FDSOI and #53rdDAC – @followASN will be happy to pass it along!
This is part 2 (of 2) of ASN’s coverage of the epic FD-SOI Symposium in San Jose. In part 1 we looked at the exciting developments happening at 28nm (if you missed it, click here to read it now). Here in part 2, we’ll look at 22nm, covering the presentations by GlobalFoundries, ARM, VLSI Research and Sigma Designs. Again, the presentations are now starting to be available on the SOI Consortium website – click here to see them (they’re not all there as of today, though, so keep checking back).
Dan Hutcheson, CEO of VLSI Research, has come around to FD-SOI. His excellent talk, “FD-SOI: Disruptive or Just Another Process” (click here to download it), concluded that FD-SOI is not disruptive – but it’s an enabler of disruption. The disruption is IoT, and it’s going to be a big one. To prepare for his talk, he did an informal survey of designers at a dozen top companies. Here are some of the things he heard:
Some companies are using FinFET for some chips and FD-SOI for others, depending on the market they’re targeting – either way, the technologies will co-exist. FinFETs were generally chosen for high-density chips from large companies with lots of money; FD-SOI by those who have time-to-market constraints, are looking to differentiate their products, appreciate the much lower NRE* costs, and that are going for power, reliability and analog advantages.
People see a future with FD-SOI – it’s not a one-trick process.
The design community is happy to be able to re-use many of their favorite techniques that were lost after the 130nm node.
Top target markets for FD-SOI are (by far) IoT, automotive and low-power, followed by analog/mixed-signal, networks, RF, low-end products, mobile, peripherals, MPU/GPU, image sensors and rad-hard.
Here are a couple of his slides that sum up the technical and business reasons people cited as reasons for going to FD-SOI:
Dan then made a video recapping his San Jose presentation – it’s awesome – click here to see it.
The ballroom packed right out when GloFo VP Subramani Kengeri took the stage to present, “Enabling Next Generation Semiconductor Product Innovations with 22FDXTM.
In terms of energy efficiency, he explained, 0.4V is the minimum energy point for almost any technology – and FD-SOI gets you 0.4V. He then went on to reiterate the features of GloFo’s 22FDXTM Platform, the industry’s first 22nm FD-SOI:
Ultra-lower power with 0.4 volt operation
Software-controlled transistor body-biasing for innovative performance and power optimization
Delivers FinFET-like performance and better energy-efficiency at 28nm-like cost
Integrated RF: reduced system cost, and back-gate feature to reduce RF power up to ~50%
Integrated eNVM and RF enables lowest cost and smallest form-factor
Post-Silicon Tuning/Trimming for Analog/RF, SRAM and Power/Performance optimization
Enables innovative applications across mobile, IoT and RF markets
70% lower power than 28HKMG, 20% smaller die than 28nm bulk planar
Lower die cost than FinFETs
He then gave lots of technical details (the whole presentation is now available for download from the SOI Consortium website – click here to get it). A key point is that FD-SOI will scale to 7nm. Here’s the slide that says it all:
Also, be sure to check out the Cadence presentation when it’s posted – it looks at the solid design methodology now in place.
Following a brief mea culpa acknowledging that ARM had been missing too long from the FD-SOI table, GM of the Physical Design group Will Abbey made it clear that they are now fully onboard. In his talk, “Realize the Potential of FD-SOI”, he said in comparisons between 22nm FD-SOI and 14nm FinFET, they see a lot of space for FD-SOI. Here’s his summary slide:
They are now looking at ways to further optimize back-biasing to decrease total power in block-level implementations. And yes, he said, you’ll get performance that’s close to FinFET.
Fabless innovator Sigma Designs is focused on the connected home (especially smart TV and media connectivity) and IoT. CEO Thinh Tran presented, “Enabling the Digital Connected World with FDSOI” – you can download it here.
If you really want to optimize for power efficiency, use FD-SOI and run at 0.4V, he advised. “I’m very excited about this,” he told the San Jose audience, adding that, “It’s especially good for RF.” Here’s his slide that explains why:
So, it was a great day in San Jose for 22nm and 28nm FD-SOI. Be sure to keep checking back at the SOI Consortium website, as more presentations will become available in the days to come.
~ ~ ~
*NRE = non-recurring engineering. In a fabless scenario, there are NRE for IP and design (engineering costs, up-front and royalty-based IP costs), NRE for masks and fabrication (mask costs, wafer prototype lots, tools costs, probe cards, load-boards and other one-time capital expenditures), and NRE for qualifications (ESD, latch-up and other industry-specific qualifications, as in automotives).
The #1 take-away message from the recent FD-SOI Symposium in San Jose is that “FD-SOI is the smart path to success”. With presentations echoing that theme by virtually all the major players – including (finally!) ARM – to a packed house, it really was an epic day for the FD-SOI ecosystem. The presentations are now starting to be available on the SOI Consortium website – click here to see them (they’re not all there as of today, though, so keep checking back).
Since there’s so much to cover, we’ll break this into two parts. This is Part 1, focusing on presentations related to some of the exciting products that are hitting the market using 28nm FD-SOI. Part 2 will focus on the terrific presentations related to 22nm FD-SOI. In future posts we’ll get into the details of many of the presentations. But for now, we’ll just hit the highlights.
So back briefly to FD-SOI being smart. (A nice echo to the Soitec FD-SOI wafer manufacturing technology – SmartCutTM – that make it all possible right?) It started with the CEO of Sigma Designs (watch for their first IoT products on FD-SOI coming out soon) quipping, “FD-SOI is the poor man’s FinFET.” To which GlobalFoundries’ VP Kengeri riffed that really, “FD-SOI is the smart man’s FinFET”. And NXP VP Ron Martino, summed it up saying, “FD-SOI is the smart man’s path to success”. Yes!
Samsung now has a strong 28nm FD-SOI tape-out pipeline for 2016, and interest is rising fast, said Kelvin Low, the company’s Sr. Director of Foundry Marketing. His presentation title said it all: “28FDS – Industry’s First Mass-Produced FDSOI Technology for IoT Era, with Single Platform Benefits.” They’ve already done 12 tape-outs, are working on 10 more now for various applications: application processor, networking, STB, game, connectivity,…., and see more coming up fast and for more applications such as MCU, programmable logic, IoT and broader automotive. It is a mature technology, he emphasized, and not a niche technology. The ecosystem is growing, and there’s lots more IP ready. 28nm will be a long-lived node. Here’s the slide that summed up the current production status:
As you see, the production PDK with the RF add-on will be available this summer. Also, don’t miss the presentations by Synopsys (get it here), which has repackaged the key IP from ST for Samsung customers, Leti on back-bias (get it here), Ciena (they were the Nortel’s optical networking group) and ST (it’s chalk-full of great data on FD-SOI for RF and analog).
If you read Ṙon’s terrific posts here on ASN recently, you already know a lot about where he’s coming from. If you missed them, they are absolute must-reads: here’s Part 1 and here’s Part 2. Really – read them as soon as you’re done reading this.
As he noted in his ASN pieces, NXP’s got two important new applications processor lines coming out on 28nm FD-SOI. The latest i.MX 7 series combines ultra-low power (where they’re dynamically leveraging the full range of reverse back biasing – something you can do only with FD-SOI on thin BOX) and performance-on-demand architecture (boosted when and where it’s needed with forward back-biasing). It’s the first general purpose microprocessor family in the industry’s to incorporate both the ARM® Cortex®-A7 and the ARM Cortex-M4 cores (the series includes single and dual A7 core options). The i.MX 8 series targets highly-advanced driver information systems and other multimedia intensive embedded applications. It leverages ARM’s V8-A 64-bit architecture in a 10+ core complex that includes blocks of Cortex-A72s and Cortex-A53s.
In his San Jose presentation, Ron said that FD-SOI is all about smart architecture, integration and differentiating techniques for power efficiency and performance. And the markets for NXP’s i.MX applications processors are all about diversification, in which a significant set of building blocks will be on-chip. The IoT concept requires integration of diverse components, he said, meaning that a different set of attributes will now be leading to success. “28nm FD-SOI offers advantages that allows scaling from small power efficient processors to high performance safety critical processor,” he noted – a key part of the NXP strategy. Why not FinFET? Among other things, it would bump up the cost by 50%. Here are other parts of the comparison he showed:
For NXP, FD-SOI provides the ideal path, leading to extensions of microcontrollers with advanced memory. FD-SOI improves SER* by up to 100x, so it’s an especially good choice when it comes to automotive security. Back-biasing – another big plus – he calls it “critical and compelling”. The icing on the cake? “There’s so much we can do with analog and memory,” he said. “Our engineers are so excited!”
You know how using mapping apps on your smartphone kills your battery? Well now there’s hope. Sony’s getting some super impressive results with their new GPS using 28nm FD-SOI technology. These GPS are operated at 0.6V, and cut power to 10x (!) less than what it was in the previous generation (which was already boasting the industry’s lowest power consumption when it was announced back in 2013).
In San Jose, Sony Senior Manager Kenichi Nakano presented, “Low Power GPS design with RF circuit by the FDSOI 28nm”, proclaiming with a smile, “I love FD-SOI, too!” All the tests are good and the chip is production ready, he said. In fact, they’ve been shipping samples since March.
As of this writing, his presentation is not yet posted. But til it is, if you’re interested in the background of this chip, you can check out the presentation he gave in Tokyo in 2015 here.
SERDES (Serializer/Deserializer) IP is central to many modern SOC designs, providing a high-speed interface for a broad range of applications from storage to display. It’s also used in high-speed data communications, where it’s had a bad rep for pulling a huge amount of power in data centers. But Analog Bits has been revolutionizing SERDES IP by drastically cutting the power. Now, with a port to 28nm FD-SOI, they’re claiming the industry’s lowest power.
In his presentation, “A Case Study of Half Power SERDES in FDSOI”, EVP Mahesh Tirupattur described FD-SOI as a new canvas for chip design engineers. The company designs parts for multiple markets and multiple protocols. When they got a request to port from bulk to 28nm FD-SOI, they did it in record time of just a few months, getting power down to 1/3 with no extra mask steps. Plus, they found designing in FD-SOI to be cheaper and easier than FinFET, which of course implies a faster time to market. “The fabs were very helpful,” he said. “I’m pleased and honored to be part of this ecosystem.”
Listening to a presentation by Stanford professor Boris Murmann gets you a stunning 30,000 foot view of the industry through an amazing analog lens. He’s lead numerous explorations into the far reaches of analog and RF in FD-SOI, and concludes that the technology offers significant benefits toward addressing the needs of: ultra low-power “fog” computing for IoT (it’s the next big thing – see a good Forbes article on it here); densely integrated, low-power analog interfaces; universal radios; and ultra high-speed ADC. Get his symposium presentation, “Mixed-Signal Design Innovations in FD-SOI Technology” here.
So, it was a great day in San Jose for 28nm FD-SOI. Next in part 2, we’ll look at why it was also an epic day for 22nm FD-SOI. Be sure to keep checking back at the SOI Consortium website, as more presentations will become available in the days to come.
~ ~ ~
*SER = Soft Error Rates – soft errors occur when alpha or neutron particles hit memory cells and change their state, giving an incorrect read. These particles can either come from cosmic rays, or when radioactive atoms are released into the chips as materials decay.
Registration is open for GlobalFoundries’ technical webinar, “How to Implement an ARM Cortex-A17 Processor in 22FDX 22nm FD-SOI Technology” (click here to go to the registration page). The webinar will cover the optimal steps to successfully implement ARM® Cortex®-A Series* processors using 22FDXTM 22nm FD-SOI technology.
GF Design Enablement Fellow Dr. Joerg Winkler will address:
This webinar will take place April 26, 2016 at10:00 am Pacific Time.
BTW, GF’s already done quite a few 22FDX-related webinars and videos – click here to see the current list.
~ ~ ~
* Per ARM, “Cortex-A processors are specifically designed to execute complex functions and applications such as those required by consumer devices like smartphones and tablets. Their performance efficiency is also making them an increasingly popular choice for servers and enterprise applications where large core clusters can be combined for optimal solutions.”
By Ronald M. Martino, Vice President, i.MX Applications Processor and Advanced Technology Adoption, NXP Semiconductors
At NXP, we’re very excited about the prospects for our new i.MX 7 and 8 series of applications processors, which we’re manufacturing on 28nm FD-SOI.
As noted in part 1 of this article series, the new i.MX 7 series, which leverages the 32-bit ARM v7-A core, is targeting the general embedded, e-reader, medical, wearable and IoT markets, where power efficiency is paramount. The i.MX 8 series leverages the 64-bit ARM v8-A series, targeting automotive applications, especially driver information systems, and well as high-performance general embedded and advanced graphics applications.
Choosing an FD-SOI solution gave our designers some specific tools that helped them to more easily and robustly deliver the features our customers are looking for. Here in part 2, we’ll look a little more deeply into the markets each of these chip families is targeting, and the role FD-SOI plays in helping us meet our specs.
Announced last June, the first members of our new 7 series — the i.MX 7Solo and i.MX 7Dual product families — will be hitting the market shortly. We’ve been shipping samples since last year, and the response has been tremendous. (You can read about the i.MX 7 IoT ecosystem we’re helping create for our customers here and support for wearable markets here.)
Our i.MX 7 customers are building products for power- and cost-sensitive markets. That of course includes a vast array of innovative IoT solutions and wearables, but also solutions for other parts of the embedded market like handheld point-of-sale (POS) and medical devices, smart home controls and industrial products. The i.MX 7 series also continues NXP’s industry leading support for the e-reader market via integration of an advanced, fourth-generation EPD controller.
For all these markets, excellent performance is very important, but both dynamic and static power figures are really key. When you’re creating a system with power efficient processing and low-power deep sleep modes, you enable a new tier of performance-on-demand, battery-operated devices that are lighter and cheaper, and in a virtuous cycle require smaller batteries.
The next members of the NXP i.MX 7 series combine ultra-low power (dynamically leveraging the reverse back biasing you can do with FD-SOI) and performance-on-demand architecture (boosted when needed with FD-SOI’s forward back-biasing). It’s the industry’s first general purpose microprocessor family to incorporate both the ARM® Cortex®-A7 and the ARM Cortex-M4 cores (customers can choose between single or dual A7 cores). These technologies, together with our new companion PF3000 power management IC, unleash the potential for dramatically innovative, secure and power efficient end-products for wearable computing and IoT applications.
The initial offering of i.MX 7 was designed (on 28nm bulk) with Cortex-A7 cores operating up to 1 GHz, while the Cortex-M4 core operates at up to 200 MHz. The Cortex-A7 and Cortex-M4 achieve processor core efficiency levels of 100 microWatts (μW) /MHz and 70 μW /MHz respectively.
A Low Power State Retention (LPSR), battery-saving mode can be improved by FD-SOI and consumes only 250 μW, representing a 3x improvement over our previous generation (on 40nm bulk). That’s almost 50% better than our competitors. Plus it minimizes wake up times without requiring Linux reboot, while supporting DDR self-refresh mode, GPIO wakeup, and memory state retention.
The next members of the i.MX 7 series, with FD-SOI dynamic back-biasing, enable different blocks to be reverse or forward back-biased on the fly to attain always-optimal power savings or performance. Additional power optimization features are enabled to achieve leadership power efficiency. We’ve optimized FD-SOI dynamic back-biasing to enable performance-on-demand architecture through which the i.MX 7 series meets the bursty, high-performance needs (this is when forward back-biasing kicks in) of running Linux, graphical user interfaces, high-security technologies like Elliptic Curve Cryptography, as well as wireless stacks or other high-bandwidth data transfers with one or multiple Cortex-A7 cores.
When high levels of processing are not needed, low-power modes kick in with reverse back biasing of the critical subsystems, and the ongoing, real-time work is carried on by the smaller, lower powered Cortex-M4.
All things considered, it’s perhaps no surprise that we expect i.MX 7 series solutions for cost-sensitive markets to be a key driver of our long-term i.MX portfolio expansion.
Our new i.MX 8 series portfolio, based on 28nm FD-SOI process technology, targets highly-advanced driver information systems and other multi-media intensive embedded applications. It incorporates those same key attributes as the i.MX 7, but extends them into realms the industry has never experienced. We believe the i.MX 8 series is poised to revolutionize interactivity in multimedia and display applications across all kinds of industries.
i.MX 8 incorporates innovations in the processor — complex graphics, vision, virtualization and safety to help revolutionize interactivity for a wide range of uses in many, many markets. The capabilities of this family is broad, but one of the places it’s going to be the biggest game-changer is in what is becoming the e-cockpit of your car.
For almost two decades, SOI has shone in the embedded processing world. In addition, NXP counts every major automotive maker in the world amongst its customers for our devices. Entering the new e-cockpit frontier, 28nm FD-SOI is the logical choice in making the i.MX 8 series meet and exceed the stringent requirements of top automotive OEMs for years to come.
The i.MX 8 series leverages ARM’s V8-A 64-bit architecture in a 10+ core complex that includes blocks of Cortex-A72s and Cortex-A53s. All the FD-SOI advantages discussed above for the i.MX 7 are also being brought to bear here (the power envelope for automotive designers being extremely strict). But in the hot and electrically noisy automotive environment, FD-SOI also plays an important role in ensuring robust operation.
The way we see it, your car’s multimedia centric e-cockpit will revolve around the i.MX 8, a single chip that drives all displays from infotainment to heads-up-displays (HUD) to instrument clusters. It’s optimized for the intelligent transfer of data and information management from multiple subsystems within the IC – as opposed to only delivering raw performance through one or two processing blocks.
For drivers and passengers alike, we’re looking at a very different world: one that includes the spread of advanced heads-up displays, intuitive gesture control, natural speech recognition, augmented reality, enhanced convenience and device connectivity. (I wrote a blog exploring the possibilities last fall – you can read it here.)
And of course, it will be secure from hackers, and fail-safe for critical systems.
From our customers’ standpoint, they can design a single hardware platform and scale it across multiple market segments with the unique approach to pin and software compatibility within the i.MX product families.
The i.MX family has been leveraged in over 35 million vehicles since it was first launched in vehicles in 2010. So with all these new features, and low-power and robust performance, we see a very bright future for FD-SOI and the i.MX 8 in automotive. It’s going to be a great ride.
By Ronald M. Martino, Vice President, i.MX Applications Processor and Advanced Technology Adoption, NXP Semiconductors
The latest generations of power efficient and full-featured applications processors in NXP’s very successful and broadly deployed i.MX platform are being manufactured on 28nm FD-SOI. The new i.MX 7 series leverages the 32-bit ARM v7-A core, targeting the general embedded, e-reader, medical, wearable and IoT markets, where power efficiency is paramount. The i.MX 8 series leverages the 64-bit ARM v8-A series, targeting automotive applications, especially driver information systems, as well as high-performance general embedded and advanced graphics applications.
Over 200 million i.MX SOCs have been shipped over six product generations since the i.MX line was first launched (by Freescale) in 2001. They’re in over 35 million vehicles today, are leaders in e-readers and pervasive in the general embedded space. But the landscape for the markets targeted by the i.MX 7 and i.MX 8 product lines are changing radically. While performance needs to be high, the real name of the game is power efficiency.
The bottom line in chip manufacturing is always cost. A move from 28nm HKMG to 14nm FinFET would entail up to a 50% cost increase. Would it be worth it? While FinFETs do boast impressive power-performance figures, for applications processors targeting IoT, embedded and automotive, we need to look beyond those figures, taking into account:
In fact, both NXP and the former Freescale have extremely deep SOI expertise. Freescale developed over 20 processors based on partially-depleted SOI over the last decade; and NXP, having pioneered SOI technology for high-voltage applications, has dozens of SOI-based product lines. So we all understand how SOI can help us strategically leverage power and performance. For us, FD-SOI is just the latest SOI technology, this time with a design flow almost identical to bulk, but on ultra-thin SOI wafers and some important additional perks like back-biasing.
When all the factors we care about for the new i.MX processor families are tallied up, FD-SOI comes out a clear winner for i.MX SOCs.
For our designers, here’s why FD-SOI is the right solution to the engineering challenges they faced in meeting evolving market needs.
In terms of power, you can lower the supply voltage (Vdd) – so you’re pulling less power from your energy source – and still get excellent performance. Add to that the dynamic back-biasing techniques (forward back-bias improves performance, while reverse back-bias reduces leakage) available with FD-SOI (but not with FinFETs), you get a very large dynamic operating range.
By dramatically reducing leakage, reverse back-biasing (RBB) gives you good power-performance at very low voltages and a wide range of temperatures. This is particularly important for IoT products, which will spend most of their time in very low-power standby mode followed by short bursts of performance-intense activity. We can meet the requirements for those high-performance instances with forward back-biasing (FBB) techniques. And because we can apply back-biasing dynamically, we can specify it to meet changing workload requirements on the fly. [Editor’s note: click here and here for helpful ASN articles with descriptions and discussions of back-biasing, which is also sometimes called body-biasing.]
Devices for IoT also have major analog and RF elements, which do not scale nearly so well as the digital parts of the chip. Furthermore analog and RF elements are very sensitive to voltage variations. It is important that the RF and analog blocks of the chip are not affected by the digital parts of a chip, which undergo strong, sudden signal switching. The major concerns for our analog/RF designers include gain, matching, variability, noise, power dissipation, and resistance. Traditionally they’ve used specialized techniques, but FD-SOI makes their job much easier and results in superior analog performance.
In terms of RF, FD-SOI greatly simplifies the integration of RF blocks for WiFi, Bluetooth or Zigbee, for example, into an SOC.
Soft error rates (SER)* are another important consideration, especially as the size and density of SOC memory arrays keep increasing. Bulk technology gets worse SER results with each technology node, while FD-SOI provides ever better SER reliability with each geometry shrink. In fact, 28nm FD-SOI provides 10 to 100 times better immunity to soft-errors than its bulk counterpart.
Our process development strategy has always been to leverage foundry standard technology and adapt it for our targeted applications, with a focus on differentiating technologies for performance and features. We typically reuse about 80% of our technology platform, and own our intellectual property (IP). Looking at the ease of porting existing platform technology and IP, and analyzing die size vs. die cost, again, FD-SOI came out the clear choice.
In terms of manufacturing, FD-SOI is a lower-risk solution. Integration is simpler, and turnaround time (TAT) is much faster. 28nm FD-SOI is a planar technology, so it’s lower complexity and extends our 28nm installed expertise base. Throughout the design cycle, we’ve worked closely with our foundry partner, Samsung. They provided outstanding support, and very quickly reached excellent yield levels, which is of course paramount for the rapid ramp we anticipate on these products.
In the second part of this article, we’ll take a look at the new i.MX product lines, and why FD-SOI is helping us make those game-changing plays for specific markets.
~ ~ ~
* Soft errors occur when alpha or neutron particles hit memory cells and change their state, giving an incorrect read. These particles can either come from cosmic rays, or when radioactive atoms are released into the chips as materials decay.
The SOI Consortium has lined up an excellent, comprehensive FD-SOI Symposium on April 13th in San Jose. They’ll be highlighting the tremendous progress of the FD-SOI ecosystem. Headliners include Cisco, Sony, NXP, SigmaDesigns, ARM, Ciena plus the big FD-SOI foundries, EDA companies, design partners, chipmakers and analysts. There is a special session dedicated to RF and analog design innovation on FD-SOI with STMicroelectronics, Stanford and others. In short, we’re going to get a chance to see the FD-SOI ecosystem in action.
To attend, all you have to do is register in advance – click here to go to the registration page. It’s free and open to everyone who registers.
08:00AM – 09:00AM – Registration
08:55AM – 09:00AM – Welcome by Carlos Mazure, SOI Consortium
09:00AM – 09:30AM – Aglaia Kong, Cisco Systems, CTO for Internet of Everything
09:30AM – 10:00AM – Thinh Tran, Sigma Designs, CEO
10:00AM – 10:30AM – Ron Martino, NXP, VP, Application Processors & Advanced Technology Adoption
10:30AM – 10:50AM – Coffee Break
10:50AM – 11:20AM – Subramani Kengeri, GLOBALFOUNDRIES, VP CMOS Business Unit
11:20AM – 11:50AM – Will Abbey, ARM, GM Physical IP
11:50AM – 12:20PM – Kelvin Low, Samsung Semiconductor, Senior Director, Foundry Marketing
12:20PM – 1:40PM Lunch
1:40PM – 2:10PM – Kenichi Nakano, SONY, Sr. Manager, Analog LSI Business Division
2:10PM – 2:40PM – Dan Hutcheson, VLSI Research, CEO
2:40PM – 3:05PM – Mahesh Tirupattur, Analog Bits, EVP
3:05PM – 3:30PM – Mike McAweeney, Synopsys, Sr. Director, IP Division
3:30PM – 4:00PM – Coffee Break
4:00PM – 4:30PM – Naim Ben-Hamida, Ciena, Senior Manager
4:30PM – 4:55PM – Rod Metcalfe, Cadence, Group Director, Product Engineering
4:55PM – 5:20PM – Prof. Boris Murmann, Stanford, on “Mixed-Signal Design Innovations in FD-SOI Technology”
5:20PM – 5:45PM – Frederic Paillardet, STMicroelectronics, Sr. Director, RF R&D
5:45PM – 6:00PM – Ali Erdengiz, CEA-LETI, Silicon Impulse
6:00PM – 6:05PM – Closing remarks by Giorgio Cesana, SOI Consortium
Please note that if you’ve already registered last month when the first announcement went out, the location has changed. The SOI Consortium FD-SOI Symposium will be held on Wednesday, 13 April 2016, from 8am to 6:30pm at the:
Doubletree Hotel San Jose
2050 Gateway Place
San Jose, California 95110, USA
If you can’t make it, not to worry – ASN will be there taking notes for a round-up and follow-up articles. Plus we’ll be tweeting and retweeting (follow us on Twitter at @FollowASN and @AdeleHars – look for the hashtag #FDSOI). And of course you’ll want to follow the Twitter feeds of participating companies, and of the SOI Consortium @SOIConsortium.org.
This post was first published as part of Paul McLellan’s new Breakfast Bytes blog on the Cadence website.
~ ~ ~
Cadence recently put out a press release that the Cadence implementation flow had been qualified on the GLOBALFOUNDRIES 22FDX process. At ARM TechCon, Joerg Winkler and Tamer Ragheb from the design enablement group of GLOBALFOUNDRIES in Dresden, Germany, provided a lot more detail. With German precision, their talk was titled The Implementation of ARM® Cortex®-A17 Quad-Core in GLOBALFOUNDRIES 22FDX Technology Using Cadence Innovus Implementation System. But I didn’t need to wait, I got a 1:1 meeting with Joerg earlier in the week and we went into more detail.
One thing that I had been confused about that Joerg clarified when I asked him is whether double patterning is used. I knew that the metal pitch was 80nm and so, in principle, could be single patterned. But that is only true if the layer is patterned in a single direction (vertical or horizontal but not both). For metal1 and metal2 they wanted to have both directions so that Ls and Ts could be made inside the standard cells, which meant that they needed to use double patterning.
At some level, the details of the process don’t affect the implementation flow that much. The transistors are inside the standard cells and other blocks, so whether they are planar, FinFET or FD-SOI is secondary to where the pins are, how the coloring affects placement, and so on. So the basic flow through Genus, Innovus and the various signoff engines is unchanged from any other process.
One area where FD-SOI is very different, as I went through in detail in my earlier blog, is the forward and reverse body bias (fbb/rbb). This is a voltage applied to the back of the thin buried oxide (the box) that doesn’t turn the transistor on or off (the box is too thick, and the bias can only change slowly due to the high capacitance). However, it does affect the performance of the transistor. This allows various tradeoffs: lower the voltage to reduce power, and then speed it up again with fbb; reduce leakage in a hibernating IoT device with rbb. Basically fbb increases the performance and this can be taken purely as increased performance, or as reduced power at the same performance. And rbb reduces performance but also decreases leakage, so it can be used when high performance is not required but power is critical.
The challenge that Joerg and his team faced was to come up with an architecture for how to connect up the bias. It is a little like planning a power grid. There are local decisions as to how to actually connect, what layer of metal to use and so on. Then there are block-level issues such as how to distribute the signals without creating huge blockages for routing. There isn’t really a chip-level issue like for power since, except perhaps for test chips, the bias is not expected to be externally applied through the package pins, but rather generated internally on the chips with charge pumps and enabled/disabled under software control. The highest level decision is to partition the chip into areas where different biases can be applied—the same bias is not needed everywhere.
The test vehicle chosen was a Quad-Core ARM Cortex-A17 processor. They decided to create five areas where the bias could be controlled independently, each of the four cores and then everything else, which notably includes the L2 cache and its controller. The libraries used were an 8-track standard cell library from Invecas (GF’s IP development partner) with continuous RX and support for body biasing. The cache memories were built from GF evaluation memory kit, with 14 different L1 cache memory macros, 1 L2 cache memory macro with support for body biasing of the bitcell array and for the memory periphery. An additional complication is that these areas also need to support power down (so cores can be powered off completely as well as biased). The body bias is all specified in the IEEE 1801 power file (so that all the other tools can handle the power policy chosen) and in the script that drives the Innovus Implementation System during physical design to actually create the connections.
The body bias nets needed to be connected up to dedicated pins on the well-tap cells, the power switches and the memory macros. The cells were carefully aligned so that the body bias connections were straight runs of metal, and then a body-bias net ring was placed around the perimeter of the module as is often done with power nets. See the diagram below.
The actual connections can be seen in the diagram below. The 10 yellow lines running across are the 10 body bias signals for the 5 regions (they are in pairs, one for P transistors and one for N). The green vertical lines in the middle pick up the appropriate pair of bias signals and these, in turn, are connected to the actual well tap cells (where the signals effectively connect to the back gate).
In a very similar way, the bias for the memory is also picked up from a ring and then run through the core of the memories using straight runs of metal.
GLOBALFOUNDRIES also created an implementation of the smaller ARM Cortex-A9 processor. They have been using this microprocessor for several technology nodes to allow a comparison of power, performance, and area (PPA). In this particular case they wanted to compare performance at different body biases compared to the 28SLP process. The result is that 22FDX with fbb has 30% higher frequency at the same power (along with a 45% area reduction) and with rbb has 45% power reduction at the same frequency (and obviously the same 45% area reduction). This means that the implementation can vary over a huge range of power/performance with the same silicon, whereas at 28nm it would require a complete re-implementation (which is why it appears as a single red dot rather than a curve).
~ ~ ~
This post was first published under the new Breakfast Bytes blog on the Cadence website. The original is here. Many thanks to the folks at Cadence and to Paul McLellan for permission to repost it here on ASN.