ultra-low-power

The GAP9, GreenWaves Technologies latest IoT application processor -- which is being fabbed on GlobalFoundries 22FDX (FD-SOI) technology -- will be sampling in the first half of 2020, according to EETimes (read the whole article here). Mass production is slated for 2021. Greenwaves (which has been an SOI Consortium member for several years now) is a fabless semiconductor startup designing disruptive ultra-low power embedded solutions for AI processing in sensing devices at the very edge. GreenWaves marketing director Martin Croome told EETimes, “We are using the body biasing ability in FD-SOI to allow us to achieve even lower power consumption.” Compared to GreenWaves’ currently shipping product, GAP8 (which is on a 55nm bulk process), GAP9 reduces energy consumption by 5 times while enabling inference on neural networks 10 times larger. This is thanks to architectural enhancements and the move to GF's 22FDX semiconductor process. The new chip delivers a peak cluster memory bandwidth of 41.6 GB/sec and up to 50 GOPS combined compute power at an overall power consumption of 50mW. It enables customers to embed machine learning and signal processing capabilities into battery operated or energy harvesting devices such as IoT sensors in smart building, consumer and industrial markets and consumer and medical wearable devices. GAP9 was showcased at the last RISC-V Summit in San Jose (read the full press release here). [caption id="attachment_29061" align="alignnone" width="400"] GAP9 Block diagram (Courtesy: GreenWaves)[/caption] Some of the (many!) features include: 10 identical high performance, extended ISA, RISC-V ISA cores (cluster of 9 cores for compute-intensive tasks and a fabric controller core for control and communication) Dynamic voltage frequency scaling and automatic body biasing Multiple power states: deep sleep, deep sleep with retentive RAM, low activity, SOC on, SOC on cluster on Click here for a full GAP9 product brief.

Here is our second post about the SOI Consortium’s Japan Symposium this past fall. This will provide summaries of eight very informative presentations on SOI in IoT and automotive by NXP, Dolphin Design, Leti, Silvaco, Arm, I-fuse and Secure-IC. There’s a lot of content to summarize, so this post is about twice as long as those we usually do. But you’ll want to read right to the end, for sure! In case you missed our previous post on the 5G/RF-SOI presentations given at the Japan event, you can read it here. Our next and final post on the Japan event will cover photonics presentations by Cisco/Luxtera, TowerJazz, GlobalFoundries, Leti, Cadence and Soitec. By way of reminder, the Japan SOI Symposium was a great success, with both days well attended. If your company is a member of the SOI Consortium, you can now access most of these presentations on our website. You can also click on the illustrations in this post to see them in enlarged versions. [caption id="attachment_28106" align="alignleft" width="366"] (Courtesy: NXP SOI Consortium)[/caption] The IoT World Enabled Through SOI - Jon Cheek, NXP Sr. Director, Front-End Innovation For NXP, FD-SOI introduced the ability to easily add different functionalities to the technology node like ULP, eNVM, support for high-voltage and embedding RF. For them, said Cheek, it’s about the range, and with adaptive back bias, you can “get crazy”, so you can really achieve amazing things. In fact, they think they now have the lowest leakage SRAM in the industry, thanks to body biasing. The i.MX 7ULP is finding significant success in wearables. Their “crossover” chips are the latest beneficiaries of FD-SOI with body biasing. The “new normal”, they offer huge improvements for real-time operating systems – which is of course key for edge computing. (As you can imagine, the audience was intently taking notes throughout -- this was a really excellent talk!) It also is great for machine learning, as it is designed to unlock the potential of voice-assisted end nodes. The IP they needed is now available from multiple vendors, noted Cheek, such as Tensilica and VeriSilicon. Another key play will be in visuals for industrial computing. He concluded by observing that the automobile is the ultimate IoT machine, with 10x the amount of code now found in leading edge airplanes. That’s where the i.MX8 and 8X come in. [caption id="attachment_28104" align="alignright" width="323"] (Courtesy: NXP SOI Consortium)[/caption] High-Voltage SOI – Enabling Automotive- NXP Jon Cheek gave this presentation on the second day of the Japan event. Long-time followers of SOI will know that NXP has been excelling in high-voltage (HV) SOI for well over two decades now (including the pioneering work done by Philips, now part of NXP: their EZ-HV SOI patent dates back to 1993). It’s probably safe to say that NXP's SOI-based automotive chips are used by virtually every carmaker on the planet. HV follows well behind the leading edge – it’s currently mostly around 130nm (the limits are related to metalization). Reason #1 it’s on SOI? SOI-based technologies are incredibly reliable, especially in the automotive culture targeting the three zeros (0 emissions, 0 accidents and 0 time wasted). Today’s car manufacturer’s are going to a distributed environment, and SOI still provides a huge advantage, making parts that are smaller, lower power and more reliable – so it drives a lower BOM for automakers.In conclusion, said Cheek, NXP’s leadership through SOI innovation enables scalable solutions, high voltage analog integration, sensor integration, and reliable safe passenger experience. [caption id="attachment_28101" align="alignleft" width="432"] (Courtesy: Dolphin Design SOI Consortium)[/caption] Improving SoC Energy-Efficiency with Dolphin Design Platforms – Nicolaus Gaude, BizDev Product Marketing, Dolphin Design Dolphin has a series of platforms, techniques and IP for increasing speed and drastically improving energy efficiency in SoC design. Gaude introduced their Speed Platforms, which include a Power Management Platform and a Processing platform, both of which make dramatic improvements in energy efficiency. The Power Management Platform keeps control of power management from architecture to design, resulting in a 10x improvement in energy efficiency. The Processing Platform comprises configurable RTL clusters for best-in-class (100x) energy-efficiency. Gaude then turned to the Dolphin’s Adaptive-Body Bias (ABB) IP for breakthrough energy-efficiency with FD-SOI. This is real-time, “on-the-fly” body biasing: the IP does it all. It is silicon-proven on GlobalFoundries’ 22FDX with Arm cores and Invecus standards cells SRAM, with breakthrough energy efficiency. [caption id="attachment_28108" align="alignright" width="363"] (Courtesy: Silvaco SOI Consortium)[/caption] Platform Infrastructure for SOI-IP Ecosystem – Thomas Blaesi, VP of Global Marketing, Silvaco The massive use of IP is both an advantage and a challenge, began Blaesi. There are solutions out there, but they are disconnected. Typically SoC/IP designers, IP librarians and support folks use various systems, while procurement, finance and legal use others. This is a problem for both the providers and the consumers of IP. Silvaco has a new system called Xena that centrally organizes all IP data: it’s an IP repository for tracking accounts, products, contracts, devices, support, compliance and reporting. One of the first beneficiaries of Xena will be the SOI ecosystem, as providers of SOI IP are already signing on. Beyond the organizational advantages, Xena has patented “finger printing” and “DNA analysis”, so there is a digital representation of each IP on an SoC that can’t be reverse engineered. Each fingerprint contains list of unique signatures of each file in an IP or SoC. A file’s unique signature is created from the entire file content, and that signature is guaranteed to be unique to that content. It enhances support for all versions of common design files: hard IP, soft IP, and embedded software. Because it’s cloud or enterprise based, it will be particularly useful for large organizations. Fingerprinting and DNA analysis are vendor agnostic, universal, and easy-to-use tools and methodologies for IP lifecycle management, he concluded. [caption id="attachment_28103" align="alignleft" width="463"] (Courtesy: Leti SOI Consortium)[/caption] Ultra-low power, FD-SOI based IP, in the space of IoT, Health Care, Smart Connectivity 5G – Michael Tchagaspanian, EVP Industrial Partnerships, CEA-Leti This presentation began with a review of the explosion in devices with IoT and related investments, then connected all the ways in which innovations powerhouse Leti is contributing – from the SOI wafer level to the chip level – which is to say practically everywhere! Especially hot topics in FD-SOI included: the roadmap to sub-10nm; CoolCube monolithic 3D; new embedded memories; power amplifiers; Ultra-Wide Range DSP; smart sensing local processing (including haptics, imaging, infrared advanced processing); local processing with edge AI; and spike coding for deep neural networks. He showed information on two always-on/on-demand transmission 28nm FD-SOI IoT test chips that taped out in mid-2019: the Warrior and the Samurai. And finally, he covered silicon-proven IP that Leti has for FD-SOI including power management blocks, lots of RF IP (including low-power RF wake-up), sensor interfaces, clockless network-on-chip and new SRAM technologies. These and more will be covered at the next Leti Innovation Days in Grenoble (June 2020) – during which in parallel, btw, there will also be a European SOI Summit hosted by the SOI Consortium. [caption id="attachment_28099" align="alignright" width="475"] (Courtesy: Arm SOI Consortium)[/caption] FDSOI Enablement for a Total Compute Future – Manuj Rahor, Director Emerging Technologies Product Marketing, Arm Subtitled A perspective on system optimization with Arm FDSOI IP, this presentation reviewed how Arm is enabling system gains through optimization across IP boundaries. This is work happening in the Arm Artisan Physical Design Group (PDG), which provides logic, memory and POP (processor optimization package) IP as well as various products to help ease implementation challenges for advanced nodes. In this case the focus is on Total Compute enablers on Samsung 28nm FD-SOI (called 28FDS) – specifically three building blocks recently launched on FD-SOI. The first is the 128Mb Wide Capacity embedded MRAM (an eNVM to replace eFlash) compiler for storage delivered to Samsung in July `19. It was demonstrated in silicon in the Musca-S1 Smart IoT Device Demonstrator on 28FDS, an energy efficient IoT device with eMRAM secure boot on-chip storage. [Read our coverage from March 2019 here.] The second is a novel design developed with Spin Memory. It recently taped out on 28FDS and is slated for delivery in 2020. Adding an “Endurance Engine to the eMRAM that was delivered in 2019, the ARM-Spin innovation delivers RAM-like performance with increased speed and endurance. What’s at issue here is a change in use cases. Use cases served by eFlash were not written to that often; now with sensors (as in IoT) that continually gather and write data, eFlash endurance is not sufficient. The third is billed as an SRAM replacement compiler. Its MRAM as RAM in A-class systems, with significant energy and performance gains. Again, this is a use-case issue: retention is lower (this is for weeks months, whereas the other solutions are for 10 years). But you can get more RAM than SRAM into the same footprint, so you get a 60% reduction in DRAM traffic and increased performance. Delivery for this is marked as 2020+. [caption id="attachment_28100" align="alignleft" width="294"] (Courtesy: Attopsemi SOI Consortium)[/caption] I-fuse™: A Disruptive OTP Technology – Dr. Shine Chung, Chairman, Attopsemi I-fuse is a disruptive OTP (One-Time Programmable) technology without disrupting a fuse. The goal was a 100x increase in reliability at 1/100th of the cell size and 1/10th the power. It has now been demonstrated in GlobalFoundries’ 22FDX FD-SOI technology for energy harvesting applications. In the OTP IP technologies, explained Dr. Chung, they defied the conventional wisdom of breaking a fuse to maintain a permanent programmed state forever: Attopsemi’s I-fuse™ is actually a “non-breaking” fuse. “I don’t mind to break a fuse, but I do care about breaking a fuse by explosion”, said Dr. Chung. “The I-V curve of programming a fuse beyond the break point actually shows more like an explosion. The anti-fuse OTP also ruptures gate oxide by explosion. On the contrary, I-fuse™ is a disruptive OTP technology without disrupting a fuse.” He concluded, “By using MOS as switches to enable discharging two capacitors, through cell and reference cell respectively, and compare the discharge rates, the resistance in the cell can be determined higher or lower than the reference resistance so as to convert into logic data. The read energy consumed is only 1/100 of the conventional sensing, which is good for energy harvest IoT applications. Eventually most IoT devices will be battery-less.” [caption id="attachment_28107" align="alignright" width="398"] (Courtesy: Secure-IC SOI Consortium)[/caption] AIoT Embedded Security Using FD-SOI – Yan-Taro Clochard, Japan Sales Director, Secure-IC In addition to opportunities, the impact of AI on IoT (aka AIoT) adds new threats to edge devices. Design for security and in-depth security is required, down to the physical layer. For example in automotive, sensors gather data and AI analyzes it – but the enabler is security. The challenge of AI is the increase in data and connectivity with unsecured devices. FD-SOI is a key for Secure-IC’s Securyzer security module: it leveragesFD-SOI properties to secure the AIoT world. It is flexible, and tuned for each customer. Here, FD-SOI enables the creation of physically secure systems, with secure boot and firmware updates, cryptographic services, key management and secure storage.

[caption id="attachment_12359" align="alignright" width="300"] (Courtesy: PRNewsfoto/QuickLogic Corporation)[/caption] Some great pieces of FD-SOI news from QuickLogic. The company recently demonstrated its ultra-low power ArcticPro™ embedded FPGA (eFPGA) solutions at the GlobalFoundries Technology Conferences in Santa Clara, California, Munich and Shanghai. The technology is available now. ArcticPro is the industry's first eFPGA offering for GF's 22FDX® process (btw they've been shipping it in volume for GF's 65nm and 40nm bulk processes for years). The company says its ultra-low power eFPGA architecture and mature software offer semiconductor and system companies the ability to integrate programmable hardware accelerators to lower power consumption and the flexibility to reconfigure a device's functionality in the field. [caption id="attachment_12360" align="alignleft" width="300"] (Image courtesy: QuickLogic)[/caption] QuickLogic has also announced that the technical university ETH Zurich will integrate QuickLogic's ArcticPro technology onto the university's PULP platform. PULP is a silicon-proven open-source parallel platform for ultra-low power computing created with the objective of delivering high compute bandwidth combined with high-energy efficiency. ETH will become the first licensee of eFPGA technology from QuickLogic on GF's 22FDX process node. They will develop an SoC integrating ETHZ's open-source RISC-V cores and eFPGA technology, enabling users to offload critical functions from the processor(s) and implement them in eFPGA fabric. This approach creates multiple hardware co-processors that increase system efficiency and performance while decreasing power consumption. "The main goal of the PULP program is to use a multi-disciplinary approach to achieve extremely high-power efficiency for computing applications," said QuickLogic CTO Dr. Timothy Saxe. "QuickLogic has a tremendous depth of experience in achieving low power consumption across a broad range of applications, including AI and IoT at the edge and security, and we look forward to contributing what we've learned along with our eFPGA technology to this groundbreaking initiative in low power computing." ETH's PULP platform with the fully integrated eFPGA is expected to be available Q1' 2019. QuickLogic is part of GF's fast-growing FDXcelerator™ partner ecosystem, offering customers ultra-low power (eFPGA) Intellectual Property, complete software tools and a compiler.

Here’s why the embedded community should care whether the chips they use are built on FD-SOI. FD-SOI has “…dramatically improved the landscape for power efficiency,” NXP VP Joe Yu explains in a recent Embedded Systems Engineering piece (you can read it here). He gets into the hows and whys of the i.MX7ULP chip design, taking a deep dive into the things that the embedded folks really care about. He details how FD-SOI decreases leakage and dynamic power, including the roles played by forward and reverse body biasing. He then goes on to explain why it’s better for analog, and how it prevents latch-up. FD-SOI enables new features, too, he points out, like ultra-low power consumption and deep sleep suspend. And perhaps most importantly, he explains how bursty high-performance and ultra energy efficiency are dynamically traded off on an as-needed basis. “Engineers no longer face a forced selection: low-power processor or high-performance processor,” he say. “Rather, the selection for performance or power efficiency can be made instantaneously, as needed, without having to reconfigure.” All of this plus the rich graphics and user interface FD-SOI enables makes the i.MX 7ULP perfect for “…IoT edge devices, as well as smart home controls, building automation, portable patient monitoring, wearables, and portable scanners.” This is an excellent read: highly recommended. Of course, ASN covered the i.mX7ULP when it was first announced (on Samsung's 28nm FD-SOI) last year – you can still read our coverage here. But it’s good to see the company explaining to their customers how FD-SOI will change the way they build products. BTW, you can get all the i.MX7ULP product details on the NXP website here. NXP has also put together a nifty video on the i.MX7ULP – see it here.

By: Tamer Ragheb,Digital Design Methodology Technical Manager at GlobalFoundries and Josefina Hobbs, Senior Manager of Strategic Alliances, Synopsys It’s clear that getting an optimal balance of power and performance at the right cost is foremost in the minds of designers today. Designers who want either high performance or ultra low-power, or ideally both, have a choice to make when it comes to migrating to next generation nodes. For applications that push the envelope in performance, FinFET would be the optimal solution. For applications that require ultra low-power and more RF integration, FD-SOI is the right solution. The two technologies have different value propositions that need to be considered while designing for applications ranging from high-performance computing and server to high-end mobile and Internet of Things (IoT). GlobalFoundries 22FDX is the industry’s very first 22nm FD-SOI platform. The 22FDX technology is specifically designed to meet the ultra low-power requirements of the next generation of connected devices. The big advantage of this platform is its ability to provide software control at the transistor level through flexible body-biasing (Fig. 1). The ability to provide real-time trade-offs between power and performance via software-controlled body-biasing of the transistor creates new options for the designer. For example, imagine designing a processor for a Smartwatch that could match its power-performance tradeoff to your typical use and modify its performance based on how you’re using it that day. [caption id="attachment_9473" align="alignleft" width="610"] Figure 1: Benefits of 22FDX body-biasing[/caption] The full impact of the body bias capability of 22FDX becomes clear when compared to incumbent high-performance process technologies (Fig. 2). 22FDX compared to a 28nm high K metal gate (HKMG) technology can provide up to 50% less power at the same frequency, or 40% faster performance at the same total power than 28HKMG. In addition, 22FDX can be further optimized with forward body bias, shown on the blue curve, to further reduce the power or to further boost the speed in a turbo operation mode. [caption id="attachment_9474" align="alignleft" width="610"] Figure 2: 22FDX Body Bias Optimizes Performance and Power[/caption] In addition to the body bias, 22FDX offers capabilities for design flexibility and intelligent control that are not available in other technologies. These include: Improved electrostatic control of the transistor acts as a performance booster and enables lower VDD (i.e., lower power consumption) while reaching significant performance Low variability and body-biasing capability that can achieve 0.4 volt operation Complete RF enablement with ‘knobs’ to reduce RF power by up to 50 percent Manufacturing success is highly sensitive to specific physical design features, with advanced nodes requiring more complex design rules and more attention to manufacturability issues on the part of designers. However, there are essentially no additional manufacturing requirements to design in 22FDX beyond what is required for 28nm designs. There are four application optimized extensions available with 22FDX (Fig. 3). These are: 22FDX ULP- an ultra low-power extension that provides logic libraries and memory compilers that are optimized for 0.4 volt operation. 22 FDX ULL- an ultra low-leakage extension that brings in an expanded device suite capable of achieving one pico-amp per micron leakage. 22 FDX UHP- an ultra high-performance extension that leverages the overdrive capabilities and body-biasing features to maximize the performance of technologies in a turbo or a burst mode. It has high performance libraries and high speed interfaces and BEOL stacks optimized for competing architectures or applications. 22 FDX RFA- an RF and analog extension that brings in full characterization and enablement for RF applications, including optimized RF layouts and P cells, BEOL passives, and IP for Bluetooth LE and WIFI applications. [caption id="attachment_9475" align="alignleft" width="610"] Figure 3: 22FDX Platform and Extensions[/caption] GlobalFoundries reference flow for 22FDX has been optimized to support forward and reverse body bias (FBB/RBB), which provides the design flexibility to optimize the performance/power trade-offs. The reference flow supports implant-aware and continuous diffusion-aware placement, tap insertion and body bias network connectivity according to high voltage rules, double-patterning aware parasitic extraction (PEX), and design for manufacturing (DFM). This provides designers with the flexibility to manage power, performance and leakage targets for the next-generation chips used in mainstream mobile, IoT and networking applications. GlobalFoundries has been collaborating with Synopsys to enable and qualify their tools for the 22FDX Reference Flow. The recent qualification of Synopsys’ Galaxy™ Design Platform for the current version ofGlobalFoundries’ 22FDX technology allows the designer to manage power, performance and leakage and achieve optimal energy efficiency and cost effectiveness. Synopsys’ Galaxy Design Platform supports body biasing techniques throughout the design flow, including both forward and reverse body bias, enabling power/performance trade-offs to be made dynamically and delivering up to 50% power reduction. Key tools and features of the Galaxy Design Platform in the 22FDX reference flow include: Design Compiler® Graphical synthesis with IEEE 1801 (UPF) driven bias-aware multi-corner multi-mode (MCMM) optimization Formality® formal verification with bias-aware equivalence checking IC Compiler™ and IC Compiler II™ layout with physical implementation support for non-uniform library floorplanning, implant-aware placement, multi-rail routing, and advanced power mesh creation StarRC™ parasitic extraction for multi-rail signoff with support for multi-valued standard parasitic exchange format (SPEF) PrimeTime® timing analysis and signoff including distributed multi-scenario analysis (DMSA) static timing and noise analysis, using AOCV and POCV technology IC Validator In-Design physical verification The 22FDX technology leverages existing design tools such as the Galaxy Design Platform, manufacturing infrastructure and the broader design ecosystem. This speeds time to market and enables the creation of differentiated products.