Tech Guides - Embedded Systems

Hot Chips 31, the premiere event for the biggest semiconductor vendors to highlight their latest architectural developments is held in August every year. The event this year was held at the Memorial Auditorium on the Stanford University Campus in California, from August 18-20, 2019. Since its inception it is co-sponsored by IEEE and ACM SIGARCH. Hot Chips is amazing for the level of depth it provides on the latest technology and the upcoming releases in the IoT, firmware and hardware space. This year the list of presentations for Hot Chips was almost overwhelming with a wide range of technical disclosures on the latest chip logic innovations. Almost all the major chip vendors and IP licensees involved in semiconductor logic designs took part: Intel, AMD, NVIDIA, Arm, Xilinx, IBM, were on the list. But companies like Google, Microsoft, Facebook and Amazon also took part. There are notable absences from the likes of Apple, who despite being on the Committee, last presented at the conference in 1994. Day 1 kicked off with tutorials and sponsor demos. On the cloud side, Amazon AWS covered the evolution of hypervisors and the AWS infrastructure. Microsoft described its acceleration strategy with FPGAs and ASICs, with details on Project Brainwave and Project Zipline. Google covered the architecture of Google Cloud with the TPU v3 chip.  And a 3-part RISC-V tutorial rounded off by afternoon, so the day was spent well with insights into the latest cloud infrastructure and processor architectures. The detailed talks were presented on Day 2 and Day 3, below are some of the important highlights of the event: IBM’s POWER10 Processor expected by 2021 IBM which creates families of processors to address different segments, with different models for tasks like scale-up, scale-out, and now NVLink deployments. The company is adding new custom models that use new acceleration and memory devices, and that was the focus of this year’s talk at Hot Chips. They also announced about POWER10 which is expected to come with these new enhancements in 2021, they additionally announced, core counts of POWER10 and process technology. IBM also spoke about focusing on developing diverse memory and accelerator solutions to differentiate its product stack with heterogeneous systems. IBM aims to reduce the number of PHYs on its chips, so now it has PCIe Gen 4 PHYs while the rest of the SERDES run with the company's own interfaces. This creates a flexible interface that can support many types of accelerators and protocols, like GPUs, ASICs, CAPI, NVLink, and OpenCAPI. AMD wants to become a significant player in Artificial Intelligence AMD does not have an artificial intelligence–focused chip. However, AMD CEO Lisa Su in a keynote address at Hot Chips 31 stated that the company is working toward becoming a more significant player in artificial intelligence. Lisa stated that the company had adopted a CPU/GPU/interconnect strategy to tap artificial intelligence and HPC opportunity. She said that AMD would use all its technology in the Frontier supercomputer. The company plans to fully optimize its EYPC CPU and Radeon Instinct GPU for supercomputing. It would further enhance the system’s performance with its Infinity Fabric and unlock performance with its ROCM (Radeon Open Compute) software tools. Unlike Intel and NVIDIA, AMD does not have a dedicated artificial intelligence chip or application-specific accelerators. Despite this, Su noted, “We’ll absolutely see AMD be a large player in AI.” AMD is considering whether to build a dedicated AI chip or not. This decision will depend on how artificial intelligence evolves. Lisa explained that companies have been improving their CPU (central processing unit) performance by leveraging various elements. These elements are process technology, die size, TDP (thermal design power), power management, microarchitecture, and compilers. Process technology is the biggest contributor, as it boosts performance by 40%. Increasing die size also boosts performance in the double digits, but it is not cost-effective. While AMD used microarchitecture to boost EPYC Rome server CPU IPC (instructions per cycle) by 15% in single-threaded and 23% in multi-threaded workloads. This IPC improvement is above the industry average IPC improvement of around 5%–8%. Intel’s Nervana NNP-T and Lakefield 3D Foveros hybrid processors Intel revealed fine-grained details about its much-anticipated Spring Crest Deep Learning Accelerators at Hot Chips 31. The Nervana Neural Network Processor for Training (NNP-T) comes with 24 processing cores and a new take on data movement that's powered by 32GB of HBM2 memory. The spacious 27 billion transistors are spread across a 688mm2 die. The NNP-T also incorporates leading-edge technology from Intel-rival TSMC. Intel Lakefield 3D Foveros Hybrid Processors Intel in another presentation talked about Lakefield 3D Foveros hybrid processors that are the first to come to market with Intel's new 3D chip-stacking technology. The current design consists of two dies. The lower die houses all of the typical southbridge features, like I/O connections, and is fabbed on the 22FFL process. The upper die is a 10nm CPU that features one large compute core and four smaller Atom-based 'efficiency' cores, similar to an ARM big.LITTLE processor. Intel calls this a "hybrid x86 architecture," and it could denote a fundamental shift in the company's strategy. Finally, the company stacks DRAM atop the 3D processor in a PoP (package-on-Package) implementation. Cerebras largest chip ever with 1.2 trillion transistors California artificial intelligence startup Cerebras Systems introduced its Cerebras Wafer Scale Engine (WSE), the world’s largest-ever chip built for neural network processing. Sean Lie the Co-Founder and Chief Hardware Architect at Cerebras Lie presented the gigantic chip ever at Hot Chips 31. The 16nm WSE is a 46,225 mm2 silicon chip which is slightly larger than a 9.7-inch iPad. It features 1.2 trillion transistors, 400,000 AI optimized cores, 18 Gigabytes of on-chip memory, 9 petabyte/s memory bandwidth, and 100 petabyte/s fabric bandwidth. It is 56.7 times larger than the largest Nvidia graphics processing unit, which accommodates 21.1 billion transistors on a 815 mm2 silicon base. NVIDIA’s multi-chip solution for deep neural networks accelerator NVIDIA which announced about designing a test multi-chip solution for DNN computations at a VLSI conference last year, the company explained chip technology at Hot Chips 31 this year. It is currently a test chip which involves a multi-chip DL inference. It is designed for CNNs and has a RISC-V chip controller. It has 36 small chips, 8 Vector MACs per PE, and each chip has 12 PEs and each package has 6x6 chips. Few other notable talks at Hot Chips 31 Microsoft unveiled its new product Hololens 2.0 silicone. It has a holographic processor and a custom silicone. The application processor runs the app, and the HPU modifies the rendered image and sends to the display. Facebook presented details on Zion, its next generation in-memory unified training platform. Zion which is designed for Facebook sparse workloads, has a unified BFLOAT 16 format with CPU and accelerators. Huawei spoke about its Da Vinci architecture, a single Ascend 310 which can deliver 16 TeraOPS of 8-bit integer performance, support real-time analytics across 16 channels of HD video, and consume less than 8W of power. Xiling Versal AI engine Xilinx, the manufacturer of FPGAs, announced its new Versal AI engine last year as a way of moving FPGAs into the AI domain. This year at Hot Chips they expanded on its technology and more. Ayar Labs, an optical chip making startup, showcased results of its work with DARPA (U.S. Department of Defense's Defense Advanced Research Projects Agency) and Intel on an FPGA chiplet integration platform. The final talk on Day 3 ended with a presentation by Habana, they discussed about an innovative approach to scaling AI Training systems with its GAUDI AI Processor. AMD competes with Intel by launching EPYC Rome, world’s first 7 nm chip for data centers, luring in Twitter and Google Apple advanced talks with Intel to buy its smartphone modem chip business for $1 billion, reports WSJ Alibaba’s chipmaker launches open source RISC-V based ‘XuanTie 910 processor’ for 5G, AI, IoT and self-driving applications

Internet of Things (IoT) has gained a huge traction due to the ability to gather data from sensors embedded within a variety of IoT devices including Close-circuit cameras, vehicles, smart homes, smart appliances, and many more. Think of IoT as a network of devices which gathers raw and real-time data, analyzes them, and provides desired outputs that benefit the users. But what after the data is analyzed? What is done with the analyzed report? The data has to be acted upon. Here, Artificial Intelligence can do the needful. AI can get hold of all that data crunched by IoT devices and act on it in a successful and organized manner. Industries that already use IoT devices can automate certain mundane workflows such as documentation, machine maintenance notification alert, and so on when powered by AI. Intelligent things with AI-backed IoT The saying, ‘With great power come great responsibilities’, is true for AI powered IoT.AI backed IoT devices can make complex decisions, perform self-learning, and can carry out autonomous decision making. One can group IoT applications broadly into two categories based on who the end user is, i.e. Industrial IoT for enterprises and consumer IoT for individual consumers. Let’s look into some of the major domains that AI has enhanced. 1. Industrial IoT Also known as the IIoT, IoT has impacted industries by bringing in unprecedented opportunities. However, it has also brought in a wave of new risks to businesses. IIoT provides the internet with a new ability to control machines, factories and the industrial infrastructure. Some of the characteristics of IIoT include, Improved Interoperability where the machines and sensors communicate via IoT Availability of Transparent information with the presence of more sensors, which means abundance of information. Autonomous decision making now lies in the hands of the IoT devices, where they can detect emergency situations, for instance when a machine servicing is required and can act on it immediately.    Manufacturing Manufacturing is by far the biggest industry affected by the IoT wave. According to a report, ‘global manufacturers will invest $70 billion on IoT solutions in 2020, which is up from the $29 billion they spent in 2015’.Let’s see how some of the processes in manufacturing get a lift with AI enabled IoT: Detection of machine health using Predictive maintenance : Predictive maintenance involves collection and evaluation of data from machines in order to increase efficiency and optimize the maintenance processes. With predictive maintenance, manufacturers can determine the condition of their equipments and also predict when machines need maintenance. A startup named Konux, based in Munich, Germany, has developed a machine-learning powered monitoring system for train switches. The Konux switch sensor can be retrofitted onto existing train networks, providing real-time monitoring of track conditions and rolling stock. Data is transmitted wirelessly to the Konux Kora platform, which uses predictive algorithms based on machine learning to alert staff to specific problems as well as drive recommendations for maintenance. Supply Chain Optimization : With an IoT-optimized supply chain, manufacturers can get hold of real-time data and analyze issues to act upon them before the onset of any major problem. This in turn reduces inventory and capital requirements. In order to track a product, companies have set up smart shelves, which keep a record of when the product has been removed, the total no. of products, and so on. This smart shelf is connected to their entire network which is linked to their planning and demand sensing engine. Here, the AI powered decision support systems help to translate those demand signals into production and order processes. Read ‘How AI is transforming the manufacturing Industry’ for a more indepth look at AI’s impact on the manufacturing industry. Retail Adoption of IIoT in retail has upped the game for online retailers. Retail stores now comprise of in-store advertising and gesture walls. These walls help customers search merchandize, offers, and buy products with simple gestures. Retailers also have Automated Checkouts, or most simply self-checkout kiosks. This enables customers to avoid long queues and pay for products using a mobile app based payments system which scans the QR code embedded on the products, contactless payments or other means. With IoT enabled sensors, retailers can now extract insights about the most popular areas people pass by and where they stop to see the merchandize. Retailers can then send promotional text messages, discount coupons directly on the customer’s phone while they are in the store’s vicinity. For instance, Apple’s iBeacon enables devices to alert apps and websites about customer location. Retailers have also adopted Inventory Optimizations by using digital shelf and RFID techniques for managing their inventories effectively. Healthcare IoT in healthcare is proving to be a boon for patients by decreasing costs and reducing multiple visits to doctors. With these healthcare solutions, patient monitoring can be done in real-time. Due to this real-time data, diseases can be treated well in advance before they reach a malignant stage. These IoT enabled healthcare solutions provide accurate collection of data, automated workflows which are combined with data driven decisions.This cuts down on waste, reducing system costs and most importantly minimizes errors. Also, creation and management of drugs is a major expenditure in the healthcare industry. With IoT processes and devices, it is possible to manage these costs better. A new generation of “smart pills” is allowing healthcare organizations to ensure that a patient takes his or her medication, while also collecting other vital data. Apart from these major applications of IoT in the Industrial sectors, it has also affected sectors such as telecommunications, energy, and in the Government. Next up, we move on to explaining how AI backed IoT can affect and enhance the consumer domain. 2. Consumer IoT Consumers go for services that provide them with an easy way to do mundane tasks. Let us have a look at some examples where AI has intelligently assisted IoT for consumers benefit. Connected Vehicles Connected vehicles are vehicles that use any of a number of different communication technologies to communicate with the driver, other cars on the road (vehicle-to-vehicle [V2V]): This tech helps wirelessly exchange information about the speed and position of surrounding vehicles. This helps in avoiding crashes, ease traffic congestion, and improve the environment. roadside infrastructure (vehicle-to-infrastructure [V2I]): These technologies capture vehicle-generated traffic data wirelessly and provide information such as warnings from the infrastructure to the vehicle that inform the driver of safety, mobility, or environment-related conditions. the “Cloud” [V2C]: A Vehicle-to-Cloud infrastructure integrates NaaS (Network As A Service) into the automotive ecosystem and allows provisioning of vehicle-based services for automobile user. Connected homes These AI enabled IoT devices and services can automatically respond to preset rules, be remotely accessed and managed by mobile apps or a browser, and send alerts or messages to the user. For instance, Google Home, with a built-in Google Assistant, controls home and helps people with lists, translation, news, music, calendar and much more. Google Home can also answer any questions asked to it. This is because of Google’s huge Knowledge Graph that it is connected to. Similarly, Amazon’s Echo, a voice-controlled speaker and Apple’s homepod also assist in collecting data they get via voice. The AI can also get all devices within the home connected, with the help of Wi-Fi. With the latest IFTTT technology, your Google Home can talk to Nest and adjust the temperature of your home as per your requirement or the external temperature change. Health and lifestyle AI integrated with predictive analytics within the embedded devices such as fitness apps, health trackers, diet planners, and so on, makes them intelligent and personalized. For instance, Fitbit coach app paired with the Fitbit has a huge database. The app uses complex algorithms to extract meaningful information from the user data. This data is further used to recommend highly-tailored workout plans. Also, AthGene, uses ML algorithms to convert genetic information into valuable insights for customizing fitness regimen, diet plans, and lifestyle changes for users. IoT was only about devices monitoring data and giving insights in real-time. But AI added the efficiency factor, and also gave the power to these systems to take decisions. AI with IoT has a bright future; one can expect smart machines managed via Echo or Google Home in the future. Read Next How Google’s DeepMind is creating images with artificial intelligence Customer Relationship management just got better with Artificial Intelligence 5 Ways Artificial Intelligence is Transforming the Gaming Industry

The Internet of Things is going to transform the way we live in the future. It will change how we commute, how we work, even simple day to day activities. But one thing that’s often overlooked when we talk about the internet of things is how it will impact IT teams. We’ve seen a lot of change in the shape of the modern IT team over the last 10 years thanks to things like DevOps, but IoT is going to shape things further in the near future.  To better understand how the Internet of Things will shape IT teams in the future, we first need to understand the application of the Internet of Things, especially in the sector closest to IT teams, the enterprise sector. IoT in the enterprise sector If you look at consumer media, the most common applications of the Internet of Things are the small-scale ones like smart gadgets and smart home systems. Unfortunately, this class of IoT products hasn't really caught up with mainstream consumers; its audience is limited to hobbyists and people in the tech. However, it's a whole different story with the enterprise sector becuse companies all over the world are starting to realize the benefit of applying IoT in their line of business.  Different industries have different applications of IoT. Usually though, IoT is used to either increase efficiency or reduce cost. For example, a shipping service may apply a monitoring system on their vehicles to track their speed and mileage to find ways to reduce fuel usage. Similarly, an airline company could apply sensors on their fleet of airplanes to monitor engine conditions to maintain it properly. A company may also apply IoT to manage its energy consumption so that it can reduce unneeded expenses. What new skills does IoT demand of tech pros All of these applications of IoT are going to require new skills and maybe even new job roles. So while we’ll see efficiencies thanks to these innovations, to really make an impact its still going to need both personal and organizational investment in skills and knowledge to ensure IoT is really helping to drive positive change. IoT and the second data explosion Let’s start with the most obvious change – the growth of data. Yes, the big data explosion has been happening all around us for the last decade, but IoT is bringing with it a second explosion that will be even bigger. This means everyone is going to have to become more data-savvy. That’s not to say that everyone will need to moonlight as a data scientist, but they will need an awareness of how data is stored and processed, who needs access to it and who needs to act on it. Device management will become more important than ever IoT isn’t just about data. It’s also about devices. With more gadgets and sensors connected to a given network, device management and maintenance will be an essential part of the IT team’s work. To tackle this problem, the team will need to grow bigger to handle more work, or they will need to use a more powerful device management tool that can handle a big amount of connected devices. New security risks presented by IoT An increase in the number of connected devices also presents increased security risks. This means pressure will be on IT departments to  IT team will need to tighten up security. Managing networks is one part of that, but a further challenge will be managing the human side of security – ensuring good practice is followed by staff and taking steps to minimize social engineering threats. IT teams will have to customize IoT solutions to meet their needs IoT doesn’t yet have many standards. That means today’s organizations face opportunities and challenges in how they customize solutions and tools for their own needs. This can be daunting, but for people working in IT teams it’s also really exciting – it gives them more control and ownership of the work they are doing. Third party solutions will no doubt remain, but they won’t be quite so important when it comes to IoT. True, companies like IBM will be working on IoT solutions right now to capture the market; however, because these innovations are in their infancy there’s a limit on traditional technology corporations’ ability to shape and define the IoT landscape in the way they have done with innovations in the past.  And that's just a small bit of how the Internet of Things will affect the IT team. When IoT takes off, it will change our lives in the most unimaginable ways possible, so of course there will be even more changes that will happen with the IT teams in charge of this. But then again, the world of technology is ripe with changes and disruptions, so I'm sure we're all used to changes and will be able to adapt. Raka Mahesa is a game developer at Chocoarts who is interested in digital technology in general. Outside of work, he enjoys working on his own projects, with Corridoom VR being his latest released game. Raka also regularly tweets @legacy99.

Many cloud service providers have been marching towards adopting IoT (Internet of Things) services to attract more customers. This league includes top cloud merchants such as AWS, Microsoft Azure, IBM, and much recently, Google. Among these, Amazon Web Services have been the most popular. Its AWS IoT Core service is a fully-managed cloud platform that provides IoT devices with an easy and secure connection to interact with cloud applications and other IoT devices. AWS IoT Core can keep track of billions of IoT devices, with the messages travelling to and from them. It processes and routes those messages to the AWS endpoints and to other devices reliably and securely. This means, with the help of AWS IoT Core, you can keep track of all your devices and have a real-time communication with them. Undoubtedly, there is a lot of competition around cloud platforms to host IoT services. Users are bound to a specific cloud platform for a varied set of reasons such as a yearly subscription, by choice, or other reasons. Here are 5 reasons to choose AWS IoT core for your IoT projects: Build applications on the platform of your choice with AWS IoT Core Device SDK AWS IoT Core Device SDK is the primary mode of connection between your application and the AWS IoT core. It uses the MQTT, HTTP, or webSockets protocols to effectively connect and exchange messages with this service. The languages supported by the AWS IoT device SDK are C, Arduino, and JavaScript. The SDK provides developers with mobile SDKs for Android and iOS, and a bunch of SDKs for Embedded C, Python and many more. It also includes open-source libraries, developer guides with samples, and porting guides. With these features, developers can build novel IoT products and solutions on the hardware platform of their choice. AWS IoT Summit 2018 held recently in Sydney shed light on cloud technologies and how it can help businesses lower costs, improve efficiency and innovate at scale. It had sessions dedicated to IoT. (Intelligence of Things: IoT, AWS DeepLens, and Amazon SageMaker) Handle the underlying infrastructure and protocol support with Device Gateway The device gateway acts as an entry gate for IoT devices to connect to the Amazon Web Services (AWS). It handles multiple protocols, which ensures secure and effective connection of the IoT devices with the IoT Core. The list of protocols include MQTT, WebSockets, and HTTP 1.1. Also, with the device gateway, one does not have to worry about the infrastructure as it automatically manages and scales huge amount of devices at ease. Authentication and Authorization is now easy with AWS methods of authentication AWS IoT Core supports SigV4, an AWS method of authentication, X.509 certificate based authentication, and customer created token based authentication. The user can create, deploy and manage certificates and policies for the devices from the console or using the API. AWS IoT Core also supports connections from users’ mobile apps using Amazon Cognito, which creates a unique ID for app users and can be used to retrieve temporary, limited-privilege AWS credentials. AWS IoT Core also enables temporary AWS credentials after a device has authenticated with an X.509 certificate. This is done so that the device can more easily access other AWS services such as DynamoDB or S3. Determine device’s current state automatically with Device Shadow Device shadow is a JSON document, which stores and retrieves the current state for a device. It provides persistent representations such as the last reported state and the desired future state of one’s device even when the device is offline. With Device Shadow, one can easily build applications to interact with the applications by providing REST APIs. It aids applications to set their desired future state without having to request for device starting state. AWS IoT core differentiates between the desire state and the last reported state. It can further command the device to make up the difference. Route messages both internally and externally using AWS Rules Engine The Rules Engine helps build IoT applications without having to manage any infrastructure. Based on the rules defined, the Rules engine evaluates all the incoming messages within the AWS IoT Core, transforms it, and delivers them to other devices or cloud services. One can author or write rules within the management console using the SQL-like syntax The Rules Engine can also route messages to AWS endpoints such as AWS Lambda, Amazon Kinesis, Amazon S3, Amazon Machine Learning, Amazon DynamoDB, Amazon CloudWatch, and Amazon Elasticsearch Service with built-in Kibana integration. It can also reach external endpoints using AWS Lambda, Amazon Kinesis, and Amazon Simple Notification Service (SNS). There are many other reasons to choose AWS IoT Core for your projects. However, it is purely one’s choice as many might already be using or bound to other cloud services. For those, who haven’t yet started, they may choose AWS for a plethora of other cloud services that they offer, which includes AWS IoT Core too.