Memory, or the ability to store information in a readily accessible way, is an essential operation in computers and human brains. A key difference is that while brain information processing involves performing computations directly on stored data, computers shuttle data back and forth between a memory unit and a central processing unit (CPU). This inefficient separation (the von Neumann bottleneck) contributes to the rising energy cost of computers.Since the 1970s, researchers have been working on the concept of a memristor (memory resistor); an electronic component that can, like a synapse, both compute and store data. But Aleksandra Radenovic in the Laboratory of Nanoscale Biology (<a href="https://www.epfl.ch/labs/lben/" target="_blank">LBEN</a>) in EPFL’s School of Engineering set her sights on something even more ambitious: a functional nanofluidic memristive device that relies on ions, rather than electrons and their oppositely charged counterparts (holes). Such an approach would more closely mimic the brain’s own – much more energy efficient – way of processing information.“Memristors have already been used to build electronic neural networks, but our goal is to build a nanofluidic neural network that takes advantage of changes in ion concentrations, similar to living organisms,” Radenovic says.“We have fabricated a new nanofluidic device for memory applications that is significantly more scalable and much more performant than previous attempts,” says LBEN postdoctoral researcher Théo Emmerich. “This has enabled us, for the very first time, to connect two such ‘artificial synapses’, paving the way for the design of brain-inspired liquid hardware.”The research has <a href="https://www.nature.com/articles/s41928-024-01137-9" target="_blank" rel="noopener">recently been published in Nature Electronics.</a><h2>Just add water</h2>Memristors can switch between two conductance states – on and off – through manipulation of an applied voltage. While electronic memristors rely on electrons and holes to process digital information, LBEN’s memristor can take advantage of a range of different ions. For their study, the researchers immersed their device in an electrolyte water solution containing potassium ions, but others could be used, including sodium and calcium.“We can tune the memory of our device by changing the ions we use, which affects how it switches from on to off, or how much memory it stores,” Emmerich explains.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1712102720846-6b9cf0f3.jpg" style="width: 100%px;" class="fr-fic fr-dib">Illustration of a HAC circuit. 2023 EPFL/Andras Kis CC BY SA 4.0The device was fabricated on a chip at EPFL’s&nbsp;<a href="https://www.epfl.ch/research/facilities/cmi/" target="_blank">Center of MicroNanoTechnology</a> by creating a nanopore at the center of a silicon nitride membrane. The researchers added palladium and graphite layers to create nano-channels for ions. As a current flows through the chip, the ions percolate through the channels and converge at the pore, where their pressure creates a blister between the chip surface and the graphite. As the graphite layer is forced up by the blister, the device becomes more conductive, switching its memory state to ‘on’. Since the graphite layer stays lifted, even without a current, the device ‘remembers’ its previous state. A negative voltage puts the layers back into contact, resetting the memory to the ‘off’ state.“Ion channels in the brain undergo structural changes inside a synapse, so this also mimics biology,” says LBEN PhD student Yunfei Teng, who worked on fabricating the devices – dubbed highly asymmetric channels (HACs) in reference to the shape of the ion flow toward the central pores.LBEN PhD student Nathan Ronceray adds that the team’s observation of the HAC’s memory action in real time is also a novel achievement in the field. “Because we were dealing with a completely new memory phenomenon, we built a microscope to watch it in action.”By collaborating with Riccardo Chiesa and Edoardo Lopriore of the <a href="https://www.epfl.ch/labs/lanes/" target="_blank">Laboratory of Nanoscale Electronics and Structures</a>, led by Andras Kis, the researchers succeeded in connecting two HACs with an electrode to form a logic circuit based on ion flow. This achievement represents the first demonstration of digital logic operations based on synapse-like ionic devices. But the researchers aren’t stopping there: their next goal is to connect a network of HACs with water channels to create fully liquid circuits. In addition to providing an in-built cooling mechanism, the use of water would facilitate the development of bio-compatible devices with potential applications in brain-computer interfaces or neuromedicine.<hr>ReferencesEmmerich, T., Teng, Y., Ronceray, N. et al. Nanofluidic logic with mechano–ionic memristive switches. Nat Electron (2024). https://doi.org/10.1038/s41928-024-01137-9

Co-authors Nathan Ronceray and Théo Emmerich. 2023 EPFL/Titouan Veuillet CC BY SA 4.0

In a step toward nanofluidic-based neuromorphic – or brain-inspired – computing, EPFL engineers have succeeded in executing a logic operation by connecting two chips that use ions, rather than electrons, to process data.

Artificial nanofluidic synapses can store computational memory

Encouraged by her family, Lavender Tessmer explored various creative pursuits from a young age, particularly textiles, including knitting and crocheting. When she came to MIT, she figured that working with textiles would remain just a hobby; she never expected them to become integral to her career path.However, when she interviewed for a research assistant position in <a href="https://selfassemblylab.mit.edu/" target="_blank">Self Assembly Lab</a>, it just so happened that the lab had recently received funding from the Advanced Functional Fabrics of America, one of the <a href="https://obamawhitehouse.archives.gov/the-press-office/2016/06/20/fact-sheet-president-obama-announces-winner-new-smart-manufacturing" target="_blank">manufacturing institutes launched during the Obama administration</a>, for a textile-based project.Tessmer, now a fifth-year doctoral student in design and computation within the School of Architecture of Planning, took on the project, working with <a href="https://architecture.mit.edu/people/skylar-tibbits" target="_blank">Skylar Tibbits</a>, associate professor of design research, and <a href="https://cee.mit.edu/people_individual/caitlin-mueller/" target="_blank">Caitlin Muller</a>, associate professor in building technology. “At MIT, my interest in textiles really exploded and became the center of everything,” Tessmer says.While textiles may appear commonplace, the Covid-19 pandemic underscored the need for textile products in safeguarding our general health and safety, particularly through the filtration necessary for masks. Recognizing the importance of manufacturing capabilities for textiles, Tessmer’s research has focused on programming textiles with specific functional properties while also considering the feasibility of large-scale manufacturing of such products.<h2>A nonlinear path to MIT</h2>Tessmer studied music as an undergraduate student at Duquesne University, pursuing a passion that bloomed as a high schooler. One assignment opened her eyes to a different career path: She was told to compare a piece of music to some other artistic medium. Through this assignment, she discovered the world of architecture by underscoring the systematic nature of both disciplines, emphasizing the need for repetition and structure to unleash creativity. “I immediately realized that’s what I want to do,” she says.Tessmer switched gears and decided to devote the year after college to architecture, instead of auditioning for music ensembles. She says, “I always liked making things, and then, with architecture, I realized that you can make things as part of your profession.” She relied on the basic drafting skills that her father had taught her, and channeled these into building her architecture portfolio.Ultimately, she decided to pursue a master’s degree in architecture at Washington University in St. Louis. She graduated with her master’s at the end of the 2007 economic recession, a time when jobs in architecture were scarce. She eagerly accepted a part-time role teaching at WashU. Over the next five years, this role evolved into a full-time lecturer role, where she taught students while also independently establishing her own design practice and leading various installation design projects. Fittingly, all of the installations were inspired by textiles. “They were these high-performance carbon-fiber braided structures that we hand-made into large-scale braided nets with specific geometries,” Tessmer explains.<h2>“Squeezing everything” out of graduate school</h2>Teaching at WashU was a great experience, but the practice-oriented nature of the architecture department motivated Tessmer to seek complementary perspectives on design. “I wanted a totally new venue that was supportive of research and pushing the boundaries of design. I wanted to see what other approaches were out there,” she says. As her interests continued to grow in that direction, she learned that MIT has some renowned researchers in the field. She decided to apply for a master’s degree in architecture studies, and ultimately a doctorate in design and computation, within the School of Architecture and Planning.MIT’s program stood out to Tessmer because of the interdisciplinary approach of the architecture department. She says, “If you are an architect or designer, it is not strange to end up in a class full of people who are not architects, and that’s totally normal and even expected.” The integrated nature of her program is a shift from her previous academic experiences, where each discipline had been distinct and separate. She also values the lack of hierarchy between different disciplines within the architecture department here. “There is respect across disciplines for the contribution from each participant,” she says.As an older student, Tessmer has a slightly different approach to graduate school, compared to her peers. She says, “MIT is amazing because there is so much variety and so many things that you can get involved in. But my style is to be hyperfocused on my interests. For me, there have been huge benefits to focusing on this specific thing and squeezing everything I can out of it, even in the face of all of these other opportunities.”Tessmer has devoted herself to several projects throughout grad school, but all share a common thread: an emphasis on fiber development and textile programming. As a master’s student in the Self Assembly Lab, she utilized the inherent properties of materials and optimized their configurations for specific functions by integrating computation into the material itself. “At MIT, I learned a much broader definition of computation,” she says. “For example, in the Self Assembly Lab, we believe that material is a storage format of information and that you can program material to behave in certain ways.” &nbsp;The first project Tessmer worked on was designing a fiber that could respond to temperature fluctuations. Another project focused on embedding many different properties within a single fabric, potentially for astronauts. “The human body is so varied in the number of properties that you need to match,” she says. In conjunction with collaborators across multiple MIT departments, she designed a spacesuit sleeve with embedded padding, stretchable areas, a compression gradient, and various sensors. Her third project has focused on embedding shapechange behavior into fabric structures to enhance human comfort or fit, as an alternative to manual tailoring. Finally, in a return to her architectural roots, she is also working on designing a reinforced concrete beam using textiles, a more sustainable solution to building with concrete, which has a significant carbon footprint.Another crucial aspect of Tessmer’s research is her focus on the feasibility of large-scale manufacturing for a product. She regularly relies on industrial-scale machinery and consults with manufacturing partners. She says, “The way research is being conducted in the lab is a close parallel to how it would be made in real life. The potential for a direct bridge between one and the other is a high priority for me and a constraint that I have tried to layer on to all of my projects.”<h2>Dabbling in entrepreneurship</h2>Tessmer says with a laugh, “My entire hobby [textiles] has now been absorbed into my research. So I am in the market for a new hobby.” For now, that hobby has taken the form of entrepreneurship. She has been exploring the commercialization potential of her technologies, having filed multiple patents and completed the <a href="https://engine.xyz/blueprint" target="_blank">Blueprint</a> program with The Engine Accelerator. She hopes that one day her method for embedding properties in textiles, while also reducing manufacturing process steps, will be used for commercial fabrics.As an example, she points to shoe manufacturing. “Your shoes are normally an assembly of lots of different materials and lots of different layers. Instead, my proposal to The Engine focused on embedding all of these properties in an automated way, eliminating the need for an extensive assembly process.” Tessmer envisions entrepreneurship as one of her potential future paths.For the time being, however, she plans to remain in academia. “From the outside, being a professor seems like an unattainable position. However, I keep being surprised at my ability to get to the next level of the academic hierarchy.” She aims to integrate all her past experiences into a future research career, designing textiles within an architectural context, while also weaving in the constraints of manufacturing scalability.

PhD student Lavender Tessmer has devoted herself to several projects throughout grad school, but all share a common thread: an emphasis on fiber development and textile programming. “At MIT, my interest in textiles really exploded and became the center of everything,” she says. Photo: Gretchen Ertl

PhD student Lavender Tessmer applies computation to create textiles that behave in novel ways.

Programming functional fabrics

Nvidia’s CUDA cores are specialized processing units within Nvidia graphics cards designed for handling complex parallel computations efficiently, making them pivotal in high-performance computing, gaming, and various graphics rendering applications.

Understanding Nvidia CUDA Cores: A Comprehensive Guide

Artificial intelligence is rapidly becoming a transformative force across all sectors of the economy, and industrial engineering is no exception. This technological frontier reshapes how industries approach efficiency, maintenance, and automation. In response, Seeed Studio has recently launched its ReComputer Industrial series of artificial intelligence (AI) devices. Revealed on September 29th, these devices represent a significant advancement in industrial AI capabilities.<a href="https://www.seeedstudio.com/blog/2022/09/30/seeed-studio-announces-recomputer-based-on-new-nvidia-jetson-orin-nano-to-accelerate-edge-ai-from-development-to-production/#:~:text=,as%20announced%20at%20NVIDIA%20GTC" target="_blank">1</a>&nbsp;Incorporating NVIDIA®&nbsp;Jetson Xavier NX, Orin Nano, and Orin NX modules and leveraging the NVIDIA Ampere™ GPU architecture offer AI computational performance from 20 TOPS to 100 TOPS. As of May 16, 2023, Seeed Studio continues to enhance these devices, emphasizing its commitment to evolving AI solutions.Integrating AI into industrial operations marks a shift towards enhanced efficiency, precision, and automation. ReComputer Industrial series devices combine robust performance with industrial adaptability. They come equipped with the Jetpack 5.1 SDK, featuring NVIDIA technologies and CUDA-X accelerated libraries, ready for complex AI tasks, including video analytics and natural language processing. Their fanless construction guarantees consistent performance even in extreme conditions, reducing the likelihood of component failure due to dust and other environmental factors.Engineers aiming to incorporate AI into industrial systems will find that the ReComputer Industrial series AI devices fulfill stringent technical specifications and comply with industry certifications such as FCC, CE, RoHS, and UKCA. This compliance is essential for applications across various sectors, including smart cities, security, industrial automation, smart factories, and medical imaging.<h3 dir="ltr">Technical Specifications</h3>The Seeed Studio ReComputer Industrial series AI devices are equipped with an NVIDIA Jetson Xavier NX, Orin Nano, or Orin NX module, which is at the core of the module's functionality. These modules are part of the NVIDIA Ampere GPU architecture, offering AI performance that ranges from 20 TOPS to 100 TOPS. The devices' hardware is designed to operate reliably in a temperature range from -20°C to 60°C, suitable for challenging industrial environments.On the software front, the ReComputer devices come with the Jetpack 5.1 SDK pre-installed. This software development kit includes NVIDIA technologies, and CUDA-X accelerated libraries, ensuring compatibility with various AI applications. The devices support various industry-standard interfaces, including RJ-45 GbE, RS-232/RS-422/RS-485, USB 3.2, DI/DO, CAN, and TPM2.0 (module optional), providing versatile connectivity options for different industrial needs.The ReComputer's fanless design is a testament to its robust construction and ability to reduce maintenance needs and extend the device's operational lifespan by minimizing the risk of dust and contaminant ingress.For further details and specifications, visit the <a href="https://www.mouser.com/new/seeed-studio/" target="_blank" rel="noopener noreferrer">datasheets&nbsp;</a>provided by Seeed Studio and Mouser Electronics.<h2 dir="ltr">Importance of Technology</h2>AI in modern industrial applications enhances efficiency, precision, and innovation. AI's role is to interpret vast data streams, optimize operations, and facilitate predictive maintenance, thereby reducing downtime and extending the life of industrial equipment.&nbsp;ReComputer devices, such as the ReComputer J1020 v2, embody this integration by providing a compact, powerful edge AI solution that brings AI's capabilities directly to the industrial forefront.Dedicated AI devices built with modules like the NVIDIA Jetson Nano are tailored to fit seamlessly into the larger AI ecosystem. They offer a robust platform for developing and deploying AI applications, with features such as CUDA cores for processing and pre-installed JetPack systems for ease of use. The ReComputer series, with its rich set of IOs and extensive support for AI frameworks, stands as a testament to the harmonious synergy between hardware and AI software, enabling a new wave of intelligent industrial solutions.In industrial settings, the benefits of using dedicated AI devices are manifold. They provide a localized, high-performance computing solution that can process and analyze data at the source, significantly reducing latency compared to cloud-based systems. This edge processing capability is crucial for time-sensitive applications where rapid decision-making is paramount. Moreover, these devices are designed with industrial-grade components, ensuring reliability and durability in harsh environments, essential for continuous operations in sectors such as manufacturing and logistics.The ReComputer devices, with their scalable and flexible architecture, empower industries to deploy complex AI models and machine learning frameworks directly on the factory floor, leading to improved quality control, safety, and productivity. By leveraging the computational power of AI at the edge, industries can unlock new potentials and drive innovation, ensuring they remain competitive in a rapidly evolving technological landscape.<img src="https://images.wevolver.com/eyJidWNrZXQiOiJ3ZXZvbHZlci1wcm9qZWN0LWltYWdlcyIsImtleSI6ImZyb2FsYS8xNzAwODEyODA3MDcyLWN1c3RvbWVycy1ibXctZmVhdHVyZWQtMTYwMHg4MTMuanBnIiwiZWRpdHMiOnsicmVzaXplIjp7IndpZHRoIjo5NTAsImZpdCI6ImNvdmVyIn19fQ==" style="width: 300px;" class="fr-fic fr-dib">The ReComputer devices can enable applications such as large-scale immersive learning. Image credit: BMW.&nbsp;<h3 dir="ltr">Key Features</h3>The Seeed Studio ReComputer Industrial series AI devices are engineered to deliver robust AI performance with onboard processing capabilities vital for modern industrial applications. These devices leverage the NVIDIA Ampere GPU architecture and 64-bit operating capabilities, offering AI performance from 20 TOPS to 100 TOPS depending on the module selected. This performance is critical for applications requiring real-time processing, such as in smart cities, security, and industrial automation.Connectivity features a range of interfaces, including RJ-45 GbE, RS-232/RS-422/RS-485, DI/DO, CAN, USB 3.2, and TPM2.0 (optional module). Such connectivity options ensure seamless integration into existing industrial networks and facilitate communication with various peripherals and sensors.Scalability and modularity are also key features of the ReComputer series. With different NVIDIA Jetson modules, such as the Orin NX and Xavier NX, developers can select the appropriate computing power needed for their specific applications. This modularity allows for tailored solutions that scale with the evolving demands of industrial AI tasks.Energy efficiency and power consumption are critical considerations in industrial settings. The ReComputer devices are designed with a passive heatsink for efficient cooling without a fan, conserving energy and reducing the risk of component failure due to dust or other environmental contaminants. The devices operate within a power envelope conducive to continuous operation in a wide range of industrial environments, from -20°C to 60°C.<h2 dir="ltr">Applications and Use Cases</h2>The Seeed Studio ReComputer Industrial AI Devices are designed to cater to a wide range of applications, leveraging their AI capabilities to enhance efficiency and innovation in various sectors. Here are some key use cases:<h3 dir="ltr">Video Analytics</h3>&nbsp;The ReComputer devices can perform object recognition, tracking, and classification, making them suitable for security systems, traffic monitoring, and smart city applications. These capabilities are essential for modern surveillance and monitoring systems, where real-time data processing and analysis are crucial.<h3 dir="ltr">Natural Language Processing</h3>With their advanced AI performance, these devices can handle speech recognition or language translation tasks. This application is particularly relevant in customer service automation, multilingual communication systems, and interactive voice response (IVR) systems.<h3 dir="ltr">Medical Imaging</h3>The ReComputer Industrial AI Devices can analyze medical images, assisting in diagnosis or treatment planning. This application is crucial in healthcare, where accurate and fast analysis of medical imagery can significantly impact patient care and outcomes.<h3 dir="ltr">Traffic Management</h3>In the realm of smart city applications, these devices can be used for license plate detection, car detection, and pedestrian detection. This technology is vital for enhancing traffic flow, ensuring road safety, and managing city infrastructures more effectively.<h3 dir="ltr">Industry 4.0 Applications</h3>Within industrial settings, applications like helmet detection, hard hat detection, and custom personal protective equipment (PPE) detection are possible. Additionally, visual anomaly detection using technologies like NVIDIA Deepstream IoT can be employed for quality control and monitoring.<h3 dir="ltr">Retail Sector</h3>The devices can be used for sentiment analysis and detection of retail store items. These applications can enhance customer experience, inventory management, and in-store analytics.<h3 dir="ltr">Robotics</h3>In robotics, the ReComputer series applies to tasks requiring sophisticated machine vision and AI processing, such as object detection and advanced navigation systems.<h3 dir="ltr">Industry Automation and Smart Factories</h3>The devices are ideal for industrial automation, enabling digital transformation across smart factories. They support applications like object detection, which is key in process control and quality inspection.<h2 dir="ltr">Integration with Other Systems</h2>The Seeed Studio ReComputer Industrial series is designed to be highly compatible with a range of industrial protocols and standards, making it a versatile choice for integration into various systems. The hardware features industry interfaces such as RJ-45 GbE, RS-232/RS-422/RS-485, DI/DO, CAN, USB3.2, and TPM2.0 (module optional), facilitating connectivity with a wide array of sensors, actuators, and other peripherals. This ensures that the ReComputer can serve as the intelligent core for many industrial applications, from smart cities and security to industrial automation and smart factories.The ReComputer series is also equipped with software compatibility, including pre-installed Jetpack 5.1 SDK, which encompasses CUDA-X accelerated libraries and other NVIDIA technologies. This software suite provides a comprehensive development environment and supports AI platforms that are continually updated, ensuring that the ReComputer remains at the forefront of technological advancements. Including NVIDIA DeepStream SDK, NVIDIA TAO Toolkit, and NVIDIA Riva Speech AI applications allows for AI-based multi-sensor processing, model training acceleration, and speech recognition, which are essential for modern AI tasks.Moreover, the ReComputer Industrial series' potential for cloud integration is significant, with its support for intelligent video analytics, computer vision, sound AI, fleet management, and cloud-native deployment. This makes it an ideal platform for edge computing applications where real-time data processing and analytics are critical.The Seeed Studio ReComputer Industrial series stands out for its robust compatibility with industrial protocols, ability to interface seamlessly with a range of peripherals, and capacity for integration with cloud platforms and services, making it a powerful and flexible solution for industrial AI applications.<h2 dir="ltr">Safety and Reliability</h2>Seeed Studio ReComputer Industrial series AI devices are designed with safety and reliability at the forefront, ensuring they are well-suited for the demanding conditions of industrial environments. The devices feature a fanless design with a passive heatsink, which contributes to their durability by reducing the risk of failure due to moving parts and allowing them to operate effectively in a temperature range from -20°C to 60°C. This design also mitigates the ingress of dust and other contaminants, enhancing the longevity of the devices.The ReComputer Industrial series AI Devices are built to withstand vibrations up to 3Grms from 5Hz to 500Hz, random, for 1 hour per axis. This makes them robust enough to handle the mechanical stresses in industrial settings.Certification and compliance are also key aspects of the ReComputer series. The devices comply with industry standards, including FCC, CE, RoHS, and UKCA certifications. These certifications indicate the devices' adherence to quality and safety standards, essential for industrial applications.The Seeed Studio ReComputer Industrial AI Devices provide a reliable and secure solution for industrial AI applications by integrating these built-in safety features, ensuring durability, and maintaining compliance with industry standards.<h2 dir="ltr">Development and Customization</h2>The Seeed Studio reComputer Jetson series equips developers with a powerful AI development platform featuring Ubuntu 18.04 LTS and NVIDIA JetPack 4.6 pre-installed on a 16 GB eMMC. Depending on the specific model, developers can swiftly initiate projects by connecting standard peripherals and powering the device through its Type-C or DC interface.Upon first boot, developers configure the system by setting the language, keyboard layout, network connection, and user credentials. The reComputer J1010 and J1020 models support various power modes, with the MAXN mode recommended for maximum performance.When a system reinstall is necessary, Seeed Studio's comprehensive wikis facilitate the process on A20X carrier boards and the reComputer J1010 Carrier Board. These guides offer step-by-step instructions for using NVIDIA SDK Manager or command lines, catering to diverse project needs.Developers facing limited eMMC space have several options to optimize storage: system redeployment to external devices, JetPack component removal, or capacity expansion through external storage, allowing for tailored storage solutions.Seeed Studio provides multiple channels for technical support and community discussion, ensuring developers receive the help they need for a seamless product experience.Designed with a focus on developer needs, the reComputer Jetson series delivers essential tools and resources for effective AI development, offers customization to meet unique industrial demands, and fosters a collaborative development community.<h2 dir="ltr">Conclusion</h2>The ReComputer Industrial series devices by Seeed Studio mark a significant milestone in the industrial AI landscape, offering engineers a robust platform for innovation and application development. These devices provide the necessary computational power, connectivity, and reliability to meet the demands of modern industrial environments. Engineers are encouraged to delve into the capabilities of the ReComputer series, leveraging its versatility and power to drive advancements in industrial automation and AI. With comprehensive support and resources available, the ReComputer devices stand ready to underpin the next wave of industrial AI solutions.<h3 dir="ltr">References</h3>1. <a href="https://www.forbes.com/sites/forbestechcouncil/2023/06/20/how-edge-ai-is-fueling-industry-40-outcomes/#:~:text=Manufacturers%20sit%20atop%20vast%20troves,connected%20to%20the%20data%20source." target="_blank">The role of AI in Industry 4.0, Forbes</a>4. <a href="https://www.mouser.com/datasheet/2/744/reComputer_Industrial_datasheet-3217813.pdf" target="_blank">Seeed Studio reComputer Industrial datasheet.</a>5. <a href="https://www.mouser.com/" target="_blank">Mouser Electronics</a>6. <a href="https://www.seeedstudio.com/blog/2023/04/20/innovative-community-projects-that-utilized-grove-vision-ai-module-10-inspiring-stories/" target="_blank">Seeed Studio Blog. (2023, April 20). Innovative community projects that utilized Grove Vision AI Module: 10 inspiring stories.</a>7. <a href="https://docs.edgeimpulse.com/experts/image-projects/object-detection-ubidots-seeed-grove-ai" target="_blank">Edge Impulse Documentation. (n.d.). Object detection Ubidots Seeed Grove AI</a>8. <a href="https://wiki.seeedstudio.com/reComputer_Jetson_Series_Initiation/" target="_blank">Seeed Studio Wiki. reComputer for Jetson Initiation</a>9. <a href="https://www.seeedstudio.com/blog/2023/05/10/custom-design-your-own-recomputer-powered-by-nvidia-jetson/" target="_blank">Seeed Studio Blog. Custom Design Your Own reComputer Powered by NVIDIA Jetson</a>10. <a href="https://www.seeedstudio.com/blog/2022/09/30/seeed-studio-announces-recomputer-based-on-new-nvidia-jetson-orin-nano-to-accelerate-edge-ai-from-development-to-production/" target="_blank">Seeed Studio Blog. Seeed Studio Announces reComputer Based on New NVIDIA Jetson Orin Nano to Accelerate Edge AI from Development to Production</a>

Explore Seeed Studio ReComputer: Industrial AI Devices for Enhanced Efficiency and Automation

Seeed Studio ReComputer Industrial AI Devices for Engineers

Decoding the Essentials: A Detailed Overview of Ladder Logic Symbols in Automation

Ladder Logic Symbols: A Comprehensive Guide

In those latter applications, lasers that can emit extremely short pulses—those on the order of one-trillionth of a second (one picosecond) or shorter—are especially useful. Using lasers operating on such small timescales, researchers can study physical and chemical phenomena that occur extremely quickly—for example, the making or breaking of molecular bonds in a chemical reaction or the movement of electrons within materials. These ultrashort pulses are also extensively used for imaging applications because they can have extremely large peak intensities but low average power, so they avoid heating or even burning up samples such as biological tissues.In a paper appearing in the journal Science, Caltech's <a href="https://www.eas.caltech.edu/people/amarandi" target="_blank">Alireza Marandi</a>, an assistant professor of electrical engineering and applied physics, describes a new method developed by his lab for making this kind of laser, known as a mode-locked laser, on a photonic chip. The lasers are made using nanoscale components (a nanometer is one-billionth of a meter), allowing them to be integrated into light-based circuits similar to the electricity-based integrated circuits found in modern electronics."We're not just interested in making mode-locked lasers more compact," Marandi says. "We are excited about making a well-performing mode-locked laser on a nanophotonic chip and combining it with other components. That's when we can build a complete ultrafast photonic system in an integrated circuit. This will bring the wealth of ultrafast science and technology, currently belonging to meter-scale experiments, to millimeter-scale chips."Ultrafast lasers of this sort are so important to research, that this year's Nobel Prize in Physics was awarded to a trio of scientists for the development of lasers that produce attosecond pulses (one attosecond is one-quintillionth of a second). Such lasers, however, are currently extremely expensive and bulky, says Marandi—who notes that his research is exploring methods to achieve such timescales on chips that can be orders of magnitude cheaper and smaller, with the aim of developing affordable and deployable ultrafast photonic technologies."These attosecond experiments are done almost exclusively with ultrafast mode-locked lasers," he says. "And some of them can cost as much as $10 million, with a good chunk of that cost being the mode-locked laser. We are really excited to think about how we can replicate those experiments and functionalities in nanophotonics."At the heart of the nanophotonic mode-locked laser developed by Marandi's lab is lithium niobate, a synthetic salt with unique optical and electrical properties that, in this case, allows the laser pulses to be controlled and shaped through the application of an external radio-frequency electrical signal. This approach is known as active mode-locking with intracavity phase modulation."About 50 years ago, researchers used intracavity phase modulation in tabletop experiments to make mode-locked lasers and decided that it was not a great fit compared to other techniques," says Qiushi Guo, the first author of the paper and a former postdoctoral scholar in Marandi's lab. "But we found it to be a great fit for our integrated platform.""Beyond its compact size, our laser also exhibits a range of intriguing properties. For example, we can precisely tune the repetition frequency of the output pulses in a wide range. We can leverage this to develop chip-scale stabilized frequency comb sources, which are vital for frequency metrology and precision sensing," adds Guo, who is now an assistant professor at the City University of New York Advanced Science Research Center.Marandi says he aims to continue improving this technology so it can operate at even shorter timescales and higher peak powers, with a goal of 50 femtoseconds (a femtosecond is one-quadrillionth of a second), which would be a 100-fold improvement over his current device, which generates pulses 4.8 picoseconds in length.<hr>The paper describing the research is titled "<a href="https://www.science.org/doi/10.1126/science.adj5438" target="_blank">Ultrafast mode-locked laser in nanophotonic lithium niobite</a>" and appears in the November 9 issue of Science. Co-authors are Benjamin K. Gutierrez (MS '23), graduate student in applied physics; electrical engineering graduate students Ryoto Sekine (MS '22), Robert M. Gray (MS '22), James A. Williams, Selina Zhou (BS '22), and Mingchen Liu; Luis Ledezma (PhD '23), an external affiliate in electrical engineering; Luis Costa, formerly at Caltech and now with JPL, which Caltech manages for NASA; and Arkadev Roy (MS '23, PhD '23), formerly of Caltech and now with UC Berkeley .

Lasers have become relatively commonplace in everyday life, but they have many uses outside of providing light shows at raves and scanning barcodes on groceries. Lasers are also of great importance in telecommunications and computing as well as biology, chemistry, and physics research.

Ultrafast Lasers on Ultra-Tiny Chips

The world of materials science is constantly discovering or fabricating materials with exotic properties. Among them are the multiferroics, a unique class of materials that can be both magnetized and polarized at the same time, which means that they are sensitive to both magnetic and electric fields.Having both these properties in a single material has made multiferroics very interesting for research and commercial purposes with potential applications from advanced electronics to next-generation memory storage. By understanding and harnessing the properties of multiferroics, researchers aim to develop more efficient, compact, and even energy-saving technologies.Now, an international research collaboration has uncovered some fascinating properties for the multiferroic Manganese-doped Germanium Telluride (Mn-doped GeTe); the “doped” part of the name simply means that a small amount of manganese (Mn) atoms has been introduced into the germanium telluride (GeTe) crystal structure to modify its properties. The work holds promise for the future of energy-efficient computing but also offers a deeper understanding of the collective behaviors in multiferroic materials.The project was led by Professors Hugo Dil at EPFL, Gunther Springholz at Johannes Kepler University Linz, and Jan Minár at the University of West Bohemia.Mn-doped GeTe is known for its unique ferroelectric and magnetic properties. But the new study has now found that it also possesses a magnetic order distinct from typical ferromagnets, such as iron, which align with a magnetic field. Instead, the scientists found that Mn-doped GeTe exhibits the characteristics of a ferrimagnet.What is a ferrimagnet? Unlike “normal” magnets like the ones we stick to our fridges, a ferrimagnet is more like two magnets with slightly different strengths superimposed on top of each other. The discovery that Mn-doped GeTe behaves like this means that we now have more flexibility to control the direction of magnetization – an essential feature for a number of technologies.This proved to be important, as it allowed the scientists to develop a method for enhancing the efficiency of switching magnetization direction by an astounding six orders of magnitude. Instead of doing this in the traditional way of applying a large current pulse to Mn-doped GeTe, they instead used a small, constantly fluctuating (AC) electrical current, followed by a tiny current nudge at just the right moment – a bit like pushing a swing at the right time to make it go higher with less effort. The researchers named this phenomenon “stochastic resonance”.This tiny “nudge” caused a change that spread quickly throughout Mn-doped GeTe, like a ripple in a pond. This happened because the material behaves a bit like a solid and a bit like a liquid – essentially a glass: a change in one part causes a chain reaction that changes other parts.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1699111002629-694296ca.jpg" style="width: 766px;" class="fr-fic fr-dib">The method for magnetic switching on Mn-doped GeTe. Credit: Hugo Dil (EPFL)To put in a more technical way, the magnetic switch propagated swiftly across the Mn-doped GeTe through collective excitations, which are coordinated collective movements of a large number of electron spins within the material. “This is possible because the system forms a correlated spin glass, where the local magnetic moments are in a glassy state, similar to atoms in an old-fashioned window,” says Hugo Dil. “If one spin is forced to change its orientation, this information will travel like a wave through the sample and cause the other magnetic moments to also switch.”He adds: “For technological applications, this increase in switching efficiency is of course very interesting. It can eventually lead to computers that need less than one-millionth of the currently required energy to switch a bit. However, as a physicist, what really intrigues me is the collective behavior. We are now planning space- and time-resolved experiments to follow how these excitations spread and how we can control them.”Full list of contributors<ul><li>EPFL</li><li>Paul Scherrer Institut</li><li>Johannes Kepler Universität</li><li>New Technologies-Research Center University of West Bohemia</li><li>Masaryk University</li><li>P. J. Šafárik University</li><li>Slovak Academy of Sciences</li><li>CY Cergy Paris Université</li><li>Institute of Physics ASCR, v.v.i</li><li>Charles University</li></ul>ReferencesJuraj Krempaský, Gunther Springholz, Sunil Wilfred D’Souza, Ondřej Caha, Martin Gmitra, Andreas Ney, C. A. F. Vaz, Cinthia Piamonteze, Mauro Fanciulli, Dominik Kriegner, Jonas A. Krieger, Thomas Prokscha, Zaher Salman, Jan Minár, J. Hugo Dil. Efficient magnetic switching in a correlated spin glass. Nat. Comm. 14, 6127 (2023). DOI:&nbsp;<a href="https://www.nature.com/articles/s41467-023-41718-4" target="_blank" rel="noopener">10.1038/s41467-023-41718-4</a>

Dr Cinthia Piamonteze and Dr Juraj Krempasky working on an experiment of the study at the Paul Scherrer Institut. Credit: Dominik Kriegner (FZU)

A research collaboration co-led by EPFL has uncovered a surprising magnetic property of an exotic material that might lead to computers that need less than one-millionth of the energy required to switch a single bit.

Strange magnetic material could make computing energy-efficient

They come with a pre-determined amount of RAM, plus the microprocessor, memory, input/output functions and other features. Although some SBCs do now have them, expansion slots are not usually provided for peripherals, as these additional functions are already contained on the board.You’ll find SBCs in homes, schools, computer clubs and laboratories, as they offer a cost-effective way to learn coding languages and other useful computing topics like robotics or 3D printing. They’re compact, highly reliable and efficient, so they’re also used in many industrial applications requiring a simple, low-cost yet rugged board. Some of the most important applications for SBCs include military applications, deep sea and space operations, security and surveillance, monitoring and controlling equipment and industrial automation.<h2>How Do SBCs Work?</h2>SBCs will typically consist of a motherboard, power system, microprocessor, RAM, graphics and sound chips. Built-in connectors are provided for peripherals like monitors, routers, modems, USB devices, keyboards, speakers and mice. You’ll usually store your data on a removable device such as an SD card. The SBC is connected to its applications via a backplane, which provides the physical connections for the data and power. In industry, SBCs are often used in rack systems, so you can rapidly and reliably integrate them into your application.You can choose an SBC based on its size, the type of sockets it has, minimum and maximum operating temperatures and many other factors. You can use SBCs for applications in challenging environmental conditions and stresses. You can get specialised compact designs that can power embedded applications in confined spaces. It’s a very simple infrastructure design that offers great dependability, so it’s frequently used as an embedded computer controller.Modern SBCs are getting more powerful and more accessible, so that an entire computer can now fit onto a very small PCB. With newer models, opportunities are increasing for expandability, with different expansion slots. As is typical, their price has dropped over time, while their capabilities have become greater.<h2>What Uses Do SBCs Serve?</h2>You’ll often find SBCs in business applications for security and surveillance. Although you could technically use an SBC to perform some office tasks, this isn’t normal usage, as they tend not to work with most mainstream office software suites. But they are used to great effect in industry, to run basic servers or to control programmable machinery at low cost. In industrial automation and manufacturing, they’re used as the preferred embedded controller in many major production processes.SBCs play a vital role in many control and monitoring functions, such as process control and control of robotic functions. They’re capable of carrying out most automation tasks and can be specialised for individual applications. In the&nbsp;<a href="https://www.rowse.co.uk/blog/post/benefits-of-industry-4" title="The Benefits of Industry 4.0" target="_blank">industrial IoT</a>, SBCs are also used for communication and data traffic between devices. You can acquire, process and analyse data with an SBC in order to improve production quality and efficiency. Also, since they’re self-contained, replacing them is relatively easy.A high-performance SBC can handle many industry tasks involving data analysis, complex calculations or real-time processing tasks. Some of these demand considerable computing power, including edge computing, auto optical inspection, image recognition, etc. Traditional PLCs can control equipment effectively, but often don’t have enough computing power to support data processing. SBCs can handle these tasks much more efficiently and provide quicker results. If you connect one to a PLC, you’ll add the computing power of the SBC to the standard control functions. This results in a complete automated system that’s capable of every function from control to computation, data acquisition and analysis.<h2>What SBCs Are There?</h2>You’ll find many types of SBC on the market these days, mostly of the type designed for backplane connections. They work with various operating architectures, and several manufacturers are producing them. Many different models are available in popular brands like Raspberry Pi, BeagleBone and LattePanda.<h3>RASPBERRY PI</h3>Raspberry Pi is probably the best-known SBC manufacturer, and you’ll find them everywhere from labs and academic institutions to&nbsp;<a href="https://www.rowse.co.uk/blog/post/the-rise-of-the-raspberry-pi-in-industrial-settings" title="The Rise Of The Raspberry Pi In Industrial Settings" target="_blank">industrial applications</a>. They’re cheap and easy to use, and the constant evolution of new models keeps users up to date with technology. Raspberry Pi use is so widespread that most people know how to use them, or can learn quickly if they have any computing experience.<h3>ARDUINO</h3>Arduino SBCs are very similar to Raspberry Pi in their small size, low cost and expandability. This is a totally open-source production, both software and hardware, which helps to keep costs low. These are basic machines designed for learners, novelties and simple applications, so they’re not typically used for professional tasks. If you know your stuff, they can offer a cost-effective way to run machinery, but are likely to be incompatible with more mainstream systems.<h3>INTEL</h3>Intel’s high-performance SBCs make viable substitutes for desktop PCs and can be used professionally in an office or even for gaming. They come with quad-core GPUs and they’ll run on Windows or Linux. You get immediate full connectivity and professional cases, but these aren’t really for hobbyists. The high quality of these SBCs is reflected in the higher price.<h2>Critical Applications</h2>Single board computers are often smaller, lighter, more reliable and more power-efficient than comparable desktop computers, because they have fewer components, fewer connectors and very high levels of integration. Since they’re typically used in simple applications, a comparatively low-power device provides a viable means of saving money over the more costly desktop option.SBCs often come with preset I/O interfaces for particular applications such as networking, wireless communications or audio. That makes them less adaptable, but their simple unchangeable structure also means you’ll get significantly fewer problems that might cause downtime. For this reason, SBCs are the computer of choice for running critical “always on” systems, such as medical imaging systems, traffic light controllers, missile guidance systems, anti-lock braking and automotive stability control systems, as well as many others.

A Single Board Computer (SBC) defines itself by its name: it represents a complete functioning computer, where all its functions are built into just one printed circuit board (PCB). They’re small, inexpensive to purchase and run, and offer a wide range of different capabilities.

What Are Single Board Computers?

Acoustic resonators are everywhere. In fact, there is a good chance you’re holding one in your hand right now. Most smart phones today use bulk acoustic resonators as radio frequency filters to filter out noise that could degrade a signal. These filters are also used in most Wi-Fi and GPS systems.&nbsp;Acoustic resonators are more stable than their electrical counterparts, but they can degrade over time. There is currently no easy way to actively monitor and analyze the degradation of the material quality of these widely used devices.Now, researchers at the <a href="https://seas.harvard.edu/" target="_blank">Harvard John A. Paulson School of Engineering and Applied Sciences</a> (SEAS), in collaboration with researchers at the OxideMEMS Lab at Purdue University, have developed a system that uses atomic vacancies in silicon carbide to measure the stability and quality of acoustic resonators. What’s more, these vacancies could also be used for acoustically-controlled quantum information processing, providing a new way to manipulate quantum states embedded in this commonly-used material.&nbsp;“Silicon carbide, which is the host for both the quantum reporters and the acoustic resonator probe, is a readily available commercial semiconductor that can be used at room temperature,” said <a href="https://seas.harvard.edu/person/evelyn-hu" target="_blank">Evelyn Hu</a>, the Tarr-Coyne Professor of Applied Physics and of Electrical Engineering and the Robin Li and Melissa Ma Professor of Arts and Sciences, and a &nbsp;senior author of the paper. “As an acoustic resonator probe, this technique in silicon carbide could be used in monitoring the performance of accelerometers, gyroscopes and clocks over their lifetime and, in a quantum scheme, has potential for hybrid quantum memories and quantum networking.”&nbsp;The research was published in <a href="https://www.nature.com/articles/s41928-023-01029-4" target="_blank">Nature Electronics</a>.&nbsp;<h2>A look inside acoustic resonators&nbsp;</h2>Silicon carbide is a common material for microelectromechanical systems (MEMS), which includes bulk acoustic resonators.&nbsp;“Wafer-scale manufacturable silicon carbide resonators in particular are known to have the best-in-class performance for quality factor,” said Sunil Bhave,&nbsp;professor at the Elmore Family School of Electrical and Computer Engineering at Purdue and co-author of the paper. &nbsp;“But crystal growth defects such as dislocations and grain boundaries as well as resonator manufacturing defects such as roughness, tether stress, and micro-scale craters can cause stress-concentrations regions inside the MEMS resonator.”Today, the only way to see what’s happening inside an acoustic resonator without destroying it is with super powerful and very expensive x-rays, such as the broad-spectral x-ray beam at the Argonne National Lab.&nbsp;“These types of expensive and difficult-to-access machines are not deployable for doing measurements or characterization in a foundry or somewhere where you’d actually be making or deploying these devices,” said Jonathan Dietz, graduate student at SEAS and co-first author of the paper. “Our motivation was to try to develop an approach that would allow us to monitor the acoustic energy inside of a bulk acoustic resonator so you can then take those results and feed them back into the design and fabrication process.”<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1698480897314-41928_2023_1029_Fig1_HTML.jpeg" style="width: 766px;" class="fr-fic fr-dib">A piezoelectric layer (green) sandwiched between two electrodes (yellow) atop of a silicon carbide acoustic resonator (blue). Acoustic waves generated by the electrodes and piezoelectric layer put mechanical strain on the lattice, which flip the spin of the defect (red). The spin is read out with a laser focused onto the backside of the resonator. (Credit: Hu Group/Harvard SEAS)Silicon carbide commonly hosts naturally occurring defects in which an atom is removed from the crystal lattice, creating a spatially local electronic state whose spin can interact with sound waves through material strain, such as the strain generated by an acoustic resonator.&nbsp;When acoustic waves move through the material, they put mechanical strain on the lattice, which can flip the spin of the defect. Changes in the spin state can be observed by shining a laser through the material to see how many defects are “on” or “off” after perturbing them.&nbsp;“How dim or how bright the light indicates how strong the acoustic energy is in the local environment where the defect is,” said Aaron Day, a graduate student at SEAS and co-author of the paper. &nbsp;“Because these defects are the size of single atoms, the information they give you is very local and, as a result, you can actually map out the acoustic waves inside the device in this non-destructive way.”&nbsp;That map can point to where and how the system may be degrading or not operating optimally.&nbsp;<h3>Acoustic control&nbsp;</h3>Those same defects in silicon carbide can also be qubits within a quantum system.&nbsp;Today, many quantum technologies build on the coherence of spins: how long spins will remain in a particular state. &nbsp;That coherence is often controlled with a magnetic field.&nbsp;But with their technique, Hu and her team demonstrated that they could control spin by mechanically deforming the material with acoustic waves, obtaining a quality of control similar to other approaches using alternating magnetic fields.&nbsp;“To use the natural mechanical properties of a material — its strain — expands the range of material control that we have,” said Hu. “When we deform the material, we find that we can also control the coherence of spin and we can get that information just by launching an acoustic wave through the material. It provides an important new handle on an intrinsic property of a material that we can use to control the quantum state embedded within that material.”The research was co-authored by Boyang Jiang. It was supported by the National Science Foundation under the RAISE-TAQS Award 1839164 and grant DMR-1231319.&nbsp;

Control of atomic vacancies with sound waves could improve communications and offer new control for quantum computing

Using sound to test devices, control qubits

How neuromorphic computing and BrainChip's Akida solution can benefit AI training


Navigating the Challenges in AI Training

RISC-V is inevitable - it became the mantra of RISC-V, and it's true. But before we see why that is, let’s step back and discuss what RISC-V is and why we should care.Back in the early days, computers were large and bulky, taking up entire rooms and making a lot of noise.It's incredible to think about how much technology has advanced over the years. Nowadays, we're surrounded by microprocessors that make our lives easier without us even realizing it. For example, when we use our bank cards to pay at the supermarket, we're actually using a full computer that encrypts, signs and verifies our transactions. And our phones contain dozens and dozens of processors that make all the conveniences of modern life possible.<h2>What is RISC-V?</h2>RISC-V (Reduced Instruction Set Computers) is an open-source Instruction Set Architecture (ISA). You may have heard of other well-known ISAs such as x86, ARM, Power, and MIPS.&nbsp;The RISC-V ISA is the common, standardized language between the processor (hardware) and the programs running on it (software). With RISC-V, the processor itself can only do very basic things, such as (conditionally) adding or subtracting two numbers and deciding what to do based on those arithmetic outcomes. It can also repeat those arithmetic operations several times until a certain condition is reached. In fact, there are only 37 instructions in the base. But these are the very basic ingredients needed to describe essentially any problem to the processor.&nbsp;We call those commands, instructions.&nbsp;By defining the exact meaning (semantic) and spelling (syntax) of those instructions, you can start to build a common understanding between the processor and the software running on it. You can think of the ISA as a dictionary. The ISA of RISC-V is free for anyone to access here. This is incredibly important because today’s software has become so complex that no single company can manage it alone and a common standard ensures interoperability.&nbsp;RISC-V came to the right place, by the right people, at the right time.&nbsp;David Patterson, a visionary in the field of computer architecture, spearheaded the development of Reduced Instruction Set Computers (RISC) at Berkeley. The project went through five iterations, which ultimately led to the creation of the RISC-V architecture.Initially, RISC-V mainly piqued academic curiosity. However, the availability of open-source processor implementations, coupled with rising geopolitical tensions, quickly drew industrial interest as well. From that point, the project experienced a snowball effect gaining momentum and broader adoption.<h2>Open source breeds innovation</h2>The primary advantage of RISC-V lies in its licensing flexibility. Although the specifications for other ISAs are publicly available, legal constraints prevent you from implementing and selling processor hardware unless you have obtained a license for the ISA specification. With RISC-V, this limitation is removed, offering greater freedom for innovation and commercialization.The limits that most ISAs have reduce their use by smaller innovators. For example, the widely used ISA “x86” allows only a handful of licensees. Arm, another well-known provider of processor IP, has an entirely different business model. They are selling processor IPs that can be licensed and integrated into your product. Only the biggest players in the market are able to own an architectural license that allows them to implement their own processors.&nbsp;Most notable of these is Apple, which switched its entire product line from Intel x86 processors to its own Arm implementation. Their investment in this endeavor proved to be a wise decision both technically and economically, as it enabled them to distinguish their product line even further and surpass their rivals by a significant margin.In the past, designing your own processor and utilizing the expansive software ecosystem was thought to be reserved solely for large corporations. However, that's not entirely true. Alternatives like OpenRISC existed, but they weren't widely adopted because their software ecosystems were underdeveloped, and they were often considered hobbyist projects.Another notable benefit of RISC-V is its built-in modularity, which contrasts with ARM's more rigid structure. The RISC-V architecture is designed from the ground up to be modular, allowing for easier customization and innovation. While ARM tightly controls and protects its ISA specification, making it difficult to introduce your own specialized features, RISC-V offers the flexibility to add custom extensions and your “secret sauce.”This is particularly advantageous if you aim to target specific market segments and differentiate your product from competitors. Remember those 37 instructions I’ve mentioned? Turns out this is only the very base instruction set. For most applications, you would likely want to include some more specialized instructions such as floating-point support, atomic memory operations, or maybe even scalable vector extensions. You don’t need to, but you can. In the same way, your company could add your own instructions in case those are beneficial for the application in mind. This means you can still leverage the entire software ecosystem that has been and is being built and only add value on top instead of reinventing the wheel. So let’s have a look at some of these advantages.&nbsp;<h2>How can RISC-V help start-ups design their own chips?</h2>For me, the two most important aspects for a hardware-producing start-up are time-to-market and freedom to innovate.&nbsp;Most silicon/hardware start-ups innovate and try to provide value-add in their niche. The processor is a necessary, boring commodity and usually not the main star of the product. This is good and the way it should be. Otherwise, we would all be selling the same processors.&nbsp;RISC-V is an absolute game changer for that.&nbsp;Do you want to quickly try out an integration idea? Prototype something on an FPGA? Before RISC-V this either meant you needed to use some custom FPGA vendor’s ISA (which you couldn’t send to manufacture for your final product) or start negotiations with the processor vendor of your choice. Negotiations are hard for a small silicon start-up. With RISC-V you can go and grab some core from the internet, do your trials, and start building your software stack. Once happy with the prototype and you want to move forward, there are a plethora of RISC-V IP vendors you can choose from. Choice and market competition are a good thing.Expanding on the points raised earlier, the freedom to innovate with RISC-V comes without the risk of architectural lock-in. This has two advantages, first, you are free to switch IP vendors. Your software stack will still work the same because you adhered to an open standard. Second, you can provide your own extensions on top of the main RISC-V standards, and the company can provide very concentrated value-add without the need to maintain or develop commodity.&nbsp;<h3>Let’s have a look at what we’ve done at Axelera.&nbsp;</h3>At the heart of our Metis AI Processing Unit, there are four AI Cores. Each AI core provides a 512x512 matrix-vector multiplication (MVM) in-memory compute array and a vector datapath that operates on streams of data.&nbsp;To provide a generic solution that can keep up with the fast evolution in the field of neural networks we kept the datapath control as flexible as possible. We aim to push as much as possible into low-level driver software where we can innovate, correct, and adapt throughout the product’s life cycle. Therefore, each AI Core has a dedicated RISC-V, application-class core, that has full control over the datapath unit. A 5th system controller provides SoC management and workload dispatch to the individual AI cores. We’ve confidently chosen the RISC-V ISA because we could start virtual prototyping and building our software stack on open and free software far before we had done any IP negotiations. Furthermore, we knew that the IP landscape would provide us with sufficient competitive choices for our needs.&nbsp;In roughly a year, we went from concept/FPGA prototype to tape-out, and in just a couple of days after having received the chips back, we could run the first end-to-end neural networks on our lab benches. Taking the limited resources and this aggressive time schedule into account it was clear that we’ll needed to buy proven IP without long negotiations to kick-start our developments.&nbsp;Choosing to proceed with RISC-V ensured that we could continue to innovate in future generations and further distinguish our offerings.So, is it all love, peace, and harmony? Not exactly. Like anything that's still a work in progress. RISC-V has its own set of hiccups and curveballs.An ISA is mainly about instructions, but a processor is part of a bigger system that includes memory, peripherals, and other processors. This broader setup, often called a platform, isn't covered in the main ISA spec. The RISC-V Foundation aims to standardize these aspects too.However, details matter. Things like atomic handling in multi-hart systems may only be standardized for specific platforms. Also, some IP providers might have legacy elements because they developed solutions before RISC-V standards were set. Standards are scattered, so you have to keep an eye out. And given the rapid growth, it's tough to stay updated on all the ongoing activities.<h2>Looking ahead</h2>Remember those room-filling mainframes and how technology has continuously shrunk down to fit our pockets and wrists? Just like those evolutionary leaps, RISC-V aims to be a cornerstone in the next phase of computational progress.In an ideal future, RISC-V will be as "boring" as any well-established technology—boring in the sense that it's reliable, stable, and so embedded in our daily lives that we take it for granted. The long-term vision is for software to run seamlessly on RISC-V-based processors without users having to worry about compatibility issues.In the short term, however, there's still much to be done. While RISC-V is showing promise in deeply embedded systems like smartcards and IoT devices, achieving the same level of support and efficiency as ARM and x86 in more user-centric applications—like your phone or computer—is an ongoing project. Industry trends, as evidenced by companies like SiFive and Ventana and initiatives like the RISE project, suggest a competitive and collaborative future for RISC-V. From a research standpoint, its open nature makes it a fertile ground for innovation across the entire hardware-to-software stack.There's a palpable sense of momentum behind RISC-V, with heavyweight stakeholders and top-tier companies pooling their resources and expertise. Far from slowing down, the ecosystem around RISC-V is only gaining speed, setting it on a course to become as ubiquitous and "boring" as the microprocessors that have come to define modern life.&nbsp;So, looking at where things are going, ignoring RISC-V would be like missing the boat when we switched from those giant mainframes to desktop PCs. It's a game-changer, and you really don't want to be left behind.

RISC-V Evolution: From Early Computing to Tomorrow's Tech Promise 

Harnessing the RISC-V Wave: The Future is Now

This article originally appeared in <a href="https://engineering.princeton.edu/news/2023/10/11/illuminating-errors-creates-new-paradigm-quantum-computing" target="_blank" rel="noopener noreferrer">Princeton University</a> and has been republished with permission.<hr>Researchers have developed a method that can reveal the location of errors in quantum computers, making them up to ten times easier to correct. This will significantly accelerate progress towards large-scale quantum computers capable of tackling the world’s most challenging computational problems, the researchers said.Led by Princeton University’s <a href="https://ece.princeton.edu/people/jeff-thompson" target="_blank">Jeff Thompson</a>, the team demonstrated a way to identify when errors occur in quantum computers more easily than ever before. This is a new direction for research into quantum computing hardware, which more often seeks to simply lower the probability of an error occurring in the first place.A <a href="https://www.nature.com/articles/s41586-023-06438-1" target="_blank">paper</a> detailing the new approach was published in Nature on Oct. 11. Thompson’s collaborators include Shruti Puri at Yale University and Guido Pupillo at the University of Strasbourg.Physicists have been inventing new qubits – the core component of quantum computers – for nearly three decades, and steadily improving those qubits to be less fragile and less prone to error. But some errors are inevitable no matter how good qubits get. The central obstacle to the future development of quantum computers is being able to correct for these errors. However, to correct an error, you first have to figure out if an error occurred, and where it is in the data. And typically, the process of checking for errors introduces more errors, which have to be found again, and so on.Quantum computers’ ability to manage those inevitable errors has remained more or less stagnant over that long period, according to Thompson, associate professor of electrical and computer engineering. He realized there was an opportunity in biasing certain kinds of errors.“Not all errors are created equal,” he said.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1697112488641-Thompson-erasure-for-web-02.jpg" style="width: 766px;" class="fr-fic fr-dib">At the heart of this quantum computer, lasers control the states of individual atoms, manipulating information based on the extremely subtle physics of those atoms’ energy levels. Photo by Frank WojciechowskiThompson’s lab works on a type of quantum computer based on neutral atoms. Inside the ultra-high vacuum chamber that defines the computer, qubits are stored in the spin of individual ytterbium atoms held in place by focused laser beams called optical tweezers. In this work, a team led by graduate student Shuo Ma used an array of 10 qubits to characterize the probability of errors occurring while first manipulating each qubit in isolation, then manipulating pairs of qubits together.They found error rates near the state of the art for a system of this kind: 0.1 percent per operation for single qubits and 2 percent per operation for pairs of qubits.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1697112509360-Qbit-erasure-final-Illustration-2048x1152.jpg" style="width: 766px;" class="fr-fic fr-dib">A team led by Jeff Thompson, associate professor of electrical and computer engineering, pioneered an approach to more efficient error correction in quantum computers. (Illustration by Gabriele Meilikhov/Muza Productions)However, the main result of the study is not only the low error rates, but also a different way to characterize them without destroying the qubits. By using a different set of energy levels within the atom to store the qubit, compared to previous work, the researchers were able to monitor the qubits during the computation to detect the occurrence of errors in real time. This measurement causes the qubits with errors to emit a flash of light, while the qubits without errors remain dark and are unaffected.This process converts the errors into a type of error known as an erasure error. Erasure errors have been studied in the context of qubits made from photons, and have long been known to be simpler to correct than errors in unknown locations, Thompson said. However, this work is the first time the erasure-error model has been applied to matter-based qubits. It follows a <a href="https://www.nature.com/articles/s41467-022-32094-6" target="_blank">theoretical proposal</a> last year from Thompson, Puri and Shimon Kolkowitz of the University of California-Berkeley.In the demonstration, approximately 56 percent of one-qubit errors and 33 percent of two-qubit errors were detectable before the end of the experiment. Crucially, the act of checking for errors doesn’t cause significantly more errors: The researchers showed that checking increased the rate of errors by less than 0.001 percent. According to Thompson, the fraction of errors detected can be improved with additional engineering.The researchers believe that, with the new approach, close to 98 percent of all errors should be detectable with optimized protocols. This could reduce the computational costs of implementing error correction by an order of magnitude or more.Other groups have already started to adapt this new error detection architecture. Researchers at Amazon Web Services and a separate group at Yale have independently shown how this new paradigm can also improve systems using superconducting qubits.“We need advances in many different areas to enable useful, large-scale quantum computing. One of the challenges of systems engineering is that these advances that you come up with don’t always add up constructively. They can pull you in different directions,” Thompson said. “What’s nice about erasure conversion is that it can be used in many different qubits and computer architectures, so it can be deployed flexibly in combination with other developments.”Additional authors on the paper “High-fidelity gates with mid-circuit erasure conversion in a metastable neutral atom qubit” include Shuo Ma, Genyue Liu, Pai Peng, Bichen Zhang, and Alex P. Burgers, at Princeton; Sven Jandura at Strasbourg; and Jahan Claes at Yale. This work was supported in part by the Army Research Office, the Office of Naval Research, DARPA, the National Science Foundation and the Sloan Foundation.

Researchers led by Jeff Thompson at Princeton University have developed a technique to make it 10 times easier to correct errors in a quantum computer. Photos by Frank Wojciechowski

Researchers have developed a method that can reveal the location of errors in quantum computers, making them up to ten times easier to correct. This will significantly accelerate progress towards large-scale quantum computers capable of tackling the world’s most challenging computational problems, the researchers said.

Illuminating errors creates a new paradigm for quantum computing

While Raspberry Pi 3 is an industry leader in Single Board Computers, unless you're one of the Raspberry Pi foundations’ existing industrial customers, it can be a significant challenge to acquire a 3B+ at the moment. Fortunately, global tech company <a href="https://www.okdo.com/?utm_source=wevolver&utm_campaign=rock&utm_medium=affiliates&utm_content=rock3_vs_rpi3" target="_blank" rel="noopener noreferrer">OKdo</a> offers two viable alternatives to the Raspberry Pi 3B+ in the form of the ROCK 3 Model A and ROCK 3 Model C boards, both of which adhere to the 3B+ specifications and offer enhanced capabilities. In this article, we will conduct a comparative analysis of these three models.<h2>Key specifications</h2>Here we examine the technical specifications of each board in this comprehensive side-by-side comparison.<div dir="ltr" align="left"><table><tbody><tr><td>&nbsp;</td><td style="width: 23.1071%;">ROCK 3 Model A</td><td style="width: 28.9817%;">ROCK 3 Model C</td><td style="width: 29.8956%;">Raspberry Pi 3 Model B+</td></tr><tr><td>Availability</td><td style="width: 23.1071%;">Available</td><td style="width: 28.9817%;">Available</td><td style="width: 29.8956%;">Sold Out</td></tr><tr><td>SoC</td><td style="width: 23.1071%;">Rockchip RK3568</td><td style="width: 28.9817%;">Rockchip RK3566</td><td style="width: 29.8956%;">Broadcom BCM2837</td></tr><tr><td>Processor</td><td style="width: 23.1071%;">Quad‑core Arm® Cortex®‑A55 (ARMv8) 64‑bit @ 2.0GHz</td><td style="width: 28.9817%;">Quad‑core Arm® Cortex®‑A55 (ARMv8) 64‑bit @ 1.6GHz</td><td style="width: 29.8956%;">Quad‑core Arm®, Cortex®-A53 (ARMv8) 64-bit @ 1.4GHz</td></tr><tr><td>Clock</td><td style="width: 23.1071%;">2.0GHz</td><td style="width: 28.9817%;">1.6GHz</td><td style="width: 29.8956%;">1.4GHz</td></tr><tr><td>GPU</td><td style="width: 23.1071%;">Arm Mali™ G52-2EE</td><td style="width: 28.9817%;">Arm Mali™ G52-2EE</td><td style="width: 29.8956%;">Broadcom Videocore-IV</td></tr><tr><td>Memory</td><td style="width: 23.1071%;">2GB LRDDR4 SRAM</td><td style="width: 28.9817%;">1GB / 2GB LRDDR4 SRAM</td><td style="width: 29.8956%;">1GB LPDDR2 SDRAM</td></tr><tr><td>Power</td><td style="width: 23.1071%;">USB Type-C™ PD / Qualcomm® Quick Charge™ / GPIO</td><td style="width: 28.9817%;">USB Type-C™ / GPIO</td><td style="width: 29.8956%;">Micro-USB 5V/2.5A</td></tr><tr><td>HDMI</td><td style="width: 23.1071%;">1x HDMI 4K@60fps</td><td style="width: 28.9817%;">1x HDMI 1080p@60fps</td><td style="width: 29.8956%;">1x HDMI 1080p@30fps</td></tr><tr><td>USB</td><td style="width: 23.1071%;">2x USB2 HOST / 1x USB3 HOST / 1 x USB 3.0 OTG-HOST</td><td style="width: 28.9817%;">2x USB2 HOST / 1x USB3 HOST / 1x USB2 OTG-HOST</td><td style="width: 29.8956%;">4 USB 2.0 PORTS</td></tr><tr><td>Audio Ports</td><td style="width: 23.1071%;">3.5mm jack – stereo output/mic input</td><td style="width: 28.9817%;">3.5mm jack – stereo output/mic input</td><td style="width: 29.8956%;">3.5mm jack – stereo output</td></tr><tr><td>Connectivity</td><td style="width: 23.1071%;">M.2 E Key / Gigabit Ethernet</td><td style="width: 28.9817%;">2.4GHz and 5GHz IEEE 802.11b/g/n/ac Wireless LAN / Bluetooth 5.0 / Gigabit Ethernet</td><td style="width: 29.8956%;">2.4GHz and 5GHz IEEE 802.11.b/g/n/ac Wireless LAN / Bluetooth 4.2 / Ethernet (300bps)</td></tr><tr><td>MIPI</td><td style="width: 23.1071%;">1x 2 lane DSI / 1x 2 lane CSI</td><td style="width: 28.9817%;">1x 2 lane DSI / 1x 2 lane CSI</td><td style="width: 29.8956%;">MIPI DSI / MIPI CSI</td></tr><tr><td>Storage</td><td style="width: 23.1071%;">Micro SD / eMMC / M.2 M Key NVME / SATA</td><td style="width: 28.9817%;">Micro SD / eMMC / M.2 M Key NVME / SATA</td><td style="width: 29.8956%;">Micro SD</td></tr><tr><td>UARTS</td><td style="width: 23.1071%;">5x UART</td><td style="width: 28.9817%;">5x UART</td><td style="width: 29.8956%;">2x UART</td></tr><tr><td>RTC</td><td style="width: 23.1071%;">Yes</td><td style="width: 28.9817%;">No</td><td style="width: 29.8956%;">No</td></tr><tr><td>Power On/Of</td><td style="width: 23.1071%;">Yes</td><td style="width: 28.9817%;">Yes</td><td style="width: 29.8956%;">No</td></tr><tr><td>Software</td><td style="width: 23.1071%;">Debian / Ubuntu / Android 11/ Armbian / etc</td><td style="width: 28.9817%;">Debian / Ubuntu / Android 11/ Armbian / etc</td><td style="width: 29.8956%;">Raspberry Pi OS / Ubuntu / LibreELEC / etc</td></tr><tr><td>Operation Temperature</td><td style="width: 23.1071%;">0~50 ℃</td><td style="width: 28.9817%;">0~50 ℃</td><td style="width: 29.8956%;">0~50 ℃</td></tr></tbody></table></div>Download the full comparison table <a href="https://www.okdo.com/wp-content/uploads/2023/09/OKdo_SPS_LogoStripROCK3vsPi3.pdf" target="_blank" rel="noopener noreferrer" data-lf-fd-inspected-kn9eq4roawkarlvp="true">here</a>.<h2 dir="ltr"><iframe width="767" height="360" src="https://www.youtube.com/embed/tkIyzBvriS0?enablejsapi=1&amp;origin=https://www.wevolver.com" title="ROCK 3 vs. Raspberry Pi 3: A Comparative Engineer’s Review | OKdo" frameborder="0" allowfullscreen="" class="fr-draggable" data-lf-form-tracking-inspected-kn9eq4roawkarlvp="true" data-lf-yt-playback-inspected-kn9eq4roawkarlvp="true" data-lf-vimeo-playback-inspected-kn9eq4roawkarlvp="true" data-gtm-yt-inspected-10="true" id="506105029">&nbsp;&nbsp;&nbsp;</iframe></h2><h2 dir="ltr">ROCK 3 Model C</h2>The&nbsp;<a href="https://www.okdo.com/p/okdo-rock-3-model-c-2gb-single-board-computer-rockchip-rk3566-arm-cortex-a55/?utm_source=wevolver&utm_campaign=rock&utm_medium=affiliates&utm_content=rock3_vs_rpi3" target="_blank" rel="noopener noreferrer">ROCK 3 Model C</a> shares many similarities with the Raspberry Pi 3 Model B+. However, it distinguishes itself with several notable additional features that outperform the RPI 3B+. In this analysis, we primarily focus on these differentiating aspects. Both boards feature a 64-bit, quad-core Arm System-on-Chip (SoC). While the Raspberry Pi 3 Model B+ incorporates the Cortex A-53 mid-range processor, the ROCK 3C has the low-power Cortex A55 cores.&nbsp; When subjecting both boards to the same benchmark test, such as Sysbench to fully utilize all four cores for computation, the results are nearly identical. However, the ROCK 3C consumes nearly half the power and maintains lower operating temperatures, offering significant advantages for eco-conscious applications.<img alt="ROCK 3C sbc" src="https://images.wevolver.com/eyJidWNrZXQiOiJ3ZXZvbHZlci1wcm9qZWN0LWltYWdlcyIsImtleSI6ImZyb2FsYS8xNjk2MjQzMDM4MTMxLTE2OTYyNDMwMzgxMzEucG5nIiwiZWRpdHMiOnsicmVzaXplIjp7IndpZHRoIjo5NTAsImZpdCI6ImNvdmVyIn19fQ==" width="499" height="499" class="fr-fic fr-dii">Thanks to its lower-power Cortext A55 cores, the Rock 3C uses nearly half the power compared to the Raspberry Pi 3B+. &nbsp;Image credit: Okdo<h3 dir="ltr">ROCK 3C – high-quality graphics processing</h3>For those seeking superior graphics processing capabilities, the ROCK 3C provides HDMI 1080p@60fps, in contrast to the 1080p@30fps of the RPI 3B+. This difference in video output quality is attributed to the ROCK 3C's Mali GPU, which offers broader support for graphics rendering APIs. Furthermore, the 2GB RAM configuration of the ROCK 3C enhances performance, surpassing the 1MB RAM limitation of the RPI 3B+.Both boards have the same Wireless LAN capabilities. The difference comes when using Bluetooth. The ROCK 3C complies with Bluetooth v5.0, which can give a significantly higher range, broadcast capability and throughput than the v4.2 supported by the RPI 3B+.<h3 dir="ltr">More connectivity and storage options&nbsp;</h3>In terms of wired connectivity, the ROCK 3C has true Gigabit Ethernet, while the Raspberry Pi 3 Model B+ is constrained to 300bps due to bandwidth sharing with the USB bus. Both boards support Power over Ethernet (PoE).A significant advancement with the ROCK 3C is its mitigation of the Raspberry Pi 3 Model B+'s reliance on SD card media for storing the operating system and data. The 3C offers the option to utilize dedicated eMMC memory, delivering slightly improved performance and significantly enhanced reliability and security. The eMMC connector is located on the underside of the board, and modules are available in capacities ranging from 16GB to 128GB.While the Raspberry Pi 3 Model B+ accommodates USB to SSD as a storage solution, it is constrained by its 4 x USB 2.0 ports, which impact throughput and share bandwidth with Ethernet. In contrast, the ROCK 3C offers 1 x USB 3.0 and 3 x USB 2.0 ports, with one configured as OTG.&nbsp; <img src="https://images.wevolver.com/eyJidWNrZXQiOiJ3ZXZvbHZlci1wcm9qZWN0LWltYWdlcyIsImtleSI6ImZyb2FsYS8xNjk2MjQzMDM2ODcyLTE2OTYyNDMwMzY4NzIucG5nIiwiZWRpdHMiOnsicmVzaXplIjp7IndpZHRoIjo5NTAsImZpdCI6ImNvdmVyIn19fQ==" width="602" height="601" class="fr-fic fr-dii" style="width: 758px;">With 1 x USB 3.0 and 3 x USB 2.0 ports, the ROCK 3C offers versatile connectivity options. Image credit: OKdo.For those seeking high-performance, large capacity data storage, the ROCK 3C has an M.2 M‑Key 2230 connector with single lane, PCIe 2.1 and SATA 3.0 combined interfaces. The M.2 slot can be configured either to support NVMe or SATA SSD devices, though an additional adapter board is required for SATA support.Additional features include a 2-pin fan header equipped with PWM (Pulse Width Modulation) capability, as well as an on/off push-button for safe system power management. The device also includes a microphone input integrated into the audio jack.Physically, the ROCK 3C shares the same dimensions and mounting hole geometry as the RPI 3B+, with the exception of the USB-C power connector, so it fits existing RPI 3B+ cases by widening the power connector aperture.<h2 dir="ltr">ROCK 3 Model A</h2>Now we will examine the <a href="https://www.okdo.com/p/okdo-rock-3-model-a-2gb-single-board-computer-rockchip-rk3568-arm-cortex-a55/?utm_source=wevolver&utm_campaign=rock&utm_medium=affiliates&utm_content=rock3_vs_rpi3" target="_blank" rel="noopener noreferrer">ROCK 3 Model A</a>. While it comes at a slightly higher price point compared to the ROCK 3 Model C and Raspberry Pi 3 Model B+, it offers increased processing power.The board is based on the RK3568, a 64-bit quad-core Arm System-on-Chip SoC, with the same Cortex A55 low-power cores and Mali G52 as the ROCK 3C but with the addition of a 0.8TOPS (Tera Operations Per Second) NPU (Neural Processing Unit) tailored for AI workloads. In Sysbench-based compute load tests, marginal disparities were observed among all three boards. However, in the realm of graphics processing, the ROCK 3A surges ahead, delivering noticeably superior image rendering quality and speed via its HDMI 4K video output.<img alt="ROCK 3A sbc" src="https://images.wevolver.com/eyJidWNrZXQiOiJ3ZXZvbHZlci1wcm9qZWN0LWltYWdlcyIsImtleSI6ImZyb2FsYS8xNjk2MjQzMDM5NTAxLTE2OTYyNDMwMzk1MDEucG5nIiwiZWRpdHMiOnsicmVzaXplIjp7IndpZHRoIjo5NTAsImZpdCI6ImNvdmVyIn19fQ==" width="499" height="499" class="fr-fic fr-dii" style="width: 758px;">Despite its slightly higher price point, The ROCK 3A offers enhanced processing capabilities. Image credit: OKdo.<h3 dir="ltr">Wireless Capability Sets ROCK 3A Apart</h3>Among the primary distinctions between the ROCK 3A and the ROCK 3C lies in their wireless capabilities. The ROCK 3A features an M.2 E-key slot on the top side of the board, designed for a Wireless LAN module. This complements the M.2 M-key slot on the bottom side, intended for NVMe or SATA SSD drives, and an eMMC connector dedicated to OS and data storage.In addition to fixed voltage power supplies, the ROCK 3A can make use of USB Type‑C™ PD and Qualcomm® Quick Charge™ adapters, making it more flexible and efficient for powering ancillaries. The front edge of the board hosts Gigabit Ethernet, with two USB 3.0 connectors available. The top one is capable of switching between Host and OTG modes through a small slide switch located between the USB blocks on the underside of the board. Additionally, there are two USB 2.0 connectors. <img src="https://images.wevolver.com/eyJidWNrZXQiOiJ3ZXZvbHZlci1wcm9qZWN0LWltYWdlcyIsImtleSI6ImZyb2FsYS8xNjk2MjQzMDM3MzA3LTE2OTYyNDMwMzczMDcucG5nIiwiZWRpdHMiOnsicmVzaXplIjp7IndpZHRoIjo5NTAsImZpdCI6ImNvdmVyIn19fQ==" width="602" height="601" class="fr-fic fr-dii" style="width: 758px;">The ROCK 3A features an external battery connector on the underside of the board for the RTC and streamlining designs. Image credit: OKdo.&nbsp;The board offers the convenience of an on/off switch, along with the inclusion of a 3-pin fan connector, enabling precise control and feedback for fan operation. An additional noteworthy feature is the presence of an external battery connector situated on the underside of the board, specifically intended for the RTC (Real-Time Clock). This reduces the Bill of Materials (BOM) and streamlines designs.&nbsp;The ROCK 3A has the same size, form factor and mounting holes as the Raspberry Pi 3 Model B+, the difference being the USB-C power connector, so RPI 3B+ cases may need to be modified accordingly.Presently, the ROCK 3A is readily available in a 2GB RAM configuration.&nbsp;<h2 dir="ltr">Operating Systems</h2>All of these boards provide a range of operating system options, including various versions of the Linux kernel. The Raspberry Pi 3 Model B+ offers its 32-bit Debian Bullseye version, along with a 64-bit version.The ROCK 3A and ROCK 3C come with officially supported 64-bit Debian Bullseye, Ubuntu Server 22.04 (Jammy), and Android 11 images. Currently, both boards are running on kernel version 5.10, and users have the flexibility to deploy either stable (version 6.1) or the latest (version 6.4) mainline kernels on the ROCK 3A, depending on their specific needs.Additionally, users can explore a wide range of third-party OSs to suit their requirements across all three boards.<img src="https://images.wevolver.com/eyJidWNrZXQiOiJ3ZXZvbHZlci1wcm9qZWN0LWltYWdlcyIsImtleSI6ImZyb2FsYS8xNjk2MjQzMDM4NjEyLTE2OTYyNDMwMzg2MTIucG5nIiwiZWRpdHMiOnsicmVzaXplIjp7IndpZHRoIjo5NTAsImZpdCI6ImNvdmVyIn19fQ==" width="602" class="fr-fic fr-dii" style="width: 748px;">Side-by-side: The Raspberry Pi 3B+, the ROCK Model A, and the ROCK Model C all share the same dimensions, but offer different capabilities. Image Credit: OKdo.&nbsp;<h2 dir="ltr">Summary</h2>The Raspberry Pi 3B+ has been a popular choice for a powerful and cost-effective Single Board Computer (SBC) in embedded applications. However, newer options like the ROCK 3 Model A and ROCK 3 Model C boards which are available at a similar price point, offer comparable or superior performance and additional features, making them compelling alternatives to the 3B+.&nbsp;This article first appeared on <a target="_blank" rel="noopener noreferrer" href="https://www.okdo.com/blog/rock-3-vs-raspberry-pi-3-an-engineers-review/">OKdo.com</a> written by&nbsp;Peter Milne, engineer and Linux advocate.&nbsp;

A comparative analysis of the Raspberry Pi 3B+, the ROCK 3 Model A, and the  ROCK 3 Model C. 

ROCK 3 vs. Raspberry Pi 3: A Comparative Analysis

computing

Profiles