It feels as though our world is spinning faster and faster – and the innovation cycles of the electronics industry are getting shorter and shorter. In the more than 50 years since its foundation, Rutronik has consistently kept pace with industry developments, thereby providing its customers with access to the latest innovations. In the current landscape of artificial intelligence (AI), hardware with AI accelerators, such as neural processing units, can now be found across all product areas and performance classes. These accelerators function as coprocessors, ensuring applications, from field use to data centers, are fit to meet future requirements. Further, there are various legal aspects that companies need to be aware of.
The end of Windows 10 and the consequences
Many companies still have notebooks running the Windows 10 operating system. However, Microsoft has already announced the official end of support on October 14, 2025. And it is not just technical support that is on its way out. The lack of security updates in particular is forcing companies in environments with increased criticality to make a transition due to the Network and Information Security Directive (NIS2), which is set to come into effect a year earlier (October 2024). That said, all other businesses, as well as private users, should also be aware of future security risks, especially as potential threats are bound to increase.
Users who simply want to update their existing computer environment to Windows 11 might fail due to a lack of compatibility or increased system requirements. Owners of computers with Intel Core processors up to and including the seventh generation would, at best, need to switch to Linux or operate offline to avoid a disaster. Most companies and users will likely see the end of support as an opportunity to upgrade their hardware when transitioning to Windows 11. Additionally, the peripherals in the latest devices have also advanced. Users also benefit from newer Bluetooth and Wi-Fi specifications, as well as enhanced cameras, displays, and battery life. When purchasing new computers as part of the transition to Windows 11, it is advisable to choose devices with an integrated NPU. This not only relieves the CPU and GPU but also provides enormous speed and energy management benefits for the ever-increasing use of AI processes.
Expert advice and individual configuration options
With the exone go Business 1490 X14 and exone go Business 1690 X14 notebooks, Rutronik provides its customers with two models featuring a 14" and 16" display, both powered by the new AI-capable Intel Core Ultra 7-155U processor. In a durable and lightweight housing (overall weight approx. 1 kg), security functions ensure data protection. Advanced connectivity options ensure that a connection can be established with the computer anytime, anywhere. The supplier stipulates a battery life of 16 hours, thus guaranteeing that even after years of use, there should still be enough power for a whole working day (Fig. 1).
The experts from the relatively new Rutronik IT Electronics division offer advice on configuration options, expansions, peripherals, and AI accelerator cards for existing hardware, for example. The team is also happy to assist with the individual configuration of a computer. As an official Intel franchise distributor, Rutronik provides the new Core Ultra processors together with compatible motherboards from Kontron, Asus, and Advantech. Along with a portfolio of main memories, ROMs (also in conjunction with Windows licenses), power packs, housings, and peripherals, it is possible to assemble customized AI computers.
QR code to access Rutronik IT Electronics
Scan the QR code for more information and to contact Rutronik IT Electronics: https://www.rutronik.com/rutronik-it-electronics#picture-65931
A look at the computer market
Looking beyond the present and into the future: When the COVID-19 pandemic struck and many people started working from home, all things computer-related experienced a boom in popularity. Laptops and peripherals were acquired for the now mobile workforce, along with servers that facilitated networked, decentralized working. The supply chain was put to a hard test. Many semiconductor components were allocated to the computer industry, which obviously had the added advantage of placing large orders. For every car produced, there were countless new laptops manufactured. Some sectors, like the aforesaid automotive industry, were left trailing when it came to chip supplies. They sought to acquire the coveted resources on the market by means of increased forecasts and placing double orders, thus aiming to broadly maintain the production levels of their end products. When the market stabilized in 2022/2023, storage facilities were suddenly overwhelmed with backlogged orders. Even in 2024, stock coverage of up to two years can still be observed, which is arguably a historical record.
Due to the full storage facilities of electronics processing companies and a weakening economy, new orders are scarce, as delivery times and semiconductor availability are no longer critical concerns. Caution is advised: Semiconductor suppliers are being forced to scale back their production capacities and to adjust their supply chains accordingly.
This yo-yoing of store inventory and production volumes has been known colloquially for decades as the hog cycle. Despite the dramatic fluctuations in recent years, the current situation is quite different from the familiar patterns of the past. The uncertain economic climate, marked by interest rates higher than those of the past decade, is causing companies to be more cautious about investing, with hopes that interest rates will soon start to fall. The end of support for Windows 10 and the competitive pressure to use AI are set to give the computer market a new boost in 2025. By then, the laptops from the early days of the pandemic will have been fully depreciated and can be replaced. This anticipated boom, coupled with a decrease in the production capacity for small electronic components, signals the onset of the next phase of allocation. The entire electronics industry has been emphasizing for several weeks now that long-term demand plans from customers are essential to secure supply for 2025 and beyond. All IT buyers are well advised not to wait until Microsoft’s support ends. A bit more foresight and a time buffer can help avoid a depleted market and provide better purchasing conditions than waiting around until the shelves of distributors and retailers are empty again.
Powerful AI applications and developer tools
AI is, however, not just a topic for those of us sitting in front of a PC. Even in professional applications, such as those observed in industry, medicine, robotics, and automation, sensor patterns, camera images, and audio sources are frequently analyzed using large language models (LLMs) or large action models (LAMs), with conclusions drawn according to the laws of probability.
Let us start with sensors and actuators – in other words, very close to the workpiece. Sensor data are analyzed and evaluated for patterns before proceeding with further processing. Edge Impulse, for example, provides machine learning software for microcontrollers from Nordic Semiconductor and Infineon. Some smart rings that monitor a wearer’s health parameters and transmit the data to a smartphone utilize low-power hardware from Nordic Semiconductor, along with respective ML/AI software from Edge Impulse. The fact that a ring equipped with sensors, Bluetooth, and AI calculations runs on a compact battery highlights the sgnificant advancements in energy efficiency achieved in recent years.
The new nRF54L and nRF54H families, which represent the latest and enhanced generations of multicore microcontrollers in a pin-sized chip, are set to be released at the end of 2024 and the beginning of 2025 respectively. With the PSoC 63 Bluetooth microcontroller from Infineon, customers can choose between Edge Impulse and the Imagemob software, which is now integrated into Infineon’s offerings. Both software solutions focus on machine learning tasks in the controller. Imagemob is already available in the Infineon SDK Modbus Toolbox. Installation instructions for use of the third-party supplier software Edge Impulse can also be found there.
Meeting the challenges of edge computing
If you require more processing power beyond evaluating movements, assigning sound patterns to events, consolidating environmental parameters, and making forecasts, Rutronik provides complete computers, semiconductor chips, and AI-capable embedded boards. This class is typically used for edge computing applications – in other words, a few meters away from the workpiece. Edge computing applications are powered by fixed voltages, thereby allowing them to convert more energy than battery-powered applications. The processing power is correspondingly higher, enabling video analyses and high-frequency, often real-time calculations to be performed here.
The Intel Tiber Development Cloud offers developers access to numerous AI tools, such as OpenVino, OneAPI, PyTorch, TensorFlow, Llama2, and many other LLMs, frameworks, and toolkits. It is possible to design your own computer as a hardware platform, with Rutronik providing support for AI coprocessors using Intel Core Ultra processors (Fig. 2) or Intel Gaudi. Alternatively, a board from a third-party supplier that installed, verified, and certified the semiconductor chips can be used.
Edge systems for processing large amounts of data
Advantech AIR-030 is the next stage of expansion and represents a fully assembled AI computer for edge use. Based on Nvidia Jetson AGX Orin, the computer offers heavy industry certification pursuant to ISO 61000-6-4 as well as USB-C, CanBus, Ethernet, and various expansion options, e.g. using M.2 cards. In addition to this computer, Advantech provides 12 further edge systems – based on Nvidia Jetson Nano, TX2 NX, AGX Xavier, or Nvidia Jetson Xavier NX. The data prefiltered in this manner correspond to only a fraction of the volume of the raw data generated by the sensors. In many cases, data transport to data centers is much less expensive and faster. For certain applications, AI calculations require processing data from various global sources, meaning that data centers will remain essential and continue to grow despite the expansion of edge computing. With the MIC-735 series, Advantech is offering complete servers based on the Nvidia IGX Origin that enable fast calculations of up to 275 TOPS (Fig. 3). An Nvidia RTX A6000 additionally powers the server family for complex moving image analyses. With the MIC-71xx family from Advantech, Rutronik provides its customers with six development environments based on various Nvidia Jetson chips for customer-specific device development.
AI chips and boards for smart cameras
The ICAM-500, 520 and 540 models from Advantech clearly demonstrate what a camera can achieve when AI chips from Nvidia are combined with image sensors from Sony (Fig. 4). Up to 8 MP@50 FPS or 1.6 MP@60 FPS enable high-resolution video signals even with rapidly changing images. The camera recognizes and evaluates the filmed scenarios within its housing so that only the conclusions from the scene need to be transferred to the externally connected computer, thereby requiring less external processing power. AI cameras from Advantech complement Rutronik’s lineup of smart cameras, which have achieved significant market penetration in recent years due to the success of Intel RealSense cameras, camera modules, and chip solutions. Particularly due to the support offered by OpenVino and other AI software tools, Intel RealSense products also facilitate image analysis without excessively overloading the CPU. Adlink is another provider of AI cameras in the Rutronik portfolio, and its combination with Edge Vision Analytics software is equally interesting for a wide range of tasks.
When looking for edge computing systems to operate in particularly harsh environments, it is worth considering the Advantech MIC-71x family or the MIC-730IVA. This type of computer can capture and analyze up to eight video streams – even without a fan, which is crucial for use in dusty environments, for instance. The performance range spans from 512 GOPS to 21 and 32 TOPS, all the way up to 100 TOPS for the high-end MIC-715OX and MIC-717OX models. SECeDGE software on the Advantech computers protects the customer’s AI models, ensures data integrity, eliminates the need for an external TPM chip, and, at the same time, enables secure boot, firmware updates, failure recovery, encryption, credentials management, and cloud integration (Fig. 5).
Computer, processor, and boards for AI applications
An AI solution from the Italian PCB supplier Seco is the ideal choice for certain applications. Seco Titan 300 TGL-UP3 AI is a fanless multicore computer that utilizes Intel Core and Intel Celeron processors, allowing tasks to be performed on the most energy-efficient processing units (Fig. 6). Moreover, Seco has added an NPU as a coprocessor to the two CPUs. An Axelera Metis, also referred to as an AIPU (artificial intelligence processing unit), can deliver up to 120 TOPS. Via Technologies, Rutronik’s relatively new franchise partner, utilizes intelligent semiconductors from MediaTek for its AI boards. Via ARTiGO A5000 is an ultra-compact and fanless ARM system with eight performance cores (two MediaTek Genio 700 or two ARM Cortex-A78 units) for AI applications on the edge. Two AHD cameras can also be connected to Via SOM-5000 using MIPI and touch panels using LVDS and HDMI (Fig. 7). The 4K hardware acceleration supporting H.265/H.264 enables video compression and decompression without placing any additional load on the main computer. The four TOPS of the AI unit should be adequate for many video tasks, allowing for savings in both energy consumption and acquisition costs from the product selection stage onward.
For retrofitting existing workstations or servers with AI accelerators, the iEi Mustang-T100-T5 and Mustang-M2BM offer corresponding expansion cards. The T100 card utilizes five Coral Edge TPUs from Google, while the M2BM card performs calculations using two Intel Movidius Myriad X VPUs. The Google card already delivers 20 TOPS while consuming only about 15 W. Alongside iEi, Molex’s sister company, BittWare, also utilizes Intel technology with its IA-440i accelerator card. However, it uses an Agilex 7 FPGA chip from Altera, launched as a stand-alone FPGA company by Intel. The card is ideal for high-speed data storage tasks, making it well-suited for use in data loggers and the construction of storage servers.
Modular solutions for computationally intensive applications
AI has also made its way into the ever-popular segment of standardized open standard modules (OSM). With Advantech ROM-2620, ROM-2820, and ROM-2860, Rutronik offers three modules that are also referred to by the supplier as power saver, AI enabler, and transboundary star – providing tailored solutions for field calculations based on specific requirements. While the first two modules operate on the Yocto operating system, the last variant is available exclusively for Windows 11 on ARM or Ubuntu. All three OSM modules are guaranteed to be available until 2034, ensuring that today’s investment in development has ample time to yield a return on investment. The Yocto modules support ARM Cortex-A35 or A55 cores. The ROM-2820 can access a 500 GOPS Ethos-U65 NPU, which can execute AI algorithms up to ten times faster than the Cortex-A55 core (Fig. 8). The flagship ROM-2860 is based on a Qualcomm QCS6490, which contains the Kyro 670 CPU and the Hexagon NPU from the same company. The processing power is often used in dash cams, lawn mowers, vacuum cleaners, or industrial scanners for ultrafast camera analysis – the 12 TOPS make this possible. With ARM64 compatibility and integrated Wi-Fi 6E, prefiltered data can be seamlessly transmitted to Microsoft Azure Cloud or leveraged with Azure AI Services, along with an extensive range of other development tools.
Even larger form factors, such as computer-on-modules, have long been available from Rutronik with integrated AI. The Advantech AOM-3511, with five TOPS, is just 5 cm deep, whereas the AOM-3411 has a depth of 80 mm. The former is based on an ARM Cortex Multicore. Six A55@2 GHz, one M7@800 MHz and one M33@250 MHz allow tasks to be addressed by the most appropriate processing unit. The AOM-3411 goes a step further: Powered by a Rockchip RK3576, four A72@2.2 GHz, four A53@1.8 GHz and one M0@800 MHz are available here. The integrated NPU performs up to six TOPS.
The variety of microcontroller architectures and the operating systems available can still be somewhat overwhelming for newcomers, those transitioning, or those returning to the industry. In June 2024, the first laptops utilizing ARM architectures rather than x86 were introduced, featuring Microsoft Windows 11 on ARM in the substructure and Microsoft Copilot in the superstructure. All laptop suppliers, including Microsoft with its Surface models, rely on the new Qualcomm Snapdragon X Elite processor as the drive unit. This new processor is also already used in the AOM-7721 computer-on-module from Advantech (Fig. 9). With a mighty impressive 45 TOPS, AI performance in end devices now rivals what was once only achievable in energy-intensive servers. The Advantech RSB-3810 is a single-board computer in the 2.5"" Pico-ITX form factor, ideal for those who are unable to use a baseboard for computer-on-modules. The computer is powered by a MediaTek Genio 1200, which contains a CPU, VPU, and MDLA, and is, therefore, also suitable for more computationally intensive operations, especially those based on TensorFlow Lite.
QR code for the “AI Highlights” brochure
Take a direct look at the AI Highlights brochure
Link: https://www.rutronik.com/downloads/brochures
Product overview in the new Rutronik AI Highlights brochure
If this overview of our product categories, ranging from compact AI microcontrollers to robust AI servers, has piqued your interest in the full spectrum of performance, the Rutronik AI Highlights brochure provides an informative comparison. It also includes products from DFI, AAEON, Cherry Embedded Solutions, Adlink, and specialized storage solutions from various suppliers. With this topic in particular, it is worth noting that as the world continues to spin faster and faster, Rutronik is expanding its product range with new AI solutions almost every day. As such, a chat with the experts at Rutronik cannot be replaced by AI or the catalog. Please do not wait until the official end of support for Windows 10, as prices and delivery times are likely to increase once again.
For more information and a direct ordering option, please visit our e-commerce platform at www.rutronik24.com.
Subscribe to our newsletter and stay updated.