High-Performance Computing Systems Packed into Compact
Form Factors
System-on-Module (SoM) technology is gaining traction across various industries, driven by the demand for compact, integrated solutions that allow developers to add AI and connectivity options as demands evolve. Explore the factors fueling the SoM growth and the advantages these modules offer.
In the era when inferencing is needed precisely where information is being collected and processed, it’s important to remember that the latency of the cloud is the most important aspect of Artificial Intelligence that system designers want to remove from their system performance equations. How else could we bring massive amounts of computing performance closer to the action than with a system-on-module (SoM) technology? It can deliver multiple terra operations per second (TOPS) right where the application needs it most, according to the literature supporting many of the datasheets profiled here.
On the following pages, you will see SoMs from AAeon, Axiomtek, BeagleBoard, Nvidia, Toradex, Variscite, SolidRun, and even the latest Raspberry Pi CM5.
Processors steal the spotlight of many SoMs, and here, you’ll find representatives from Intel, AMD, NXP, Broadcom, and Nvidia. If you see processor vendors here that aren’t usually associated with PC platforms, they were most like likely selected for efficiency and longetivity.
SoM presents the ultimate method of bringing computing power right to the edge. Its signature compact size characteristics entice developers to choose full-fledged PC/server-type compute density where only a cue-card-sized space is available.
If you need to speed up the pace of machine vision visual inspection rapidly, an SoM product on the market today will give you that capability. Suppose your automated guided vehicle (AGV) keeps delivering burgers and fries to the wrong address. In that case, it’s possible to swap out the old SoM and replace it with one producing next-generation navigation capabilities to ensure the AGV gets to the right place without slamming into someone’s shin bones.
The primary driver behind the increasing adoption of SOM products is the need for space-efficient and cost-effective computing solutions. SoMs integrate essential CPUs, memory, and I/O interfaces into a tiny card-sized board that is easy to remove and replace. This compact design is particularly beneficial for applications in the Internet of Things (IoT), industrial automation, and consumer electronics, where minimizing physical footprint without sacrificing performance is crucial.
Continuous improvements in semiconductor technology significantly enhance the performance and appeal of SoMs. Innovations in processing power and energy efficiency allow SoMs to meet the demands of next-generation applications that require higher computational capabilities. These advancements make SoMs attractive for developers looking to build more powerful and energy-efficient devices.
The proliferation of IoT devices and the shift toward edge computing are significant factors driving SoM adoption. As businesses seek powerful processing capabilities and reliable connectivity for real-time data processing, SoMs provide the necessary infrastructure. They enable data analysis and decision-making closer to the data source, which is critical for applications requiring immediate responses.
SoMs facilitate quicker product development cycles by simplifying hardware design. Manufacturers can customize functionalities to differentiate their products and respond swiftly to market demands. This flexibility is essential for companies that maintain a competitive edge in fast-paced environments.
In the context of Industry 4.0, SoMs are pivotal in driving automation and connectivity within smart manufacturing. They support robotics, predictive maintenance, and digital twins, which enhance productivity and operational efficiency. By integrating SoMs into manufacturing systems, companies can leverage data for improved decision-making and resource management.
AAEON UP Xtreme i11
The UP Xtreme i11 is powered by Intel’s 11th Gen Core processors, providing high performance for AI and edge computing applications. With support for up to 32GB of DDR4 RAM, it offers extensive I/O options, including dual HDMI outputs and USB 3.0. The board is designed for reliability and efficiency, making it suitable for demanding environments. Its compact size and flexible expansion options make it ideal for embedded systems in various applications, including smart cities and industrial automation. The UP Xtreme i11 supports a wide range of peripherals, enabling developers to create customized solutions.
- Intel 11th Gen Core (i3/i5/i7) processor
- Up to 32GB DDR4 RAM
- Up to 1TB NVMe SSD storage
- Dual HDMI 2.0 outputs
- Dual Gigabit Ethernet ports
- PCIe and M.2 expansion slots
- Compact dimensions: 90mm x 90mm
- Low power consumption
- Wide operating temperature range
- Extensive software support
- Real-time clock support
- Integrated security features
- Multiple audio interfaces
- High thermal efficiency
- Ideal for smart city applications
- Flexible power management
Axiomtek CAPA13
The Axiomtek CAPA13 is powered by AMD Ryzen Embedded V1807B/V1605B APUs, providing needed performance for industrial applications. With support for up to 32GB of DDR4 RAM, it offers a variety of I/O options, including dual HDMI outputs and USB 3.0. The board is designed for high reliability and efficiency, making it suitable for demanding environments. Its compact size and extensive connectivity make it an ideal choice for embedded systems in automation, transportation, and smart city applications. The CAPA13 also supports a range of expansion options for customization.
- AMD Ryzen Embedded V1807B/V1605B APUs
- AMD Radeon RX Vega Graphics
- Up to 32GB DDR4 RAM
- Up to 1TB SSD storage
- Dual HDMI 2.0 outputs
- DisplayPort++
- 1 x LVDS; 18/24-bit single/dual channel
- Dual Gigabit Ethernet ports
- PCIe expansion slots
- HD audio link
- Temp range: -20°C to +60°C (-4°F to 140°F)
- Compact dimensions: 146mm x 104mm
- Wide operating temperature range
- Extensive software support
- Designed for industrial applications
- Real-time clock support
- Integrated security features
- Multiple audio interfaces
- High thermal efficiency
- Flexible power management
BeagleBoard BeagleBone AI-64
The BeagleBone AI-64 features the Texas Instruments’ AM572x Sitara, a dual-core ARM Cortex-A15 CPU and additional AI processing capabilities, making it ideal for edge AI applications. With a focus on machine learning, it provides a rich set of I/O options, including HDMI and dual Gigabit Ethernet ports. An open-source software environment encourages innovation and collaboration. The compact design and extensive GPIO support enable integration into various projects, from robotics to smart devices. Designed for reliability, the BeagleBone AI-64 is suitable for both educational and industrial applications.
- Dual-core ARM Cortex-A15 CPU
- Up to 1GB DDR3 RAM
- 16GB eMMC storage
- HDMI 2.0 output
- Dual Gigabit Ethernet ports
- 92 GPIO pins
- SPI, I2C, and UART interfaces
- Compact dimensions: 90mm x 60mm
- Low power consumption
- Extensive software support
- Designed for AI and machine learning applications
- Community-driven development resources
- Real-time clock support
- Integrated security features
- Multiple audio interfaces
- High thermal efficiency
- Ideal for robotics and smart devices
- Wide operating temperature range
iWave Systems i.MX 8M Plus SoM
The i.MX 8M Plus SoM features a quad-core ARM Cortex-A53 CPU and integrated NPU for AI processing, making it suitable for multimedia and edge computing applications. With support for up to 8GB of LPDDR4 RAM, it provides robust performance for various tasks. The module includes HDMI 2.0, USB 3.0, and Gigabit Ethernet for versatile connectivity. Its compact 82mm x 50mm design ensures easy integration into space-constrained applications, while its low power consumption promotes energy efficiency. The i.MX 8M Plus is ideal for developers looking to implement AI and multimedia solutions in their projects.
- Quad-core ARM Cortex-A53 CPU
- Up to 8GB LPDDR4 RAM
- Up to 64GB eMMC storage
- HDMI 2.0 output
- USB 3.0 and PCIe support
- Gigabit Ethernet port
- Compact dimensions: 60mm x 25mm
- Low power consumption
- Wide operating temperature range
- Extensive software support
- Integrated NPU for AI processing
- Real-time clock support
- Multiple audio interfaces
- SPI, I2C, and UART interfaces
- Ideal for multimedia applications
- Community-driven development resources
- High thermal efficiency
- Flexible power management
Nvidia Jetson Orin Nano
The Jetson Orin Nano is designed for AI applications, featuring a six-core ARM Cortex-A78 CPU and up to 8GB of LPDDR5 RAM. A 512-core NVIDIA Ampere architecture GPU with 16 tensor cores supporting up to 67 TOPS make it suitable for edge computing, robotics, and smart devices. The board supports multiple I/O options for high-performance multimedia applications. The Orin Nano designed for low power consumption while delivering high-performance for developers looking to AI in real-world applications.
- 6-core ARM Cortex-A78 CPU
- Up to 8GB LPDDR5 RAM
- 16GB eMMC storage
- HDMI 2.1 output
- USB 3.2 Type-C ports
- Gigabit Ethernet port
- M.2 Key M slot for NVMe SSD
- 40 GPIO pins
- SPI, I2C, and UART interfaces
- Hardware video decoding support (H.265, H.264)
- Integrated AI accelerators
- Power management integrated circuit
- Real-time clock support
- Compact dimensions: 100mm x 80mm
- JetPack SDK
- Applications: Robotics, drones, AI applications
- Low power consumption design
- High thermal efficiency
Raspberry Pi Compute Module 5
The Raspberry Pi Compute Module 5 features Broadcom’s quad-core ARM Cortex-A76 CPU, and with RAM options up to 16GB. To enhance the capabilities of embedded systems, the CM5 provides dual HDMI outputs and improved connectivity options. The module retains compatibility with existing peripherals while expanding support for PCIe devices. Designed for a variety of applications, from industrial automation to DIY projects, the Compute Module 5 will offer a powerful platform for developers looking to create innovative solutions. Extensive community support ensures resources for rapid development and deployment.
- Quad-core ARM Cortex-A76 CPU
- Up to 16GB RAM
- Dual HDMI outputs
- USB 3.0 and PCIe support
- Gigabit Ethernet port
- GPIO pins for extensive connectivity
- Compact dimensions: 55mm x 40mm
- Low power consumption
- Wide operating temperature range
- Extensive software support
- Community-driven development resources
- High-performance multimedia capabilities
- Integrated security features
- Real-time clock support
- Multiple audio interfaces
- Flexible power management
- Support for various display interfaces
SolidRun HoneyComb LX2K
SolidRun’s HoneyComb LX2K features a 16-core ARM Cortex-A72 CPU, designed for high-performance applications in networking and edge computing. With support for up to 32GB of DDR4 RAM, it provides substantial processing power for data-intensive tasks. The module includes multiple I/O options, including PCIe and Gigabit Ethernet. Its compact design and energy efficiency make it suitable for various applications, from telecommunications to industrial automation. The HoneyComb LX2K is optimized for scalability, enabling creation of robust solutions tailored to specific needs.
- 16-core ARM Cortex-A72 CPU
- Up to 32GB DDR4 RAM
- Up to 64GB eMMC storage
- Dual Gigabit Ethernet ports
- PCIe expansion slots
- Compact dimensions: 100mm x 100mm
- Low power consumption
- Wide operating temperature range
- Extensive software support
- Designed for networking and edge computing
- Real-time clock support
- Integrated security features
- High thermal efficiency
- Multiple audio interfaces
- Ideal for industrial automation
- Flexible power management
- High-performance data processing capabilities
Toradex Colibri iMX8M Plus
The STM32L0 series is STMicroelectronics’ entry-level range of 32-bit MCUs designed for ultra-low power consumption levels as low as 49µA/MHz with external DC/DC. With dynamic voltage scaling, the STM32L051x6 with an ARM-based Cortex-M0+, with 64KB Flash, 8KB SRAM, 2KB EEPROM, and an ADC is aimed at energy harvesting, smoke detectors, gas and water meters and consumer devices with advanced battery management and fast charging.
- Quad-core ARM Cortex-A53 CPU
- Up to 8GB LPDDR4 RAM
- Up to 64GB eMMC storage
- USB 3.0 and USB 2.0 ports
- HDMI 2.0 output
- Gigabit Ethernet port
- 40 GPIO pins
- SPI, I2C, and UART interfaces
- Compact dimensions: 67.6mm x 36.5mm
- Wide temperature range: -40°C to +85°C
- Integrated GPU for multimedia processing
- Low power consumption
- Real-time clock support
- Security features for data protection
- Extensive software support
- Ideal for industrial and IoT applications
- Community-driven resources
- Multiple audio interfaces
Variscite DART-MX8M Plus
The Variscite DART-MX8M Plus pushes a quad-core ARM Cortex-A53 CPU with robust performance for multimedia applications. Up to 8GB of LPDDR4 RAM and 64GB eMMC storage powers high-performance for industrial and consumer embedded systems. Interfaces for HDMI 2.0, USB 3.0, and Gigabit Ethernet offer versatile connectivity and a compact footprint is suitable for space-constrained applications, while low power consumption ensures efficient operation. demands are many and supply is finite.
- Quad-core ARM Cortex-A53 CPU
- Up to 8GB LPDDR4 RAM
- Up to 64GB eMMC storage
- HDMI 2.0 output
- USB 3.0 and USB 2.0 ports
- Gigabit Ethernet port
- MIPI DSI and CSI interfaces
- Compact dimensions: 50mm x 25mm
- Low power consumption
- Wide temperature range: -40°C to +85°C
- Integrated GPU for multimedia processing
- Multiple audio interfaces
- Real-time clock support
- Security features for data protection
- Ideal for multimedia and IoT
- SDIO/MMC
- 1x PCIe v3.0
- Resistive touch controller
- Serial interfaces (ECSPI, FlexSPI, I2C, UART, CAN, JTAG)
- Single power supply: 3.3V
- Industrial temperature range: -40°C to 85°C
— ADVERTISMENT—
—Advertise Here—
Datasheet URLs:
AAEON: https://www.aaeon.com/en/product/detail/developer-board-intel-11th-up-xtreme-i11
Axiomtek: https://www.axiomtek.com/Download/Spec/en-US/capa13r.pdf
BeagleBoard: https://docs.beagleboard.org/boards/beaglebone/ai/03-design-and-specifications.html
Iwave Systems: https://d3pwr7ye9vqskz.cloudfront.net/wp-content/uploads/2023/08/iW-RainboW-G40M_i.MX_8M_Plus_SMARC_SOM-HardwareUserGuide-R2.0-REL1.0.pdf
Nvidia: https://www.nvidia.com/en-us/autonomous-machines/embedded-systems/jetson-orin/#tech-specs
Raspberry Pi: https://datasheets.raspberrypi.com/cm5/cm5-product-brief.pdf
SolidRun: https://www.solid-run.com/wp-content/uploads/2023/01/HoneyComb-LX2-Datasheet-2023.pdf
Tordex: https://docs.toradex.com/108784-verdin_imx8m_plus_datasheet.pdf
Varascite: https://www.variscite.com/wp-content/uploads/2020/10/VAR-SOM-MX8M-PLUS_Datasheet.pdf
PUBLISHED IN CIRCUIT CELLAR MAGAZINE • FEBRUARY 2025 #415 – Get a PDF of the issue
Sponsor this Article
Tom Murphy has been plying his trade as a technology journalist for more than 25 years, first as an editor/writer/reporter and then as a communications professional. He has earned industry recognition for both journalistic endeavors and technology campaigns. Before joining Circuit Cellar magazine as editor-in-chief (2024), Tom wrote tech briefs, newsletters, blogs, press releases, white papers, and technical articles for numerous clients, primarily in the semiconductor industry. His introduction into the industry came as an editor for “Electronic News” covering microprocessor companies, the fabless semiconductor phenomena, communication chips, and many others. When he’s not tapping away at the keyboard, Tom learns Dungeons and Dragons terminology from his 10-year-old daughter and walks his bulldog (slowly) around the block.
