A current International Market Insights report initiatives the one board pc market will develop from $600 million in 2018 to $1 billion by 2025. But, you don’t must learn a market analysis report to appreciate the SBC market is booming. Pushed by the developments towards IoT and AI-enabled edge computing, new boards maintain rolling off the meeting strains, a lot of them tailor-made for extremely particular purposes.
A lot of the motion has been in Linux-compatible boards, together with the insanely common Raspberry Pi. The variety of totally different distributors and fashions has exploded thanks partially to the rise of community-backed, open-spec SBCs.
Right here we study 10 of probably the most intriguing, Linux-driven SBCs among the many many merchandise introduced within the final 4 weeks that bookended the current Embedded World present in Nuremberg. (There was additionally some fascinating Linux software program information on the present.) Two of the SBCs—the Intel Whiskey Lake primarily based UP Xtreme and Nvidia Jetson Nano pushed Jetson Nano Dev Equipment—have been introduced solely this week.
Our principally open supply record additionally features a few industrial boards. Processors vary from the modest, Cortex-A7 pushed STM32MP1 to the high-powered Whiskey Lake and Snapdragon 845. Mid-range fashions embrace Google’s i.MX8M powered Coral Dev Board and a equally AI-enhanced, TI AM5729 primarily based BeagleBone AI. Deep studying acceleration chips—and customary RPi 40-pin or 96Boards enlargement connectors—are frequent themes amongst most of those boards.
The SBCs are listed in reverse chronological order in response to their announcement dates. The hyperlinks within the product names go to current LinuxGizmos stories, which hyperlink to vendor product pages.
UP Xtreme—The newest in Aaeon’s line of community-backed SBCs faucets Intel’s eighth Gen Whiskey Lake-U CPUs, which preserve a modest 15W TDP whereas boosting efficiency with as much as quad-core, twin threaded configurations. Relying on when it ships, this Linux-ready mannequin will seemingly be probably the most highly effective community-backed SBC round — and probably the most costly.
The SBC helps as much as 16GB DDR4 and 128GB eMMC and affords 4K shows through HDMI, DisplayPort, and eDP. Different options embrace SATA, 2x GbE, 4x USB three.zero, and 40-pin “HAT” and 100-pin GPIO add-on board connectors. You additionally get mini-PCIe and twin M.2 slots that assist wi-fi modems and extra SATA choices. The slots additionally assist Aaeon’s new AI Core X modules, which supply Intel’s newest Movidius Myriad X VPUs for 1TOPS neural processing acceleration.
Jetson Nano Dev Equipment—Nvidia simply introduced a low-end Jetson Nano compute module that’s form of like a smaller (70 x 45mm) model of the outdated Jetson TX1. It affords the identical 4x Cortex-A57 cores however has a good lower-end 128-core Maxwell GPU. The module has half the RAM and flash (4GB/16GB) of the TX1 and TX2, and no WiFi/Bluetooth radios. Just like the hexa-core Jetson TX2, nonetheless, it helps 4K video and the GPU affords comparable CUDA-X deep studying libraries.
Though Nvidia has backed all its Linux-driven Jetson modules with growth kits, the Jetson Nano Dev Equipment is its first community-backed, maker-oriented equipment. It doesn’t seem to supply open specs, but it surely prices solely $99 and there’s a discussion board and different neighborhood sources. Lots of the specs match or surpass the Raspberry Pi 3B+, together with the addition of a 40-pin GPIO. Highlights embrace an M.2 slot, GbE with Energy-over-Ethernet, HDMI 2.zero and eDP hyperlinks, and 4x USB three.zero ports.
Coral Dev Board—Google’s very first Linux maker board arrived earlier this month that includes an NXP i.MX8M and Google’s Edge TPU AI chip—a stripped-down model of Google’s TPU Unit is designed to run TensorFlow Lite ML fashions. The $150, Raspberry Pi-like Coral Dev Board was joined by a equally Edge TPU-enabled Coral USB Accelerator USB stick. These will likely be adopted by an Edge TPU primarily based Coral PCIe Accelerator and a Coral SOM compute module. All these units are backed with schematics, neighborhood sources, and different open-spec sources.
The Coral Dev Board combines the Edge TPU chip with NXP’s quad-core, 1.5GHz Cortex-A53 i.MX8M with a 3D Vivante GPU/VPU and a Cortex-M4 MCU. The SBC is much more just like the Raspberry Pi 3B+ than Nvidia’s Dev Equipment, mimicking the dimensions and far of the format and I/O, together with the 40-pin GPIO connector. Highlights embrace 4K-ready GbE, HDMI 2.0a, Four-lane MIPI-DSI and CSI, and USB three.zero host and Kind-C ports.
SBC-C43—Seco’s industrial, industrial temperature SBC-C43 board is the primary SBC primarily based on NXP’s high-end, as much as hexa-core i.MX8. The three.5-inch SBC helps the i.MX8 QuadMax with 2x Cortex-A72 cores and 4x Cortex-A53 cores, the QuadPlus with a single Cortex-A72 and 4x -A53, and the Quad with no -A72 cores and 4x -A53. There are additionally 2x Cortex-M4F real-time cores and 2x Vivante GPU/VPU cores. Yocto Mission, Wind River Linux, and Android can be found.
The feature-rich SBC-C43 helps as much as 8GB DDR4 and 32GB eMMC, each soldered for larger reliability. Highlights embrace twin GbE, HDMI 2.0a out and in ports, WiFi/Bluetooth, and quite a lot of industrial interfaces. Twin M.2 slots assist SATA, wi-fi, and extra.
Nitrogen8M_Mini—This Boundary Units cousin to the sooner, i.MX8M primarily based Nitrogen8M is accessible for $135, with shipments due this Spring. The open-spec Nitrogen8M_Mini is the primary SBC to characteristic NXP’s new i.MX8M Mini SoC. The Mini makes use of a extra superior 14LPC FinFET course of than the i.MX8M, leading to decrease energy consumption and better clock charges for each the 4x Cortex-A53 (1.5GHz to 2GHz) and Cortex-M4 (400MHz) cores. The downside is that you simply’re restricted to HD video decision.
Supported with Linux and Android, the Nitrogen8M_Mini ships with 2GB to 4GB LPDDR4 RAM and 8GB to 128GB eMMC. MIPI-DSI and -CSI interfaces assist non-obligatory touchscreens and cameras, respectively. A GbE port is customary and PoE and WiFi/BT are non-obligatory. Different options embrace 3x USB ports, one or two PCIe slots, and non-obligatory -40 to 85°C assist. A Nitrogen8M_Mini SOM module with comparable specs can be within the works.
Pine H64 Mannequin B—Pine64’s newest hacker board was teased in late January as a part of an formidable roll-out of open supply merchandise, together with a laptop computer, pill, and telephone. The Raspberry Pi semi-clone, which not too long ago went on sale for $39 (2GB) or $49 (3GB), showcases the high-end, however low-cost Allwinner H64. The quad -A53 SoC is notable for its 4K video with HDR assist.
The Pine H64 Mannequin B affords as much as 128GB eMMC storage, WiFi/BT, and a GbE port. I/O consists of 2x USB 2.zero and single USB three.zero and HDMI 2.0a ports plus SPDIF audio and an RPi-like 40-pin connector. Photographs embrace Android 7.zero and an “in progress” Armbian Debian Stretch.
AI-ML Board—Arrow unveiled this i.MX8X primarily based SBC early this month together with a equally 96Boards CE Prolonged format, i.MX8M primarily based Thor96 SBC. Whereas there are many i.MX8M boards today, we’re extra intrigued with the lowest-end i.MX8X member of the i.MX8 household. The AI-ML Board is the primary SBC we’ve seen to characteristic the low-power i.MX8X, which affords as much as 4x 64-bit, 1.2GHz Cortex-A35 cores, a Four-shader, 4K-ready Vivante GPU/VPU, a Cortex-M4F chip, and a Tensilica HiFi Four DSP.
The open-spec, Yocto Linux pushed AI-ML Board is focused at low-power, camera-equipped purposes reminiscent of drones. The board has 2GB LPDDR4, Ethernet, WiFi/BT, and a pair every of MIPI-DSI and USB three.zero ports. Cameras are managed through the 96Boards 60-pin, high-power GPIO connector, which is joined by the standard 40-pin low-power hyperlink. The launch is predicted June 1.
BeagleBone AI—The long-awaited successor to the Cortex-A8 AM3358 primarily based BeagleBone household of boards advances to TIs dual-core Cortex-A15 AM5729, with comparable PowerVR GPU and MCU-like PRU cores. The true story, nonetheless, is the AI firepower enabled by the SoC’s twin TI C66x DSPs and 4 embedded-vision-engine (EVE) neural processing cores. BeagleBoard.org claims that calculations for computer-vision fashions utilizing EVE run at 8x instances the efficiency per watt in comparison with the same, however EVE-less, AM5728. The EVE and DSP chips are supported by means of a TIDL machine studying OpenCL API and pre-installed instruments.
As a result of go on sale in April for about $100, the Linux-powered BeagleBone AI relies intently on the BeagleBone Black and affords backward header, mechanical, and software program compatibility. It doubles the RAM to 1GB and quadruples the eMMC storage to 16GB. You now get GbE and high-speed WiFi, in addition to a USB Kind-C port.
Robotics RB3 Platform (DragonBoard 845c)—Qualcomm and Thundercomm are initially launching their 96Boards CE kind issue, Snapdragon 845-based improve to the Snapdragon 820-based DragonBoard 820c SBC as a part of a Qualcomm Robotics RB3 Platform. But, 96Boards.org has already posted a DragonBoard 845c product web page, and we think about the board will likely be accessible within the coming months with out all of the robotics bells and whistles. A compute module model can be stated to be within the works.
The 10nm, octa-core, “Kryo” primarily based Snapdragon 845 is among the strongest Arm SoCs round. It options a sophisticated Adreno 630 GPU with “eXtended Actuality” (XR) VR expertise and a Hexagon 685 DSP with a third-gen Neural Processing Engine (NPE) for AI purposes. On the RB3 equipment, the board’s enlargement connectors are pre-stocked with Qualcomm mobile and robotics digital camera mezzanines. The $449 and up equipment additionally consists of customary 4K video and monitoring cameras, and there are non-obligatory Time-of-Flight (ToF) and stereo SLM digital camera depth cameras. The SBC runs Linux with ROS (Robotic Working System).
Avenger96—Like Arrow’s AI-ML Board, the Avenger96 is a 96Boards CE Prolonged SBC geared toward low-power IoT purposes. But, the SBC options an much more power-efficient (and slower) SoC: ST’s not too long ago introduced STM32MP153. The Avenger96 runs Linux on the high-end STM32MP157 mannequin, which has twin, 650MHz Cortex-A7 cores, a Cortex-M4, and a Vivante 3D GPU.
This sandwich-style board options an Avenger96 module with the STM32MP157 SoC, 1GB of DDR3L, 2MB SPI flash, and an influence administration IC. It’s unclear if the 8GB eMMC and WiFi-ac/Bluetooth Four.2 module are on the module or provider board. The Avenger96 SBC is additional outfitted with GbE, HDMI, micro-USB OTG, and twin USB 2.zero host ports. There’s additionally a microSD slot and the standard 40- and 60-pin GPIO connectors. The board is predicted to go on sale in April.