Embedded computing SolidRun has announced a new range of systems-on-modules (SOMs) built around the Renesas RZ/G2 system-on-chip family — offering the performance required, the company claims, to handle next-generation human-machine interfaces with artificially intelligent assistance.
“Network security cameras, video doorbells, smart office controllers, and more all require a few things in common: A powerful processor, image and video processing capabilities, a fast network connection, and varying levels of AI inference processing support,” claims Atai Ziv, PhD, SolidRun’s chief executive, in support of the devices the company has unveiled at Embedded World.
SolidRun has partnered with Renesas for a new low-power SOM, designed for “AI-enhanced HMI” and the IoT. (📷: SolidRun)
“Working with Renesas,” Ziv continues, “we’ve developed an affordable and surprisingly potent module based on their RZ/G2 line of MPUs that not only satisfies these requirements but adds serious value-adding capabilities for the price point. Furthermore, these new SOMs are pin-to-pin compatible with other SOMs we offer, which provides customers new platform options to consider for their products.”
SolidRun’s RZ/G2LC SOM offers two Arm Cortex-A55 application cores running at 1.2GHz and with floating-point and single instruction multiple data (SIMD) accelerators alongside a single Cortex-M33 real-time core and a Mali-G31 graphics processor. There’s 1GB of DDR4 RAM with in-line Error Correcting Code (ECC) support, and 8GB of eMMC storage alongside SD Card expansion.
The SoC has no on-board neural network accelerator, but instead leverages the GPU to boost AI performance. (📷: SolidRun)
The SOM also offers two USB 2.0 lanes, 100-base-T Ethernet, dual-band 802.11a/b/g/n and Bluetooth 4.2 radios, a single MIPI DSI display output and a CSI-2 camera input, one digital audio interface, three I2C buses, two SPI buses, three UART buses, two CAN buses, and general-purpose input/output (GPIO) capabilities with pulse-width modulation (PWM) support.
SolidRun is positioning the low-power parts as ideal for “AI-enhanced human-machine interface (HMI) applications, industrial and building automation, video surveillance, IoT [Internet of Things] solutions, and more” — though it comes without a dedicated coprocessor for machine learning workloads. Instead, the company is pointing to the GPU as the ideal solution — offering “advanced AI inference execution processing capabilities,” which make the whole system a claimed six times faster than previous Arm Cortex-A53 alternatives.
A development kit, which includes SOM, carrier, and accessories, is in the works but not yet released. (📷: SolidRun)
“We developed our RZ/G2 family of general-purpose microprocessors to serve a variety of entry-level HMI products, while giving much thought to lowering systems costs and maximizing performance and overall capabilities,” says Renesas’ Frank Urbe.
“We’re excited to work with SolidRun to bring a new line of SOMs based on our RZ/G2 family of MPUs [Microprocessors] to market. These compact SOMs are sure to help customers quickly take advantage of the many unique benefits of our powerful and cost-effective MPUs and will shave months of development time off roadmaps.”
More information on the SolidRun SOMs, which are drop-in compatible with carrier boards designed for its earlier releases, can be found on the company website. The company has also pledged to release an RZ/G2LC Starter Kit with HummingBoard Pulse carrier board, the RZ/G2LC SOM, and accessories — but has not yet announced pricing and availability.
