ArmSoM's RK3588 AI Module7 Offers a Drop-In Alternative to NVIDIA's Jetson Nano System-on-Module
Company boasts of broader connectivity, a faster CPU, and a dedicated NPU accelerator — but there's no CUDA support.
Shenzhen-based open hardware and software firm ArmSoM is preparing to launch an alternative to NVIDIA's Jetson Nano system-on-module, designed as a drop-in pin-compatible alternative: the RK3588 AI Module7.
"[The] RK3588 AI Module7 […] is compatible with NVIDIA's Jetson Nano interface with upgraded and improved PCIe connectivity," the company claims of its creation. "Due to this module's versatility, it caters to applications such as edge computing, cloud servers, AI [Artificial Intelligence], VR/AR [Virtual Reality/Augmented Reality], blockchain, and smart NVR [Network Video Recording] security systems, not to mention general-purpose computing."
The RK3588 Module7, brought to our attention by Linux Gizmos, is powered by Rockchip's RK3588 system-on-chip — giving it four high-performance Arm Cortex-A76 cores running at up to 2.4GHz and four energy-efficient Cortex-A55 cores running at a slower 1.8GHz. There's an Arm Mali-G610 MP4 graphics processor, and a dedicated neural processing unit (NPU) accelerator that delivers a claimed six tera-operations per second (TOPS) of compute at minimum precision for on-device machine learning and artificial intelligence workloads.
The company is positioning the module, which comes with a choice of 8GB or 32GB of LPDDR4x memory, as a upgrade over NVIDIA's entry-level Jetson Nano module — offering a higher-performance CPU, dedicated NPU, and more memory, but lacking the 128-core Maxwell CUDA-capable GPU of NVIDIA's design. The module also offers 8k30 H.264/H.265 video encoding and 8k60 H.265/VP9/AVS2, 8k30 H.264 AVC/MVC, 4k60 AVI, and 1080p MPEG-2/-1/VC-1/VP8 video decoding.
Elsewhere on the module are pins for HDMI 2.1, DisplayPort 1.4a, and embedded DisplayPort (eDP) 1.4 video outputs, a single MIPI Display Serial Interface (DSI) and 4×2 Camera Serial Interface (CSI) inputs, one USB 3.0 and three USB 2.0 ports, gigabit Ethernet, and four lanes of PCI Express Gen. 3 and one lane of PCI Express Gen. 2 connectivity, plus general-purpose input/output (GPIO) pins including three UART, two SPI, two I2S, and four I2C buses. For storage, there's a microSD card slot plus 32GB, 64GB, or 128GB of on-board eMMC 5.1 flash.
To support all this connectivity, ArmSoM has also developed a carrier board which mimics the design of the NVIDIA Jetson Nano Developer Kit but with additional connectivity to make the most use of the features of the Module7. The company has also pledged to release board schematics during the course of its crowdfunding campaign, though has not yet specified a license.
More information is available on the project's Crowd Supply page, ahead of the opening of its crowdfunding campaign.