Grinn AstraOneSBC-1680 + DEEPX

Achieving 25 TOPS of native INT8 compute performance with the DEEPX DX-M1 M.2 accelerator alongside the integrated 8 TOPS NPU, the AstraOneSBC-1680 platform is built for compute-intensive edge AI workloads that demand higher inference capacity.

The DEEPX DX-M1 incorporates a purpose-built deep learning architecture with a high-bandwidth memory subsystem, enabling efficient processing of advanced neural networks. Through the DEEPX dataflow compiler, models from TensorFlow, ONNX, Keras, and PyTorch can be seamlessly optimized and deployed.

Based on the Synaptics Astra platform, AstraOneSBC-1680 integrates a quad-core CPU and comprehensive connectivity, including mini-HDMI, Gigabit Ethernet, dual USB 3.0, CSI camera input, dual M.2 expansion, and a 40-pin header. This compact Pico-ITX solution provides the performance and flexibility required for next-generation industrial and vision-based edge AI systems.

WEEE Registration number: DE 79077745.

Wiki page

Datasheet

Image Recognition
High-performance AI
Smart Cameras
Robotics
AI-supported computer vision
Industrial automation

Processor
- 4× Arm Cortex-A73 @ 2.1 GHz, 64-bit
DEEPX Accelerator
- DEEPX DX-M1
- 25 TOPS INT8 compute
Internal AI Accelerator
- Integrated NPU (up to 8 TOPS)
RAM
- 4GB LPDDR4
Flash / Storage
- 16GB eMMC
Display
- mini-HDMI
- 2160p60
- HDR10+
- HDCP 2.3
- CEC
- eARC/eARC Rx
Network
- 1× RJ-45 GbE
- RGMII PHY
- diagnostics LEDs
USB
- 2× USB-A 3.0
- 1× USB-C 2.0
Camera
- 1× 4-lane MIPI CSI-2
- 2.5 Gbps/lane
Wireless / AI Expansion
- M.2 - DEEPX DX-M1
- M.2 (E-key USB/SDIO/UART)
GPIO
- 1× 40-pin header
- 3.3 V logic
Security
- TrustZone + Infineon OPTIGA Trust M
Power
- 12…24 V DC (2A recommended supply)
- passive cooling
Power consumption
- 2-5 W when running DEEPX-optimized AI models
Temp Support
- Extended: −20…+70 °C
- Industrial: −40…+85 °C target
Dimensions
- 100 × 72 mm (Pico-ITX)

ready-to-use

Other products

Grinn SitaraSOM-335 (ChiliSOM)

System On Module for IoT An ultra-small, extremely low-power, state-of-the-art module based on ARM Cortex-A8 processor

Grinn EvseSOM RPI HAT

The Grinn EvseSOM RPI HAT is a Raspberry Pi-compatible development and evaluation platform for integrating the Grinn EvseSOM with a host processor through a 40-pin GPIO interface. Designed for eMobility applications, it provides UART, Control Pilot (CP), Proximity Pilot (PP), and safety control interfaces for EV charging development and testing.

Grinn EvseSOM EVB

The EvseSOM EVB, built around the Grinn EvseSOM, is a comprehensive development platform designed for rapid prototyping and evaluation of eMobility charging applications. It enables developers to quickly start working with different Grinn EvseSOM™ variants by providing all required interfaces, power management, and communication peripherals in a ready-to-use environment.