Banana Pi BPI-SM9 16-ENC-A3 SoM Features SOPHGO BM1688 AI SOC

Banana Pi BPI-SM9 16-ENC-A3 Development Kit

The Banana Pi BPI-SM9 16-ENC-A3 is an AI-focused compute module built around the SOPHGO BM1688 processor, designed for low-power edge inference, multi-channel video analytics, and embedded AI systems.

The module integrates an 8-core ARM Cortex-A53 CPU, 8GB LPDDR4, and 32GB eMMC, with support for INT4/INT8/FP16/BF16/FP32 mixed-precision workloads and up to sixteen 1080p video streams for real-time analysis. It features extensive multimedia blocks, including HDR-capable ISP, hardware stitching, dewarp, and binocular depth processing, along with additional features such as dual GbE, SATA, USB3.0, HDMI 2.0, PCIe Gen3, dual MIPI-CSI, M.2 Key M/E, CAN, UART, SPI, and I2C. Applications include smart servers, industrial PCs, edge AI boxes, drones, surveillance systems, and AIoT deployments.

BPI-SM9 SoM

Banana Pi BPI-SM9 16-ENC-A3 Specifications:

• Processor:
  • SOPHGO BM1688 deep-learning processor, 8× Arm Cortex-A53 @ 1.6 GHz
  • Mixed-precision support: INT4 / INT8 / FP16 / BF16 / FP32
  • Up to 16-channel 1080p real-time video analysis
    • Decode: H.264 / H.265, up to 16× 1080p@30fps, max 8192×8192
    • Encode: H.264 / H.265, up to 10× 1080p@30fps, up to 8K
  • JPEG accelerator: 1080p@480fps, max 32768×32768
  • ISP: dual 8MP@30fps, HDR, 3DNR, LDC, 3A, dehaze
  • AI: IVE / VPSS, DPU for binocular depth, Hardware image stitching, and fisheye dewarp
• System memory/storage:
  • 8GB LPDDR4x, 64-bit, up to 4266 Mbps
  • 32GB eMMC 5.1
  • 1× microSD card interface
• Interfaces (via 260-pin SO-DIMM edge connector):
  • 2× GbE
  • 2× SATA 3.0
  • PCIe Gen3 (1×2 + additional lanes), plus PCIe 2.0 via Type-C
  • 2× MIPI-CSI camera interfaces
  • 1× HDMI 2.0 display output
  • USB: 2× USB 3.0, 1× USB 2.0 (plus internal USB 2.0 lines)
  • 1× CAN, 1× SD/SDIO
  • 4× UART, 4× I²C, 2× SPI, 6× PWM, ADC, GPIOs
• Power: Nominal 12V DC input (via carrier)
• Dimensions: 100 × 79 × 16 mm
• Operating Temp: -10°C to 60°C

The BPI-SM9 runs Ubuntu 22.04 with Linux 5.10 and integrates SOPHGO’s toolchain supporting ONNX, Caffe, and TFLite models, with compatibility layers for TensorFlow, PyTorch, Paddle, and TensorRT (BM1684/BM1684X). Developers can use the sophon-demo and sophon-stream frameworks for accelerated inference, multi-stream video pipelines, and hardware-assisted ISP/DPU workloads.

BPI-SM9 SoM Hardware Overview

The Banana Pi BPI-SM9 compute module is currently listed at USD $152.00 on the official Banana Pi store, while third-party marketplaces such as AliExpress store price the 8GB LPDDR4 + 32GB eMMC version at around US $52.50 (or US $42 after discounts). Pricing can vary significantly between platforms, so buyers should compare listings across the official shop and AliExpress, and other stores to find the best available offer.

More information about the product is available on the product’s page.

Images used courtesy of Banana Pi