Firefly unveils Core-3568J AI Core Rockchip RK3568 system-on-module & devkit

Firefly has unveiled a new 314-pin SO-DIMM system-on-module (SoM) with Core-3568J AI Core powered by Rockchip RK3568 processor and equipped with up to 8GB RAM, 128GB eMMC flash with interfaces including PCIe, SATA, dual Gigabit Ethernet, HDMI 2.0, eDP 1.3, MIPI DSI and more.

The company also provides AIO-3568J development kit for the module with all those interfaces plus an M.2 socket for 5G or WiFi 6 connectivity, and PoE support among other features.

Core-3568J AI Core system-on-module

RK3568 system-on-module


  • SoC – Rockchip RK3568 quad-core Cortex-A55 processor @ 2.0 GHz with Arm Mali-G52 2EE GPU with support for OpenGL ES 1.1/2.0/3.2, OpenCL 2.0, Vulkan 1.1, 0.8 TOPS NPU for AI acceleration, 4Kp60 H.265/H.264/VP9 video decoding, and 1080p100 H.265/H.264 video encoding
  • System Memory – 2GB, 4GB, or 8GB LPDDR4 ECC RAM up to 1600 MHz
  • Storage – 32GB, 64GB, or 128GB eMMC flash
  • 314-pin SO-DIMM edge connector with:
    • Storage – SATA 3.0 interface, SD card interface
    • Display I/F
      • HDMI 2.0 up to 4Kp60
      • 2x MIPI DSI up to 1080p60 or 2560×1440 @ 60 Hz whem combined
      • eDP 1.3 up to 2560×1600 @ 60 Hz
    • Camera I/F –  2x MIPI-CSI camera interfaces
    • Audio – Speaker and earphone output, microphone, HDMI audio output, I2S
    • Networking – 2x Gigabit Ethernet
    • USB – USB 3.0, USB 2.0
    • PCIe – 1x PCIe 3.0
    • Low-speed I/Os – SPI, UART, I2C, SDIO, PWM, ADC, GPIO
  • Supply Voltage – TBD
  • Dimensions – 82.0 x 53.5mm
  • Temperature Range – Operating: -10°C to 60°C; storage: -20°C to 70°C
  • Storage Humidity – 10%~80 %

The company will provide Android 11.0 and Ubuntu 18.04 for the module, as well as software (RKNN) for Caffe/TensorFlow/TFLite/ONNX/PyTorch/Keras/Darknet AI frameworks and models. The RK3568 system-on-module is expected to be found in smart NVR, cloud terminals, IoT gateways, and industrial control applications.

AIO-3568J RK3568 development kit

RK3568 devkit

For evaluation and development, Firefly also provides a devkit with the following connectors/features:

  • Storage
    • 1x M.2 PCIe 3.0 socket for 2242 / 2280 NVMe SSD
    • 1x SATA 3.0 for 2.5” SSD/HDD
    • MicroSD card socket
  • Video Output – HDMI 2.0 port, eDP 1.3 connector, Touch Panel header
  • Camera I/F – 2x MIPI CSI connectors
  • Audio – 3.5mm headphone jack, 3.5mm microphone jack, speaker header
  • Networking
    • Dual Gigabit Ethernet RJ45 ports including one with POE+ (802.3 AT, output power 30W) support
    • M.2 socket for 5G
    • Mini PCIe socket for 4G LTE
    • Optional WiFi 6 (802.11ax)/Bluetooth 5.0
  • USB – 2x USB 3.0 ports including one OTG port,  2x USB 2.0 ports, 2x USB 2.0 interfaces on header
  • Serial – CAN FD terminal block, 2x RS232 terminal blocks, 2x RS485 terminal blocks, 2x UART headers
  • Debugging – UART header
  • Misc – Power, reset and recovery keys, IR header, 3x ADC header, fan header
  • Power Supply
    • 6V to 24V input via DC jack (recommended 12V input)
    • DC output header with 12V, 5V, 3.3V and GND
  • Dimensions – TBD

Documentation is non-existent at this point in time, but eventually, a Wiki, firmware images, and datasheets should be released on the product page like for the company’s other modules and development boards. The price for the module and development kit is still to be determined.

Thanks to Gareth for the tip.

Share this:

Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress

ROCK 5 ITX RK3588 mini-ITX motherboard
Notify of
The comment form collects your name, email and content to allow us keep track of the comments placed on the website. Please read and accept our website Terms and Privacy Policy to post a comment.
3 years ago

This ticks a lot of boxes for me.
Looking forward to seeing it released!

3 years ago

Very likely the richest connectivity we’ve seen in a while here, and by a large margin! I find it smart to finally design so rich a module, because it offers short lead time to small-series manufacturers without the usual trouble related to having to install muxes, pcie-to-usb controllers or whatever. Just route the pins to a connector or a phy and you’re done.

3 years ago

Is the AI for BSP kernel only?

2 years ago
Khadas VIM4 SBC