Home MYIR | Single Board Computers FZ3 Card – Deep Learning Accelerator Card
MYIR | Single Board Computers

FZ3 Card – Deep Learning Accelerator Card

The FZ3 Card is a deep learning accelerator card powered by the Xilinx Zynq UltraScale+ ZU3EG MPSoC, featuring a 1.2 GHz Quad ARM Cortex-A53 and 600 MHz Dual Cortex-R5 cores, and delivering up to 1.2TOPS. It is equipped with 4GB DDR4 RAM, 8GB eMMC, and 32MB QSPI Flash, providing robust AI performance for edge computing applications..

  • Detail
  • Special Features
  • Specification
  • Processor: Xilinx Zynq UltraScale+ ZU3EG MPSoC, 1.2 GHz Quad ARM Cortex-A53, 600 MHz Dual ARM Cortex-R5, ARM Mali-400 MP2 GPU, 16nm FPGA fabric.

  • Memory: 4GB DDR4 (64-bit), 8GB eMMC, 32MB QSPI Flash, 32KB EEPROM.

  • Storage: 1 Micro SD Card slot, 1 PCIe 2.1 Interface (1-lane).

  • Connectivity: 1 USB 2.0 Host, 1 USB 3.0 Host, 1 Gigabit Ethernet, 1 MIPI-CSI Interface, 1 BT1120 Camera Interface, 1 Mini DP (4K/30fps), USB-UART, JTAG, RS232, RS485, CAN, 2 Expansion Interfaces (USB, UART, CAN, I2C, SPI).

  • Power Supply: DC 12V/2A, static power consumption: 5W.

  • Operating Temp: -40°C ~ +85°C.

  • Software: Supports PetaLinux, Xilinx Vitis Software Development Platform, PaddlePaddle AI framework.

  • High AI Performance: Provides up to 1.2TOPS computing power and runs MobileNet at 100 FPS.

  • Comprehensive Connectivity: Includes USB 2.0, USB 3.0, Gigabit Ethernet, TF, DP, PCIe, MIPI-CSI, BT1120, USB-UART, and JTAG.

  • Advanced Video Decoding: Supports 8-16 channel video decoding and 4-8 channel intelligent analysis.

  • Software Support: Ready to run with PetaLinux 2020.1 and compatible with Xilinx Vitis and Baidu’s PaddlePaddle AI framework.

  • Versatile Applications: Ideal for AI cameras, robotics, UAVs, and industrial AI applications.

  • Dimensions: 100mm x 70mm (12-layer PCB design).

  • Buttons: 1 FPGA Reset, 1 System Reset.

  • LEDs: Power LED (Red), 2 Status LEDs (Red), 2 Status LEDs (Green).

  • RTC: 3V Rechargeable RTC Battery Interface (battery not included).