Firefly EC-AGXOrin AI System Supports 8 GMSL2 Cameras and 275 TOPS

Firefly EC-AGXOrin AI System. Image used courtesy of Firefly

Firefly has recently announced EC-AGXOrin, an NVIDIA Jetson AGX Orin-based AI System with eight GMSL2 cameras, delivering up to 275 TOPS AI power for high-performance robotics, machine vision, autonomous machines, and in-vehicle AI computing.

The system integrates a 12-core Cortex-A78AE CPU, Ampere GPU with 2048 CUDA and 64 Tensor Cores, 2× NVDLA accelerators, and 64GB LPDDR5 RAM, plus 64GB eMMC with M.2 NVMe and MicroSD expansion. Connectivity options include 10GbE LAN, five GbE LAN, USB 3.0, HDMI 2.0, RS232/RS485, CAN, UART, and DIO. There is also support for up to eight GMSL2 camera inputs through FAKRA connectors. Wireless options include WiFi 6/Bluetooth, 4G, 5G, and GNSS via M.2 and Mini PCIe modules. With a rugged enclosure, wide operating temperature range (−20°C to 60°C), and standard power input, it is suitable for edge AI, robotics, industrial automation, and multi-camera analytics.

Firefly EC-AGXOrin hardware overview front side. Image used courtesy of Firefly

Firefly EC-AGXOrin hardware overview back side. Image used courtesy of Firefly

Firefly EC-AGXOrin Specifications:

• SoM: NVIDIA Jetson AGX Orin (64GB)
• CPU: 12-core Arm Cortex-A78AE (v8.2) with 3MB L2 + 6MB L3
• GPU / AI:
  • NVIDIA Ampere GPU with 2048 CUDA cores, 64 Tensor Cores @ 1.3GHz
  • 2x NVDLA v2.0 accelerators
  • PVA v2.0 vision accelerator
  • AI performance: up to 275 TOPS (INT8 @ 60W)
• Memory: 64GB 256-bit LPDDR5 @ 204.8GB/s
• Storage:
  • 64GB eMMC 5.1
  • M.2 2280 M-Key for NVMe SSD
  • MicroSD card slot
• Video:
  • Encode: up to 2× 4K60 / 8× 1080p60 (H.265)
  • Decode: up to 8K30 / 22× 1080p30 (H.265)
  • Output: HDMI 2.0 up to 4K60
• Camera: 8× GMSL2 via 2× Mini-FAKRA connectors
• Networking:
  • 1× 10GbE RJ45 LAN ports
  • 5× GbE RJ45 LAN ports
  • Optional WiFi 6/Bluetooth via M.2 E-Key
  • Optional 4G LTE (Mini PCIe) / 5G (M.2 B-Key)
  • Optional GNSS
• USB:
  • 4× USB 3.0 Type-A
  • 2× USB Type-C (debug + OTG)
• Audio: 3.5mm jack with MIC support
• Serial & I/O:
  • RS-232, RS-485
  • 2× CAN 2.0
  • UART, GPIO, 3.3V/5V pins
  • 8× DIO via DB-9
• Expansion:
  • M.2 2230 E-Key (WiFi/BT)
  • M.2 B-Key (4G/5G)
  • M.2 2280 M-Key (NVMe SSD)
  • Mini PCIe slot
• Power: 24V DC input (5.5×2.1mm jack)
• Enclosure / Environment:
  • Dimensions: 277.95 × 136.09 × 88.0 mm
  • Operating temperature: −20°C to 60°C
• Humidity: 10%~90% non-condensing

Firefly says the system runs Ubuntu 22.04 with full desktop GPU acceleration and supports CUDA, TensorRT, cuDNN, ROS robotics models, and vision models like EfficientViT, NanoSAM, and TAM. It can also run image generation models such as Flux and Stable Diffusion XL for on-device AI and robotics workloads. On top of that, it supports Ollama, ComfyUI, and frameworks like PyTorch, TensorFlow, MATLAB, and PaddlePaddle, with Docker for easy deployment. Documentation and downloads are still under development.

Firefly has yet to provide pricing information for the EC-AGXOrin AI inference system. More information can be found on the product page.

Firefly EC-AGXOrin hardware overview back side.

Images used courtesy of Firefly