e-con Systems' NVIDIA Jetson Thor-based AI Compute platform

e-con Systems' ECU accelerated by NVIDIA Jetson Thor is a compact, production-ready compute platform built for Humanoids, physical AI, industrial automation and robotics applications. Delivering up to 2070 FP4 TFLOPS of compute with 128 GB high-bandwidth memory, it combines rugged reliability with real-time multi-sensor processing. This platform enables seamless integration of multiple sensors—while leveraging e-con's advanced Holoscan camera solutions and proprietary TintE ISP® for superior image quality and optimized processing.

e-con's robust ECU solutions offer multiple interface support, high-performance AI compute, multi-sensor connectivity, and advanced imaging capabilities, achieving fast perception, decision-making, and control in dynamic, edge environments.

WHY CHOOSE E-CON'S VISION BOX?

01

Powered by NVIDIA Jetson
Thor

Unleash extraordinary AI throughput (e.g., up to 2070 TFLOPS FP4 Sparse) for complex perception, LLMs, and multimodal sensor fusion.

02

Ultra-High Bandwidth
Memory

Massive LPDDR5X (e.g., 128 GB, 256-bit) to support dense neural nets and real-time sensor data.

03

Camera Options

Holoscan-based camera integration for advanced vision AI and 8x GMSL2 camera support for high-performance robotic vision.

04

Robust Industrial Design

Rugged metal build with wide thermal tolerance
(-25 °C to +85 °C) ensures reliability in challenging field conditions.

05

Comprehensive Connectivity

Feature-rich I/O including Gigabit Ethernet, QSFP28 (4x25 GbE), USB3.2 Type-C, CAN, digital I/O, and real-time sensors (IMU, temperature, EEPROM, RTC).

06

Qualified for Physical AI

Built to handle heavy AI workloads, multimodal sensor fusion, real-time inference, and robotics control in milliseconds.

Specifications

Feature Category Specification Details
Processor Module NVIDIA Jetson Thor (e.g., Thor T6000)
Memory 128 GB LPDDR5X (256-bit, ~273 GB/s)
AI Performance Up to 2070 TFLOPS (FP4 Sparse)
Graphics Outputs 2× HDMI 2.1, 2× DP 1.4a
I/O Interfaces Gigabit Ethernet, QSFP28 (4×25 GbE), 2× USB3.2-C (with DP), Debug USB-C, CAN, RS-232/422/485, 2× Digital In, 3× Digital Out, EEPROM, IMU, Temp Sensor, RTC
Camera Support Up to 8× GMSL2 camera inputs, 10G Holoscan support
Storage M.2 Key-M 2280 (PCIe Gen5 x4), M.2 Key-M 2242 (PCIe Gen5 x2)
Power Input 18–32 VDC
Wireless Options WiFi / Bluetooth / LTE
Operating Range –25 °C to +85 °C
Operating System Ubuntu Linux 24.04, NVIDIA JetPack 7.0
Dimensions TBD
Autonomous Guided Vehicles (AGVs)
Pick & Place Robots
Industrial Automation
Industrial Automation
  • Supported AI frameworks:
    • Pytorch
    • ONNX
    • TensorFlow
    • YOLO
SMARTOdinUltra Product Specifications

Download Full HD USB Camera Documents

Target Applications

Humanoids & Autonomous Mobile Robots
Humanoids & Autonomous Mobile Robots
Smart City & Infrastructure
Smart City & Infrastructure
Industrial Automation
Industrial Automation