Jetson Thor Robot Computer
Jetson Thor
Descripción del producto
MODEL: EMB-THOREDGE-DX1
Compact. Rugged. Robotics-Grade. The EMB-THOREDGE-DX1 is a deployable heavy-duty edge AI computer engineered for next-generation robotics and autonomous systems. Powered by the NVIDIA Jetson Thor platform (T5000 or T4000), this 2.67-litre chassis delivers up to 2,070 FP4 TFLOPS of AI performance with a single-panel IO architecture designed for simplified cabling, rack installation, and vehicular deployment. With rich industrial-grade interfaces including GMSL2, CAN FD, PoE+, and QSFP28, it is purpose-built for sensor fusion, multi-camera perception, and real-time robotic control.
Up to 2,070 FP4 TFLOPS AI Performance: The NVIDIA Jetson Thor T5000 delivers 2,070 TOPS in FP4 precision for massive generative AI and multimodal robot models. The compact 2.67L design concentrates all interfaces on a single panel, with industrial-grade GMSL2 camera support for up to 8 cameras and 4x CAN FD for robotic actuator control. Supply longevity through January 2036.
FEATURE HIGHLIGHTS
● NVIDIA Jetson T5000 or Jetson T4000 System-on-Module options
● Up to 2,070 FP4 TFLOPS AI performance (T5000), 1,200 TOPS (T4000)
● 128GB LPDDR5X memory (T5000) / 64GB LPDDR5X (T4000)
● 1TB NVMe Gen4 SSD storage
● 8 x GMSL2 camera inputs via 2x Quad Fakra connectors, 12V bias voltage @ 0.6A ~ 1.1A
● 2 x HDMI 2.1 display outputs with thumbscrew lock
● 4 x USB 3.1 Type-A (5V @ 1.5W) with thumbscrew holes
● 1 x USB 3.1 Type-C (5V @ 15.0W) with thumbscrew hole
● 1 x USB 3.1 Type-C OTG (5V @ 15.0W)
● 1 x USB-C UART (USB 2.0, 5V, 900mA) for debug console
● 2 x 5Gbps PoE+ (30W) RJ45 Ethernet ports
● 4 x 25 Gbps QSFP28 (3x for T4000 variant)
● 4 x CAN FD with opto-isolation 3KV (T5000 only)
● 2 x RS-485 / RS-422 / RS-232 with 15KV ESD protection
● 4 x GPIO (5V/1A, 3V/1A, GND)
● Wi-Fi 6 (802.11ax) and Bluetooth 5 wireless connectivity
● 4G / 5G / LTE cellular support with nano SIM socket
● 7 antenna connectors: 2x Wi-Fi, 4x Cellular, 1x GPS
● TPM 2.0 security module (SLB 9672XU2.0)
● Wide operating temperature: -20°C to +85°C (main unit)
● DC 9V ~ 36V power input, barrel jack with screw lock
● CR2450 RTC battery, 10-year lifetime
● 3 x 3-pin fan headers for active cooling configuration
● Compact-first design: 185 x 170 x 85 mm, 2.67 Litres
● Single-panel IO architecture for simplified cabling and rack installation
● CE / FCC EMC certified, IEC 62368-1 safety compliant
Especificaciones del producto
APPLICATIONS
Humanoid & General Robotics
Powered by NVIDIA Isaac ROS and Isaac GR00T models for humanoid tasks. The Jetson Thor's massive AI compute (up to 2,070 FP4 TFLOPS) enables real-time multimodal perception, motion planning, and human-robot interaction for next-generation humanoid and service robots.
Autonomous Vehicles & Mobile Robots
8x GMSL2 camera inputs, 4x CAN FD for actuator control, and 2x PoE+ for LiDAR/thermal sensors provide complete sensor fusion for autonomous navigation. DC 9V~36V input and -20°C to +85°C range suit vehicular deployment.
Multi-Camera AI Perception & Safety
Process up to 8 GMSL2 cameras simultaneously for 360-degree visual perception, object detection, and safety monitoring. NVIDIA Metropolis enables visual agentic AI for video search, summarization, and anomaly detection at the edge.
Sensor Fusion & Real-Time Processing
NVIDIA Holoscan processes high-speed data from multiple sensors (LiDAR, cameras, IMU, thermal) with ultra-low latency. QSFP28 (4 x 25 Gbps) provides high-bandwidth interconnect for distributed sensor arrays and multi-unit clusters.
Industrial Automation & Smart Manufacturing
Rich industrial interfaces (CAN FD, RS-485/422/232, GPIO) enable direct integration with PLCs, motor controllers, and factory equipment. Edge AI inference for quality inspection, predictive maintenance, and process optimization.
Robotics Simulation & Digital Twin
NVIDIA Isaac Sim enables virtual robot development and testing before physical deployment. Combined with Nsight Systems 2025.5 for performance profiling, teams can iterate rapidly on AI models and control algorithms in simulation.
| System-on-Module Options | ThorEdge-DX1-T5: NVIDIA Jetson T5000 ThorEdge-DX1-T4: NVIDIA Jetson T4000 |
| System Memory | T5000: 128 GB LPDDR5X T4000: 64 GB LPDDR5X |
| Mass Storage | 1 TB NVMe Gen4 SSD M.2 M-Key 2280 Gen5x4 slot |
| AI Performance | T5000: 2,070 TOPS (FP4) T4000: 1,200 TOPS T5000 thermal mode: 40W ~ 130W T4000 thermal mode: 40W ~ 70W |
| Vision Interface | 8 x GMSL2 camera inputs (2x Quad Fakra connectors) Bias voltage: 12V, @0.6A ~ 1.1A 2 x HDMI 2.1 outputs with thumbscrew lock |
| USB Ports | 4 x USB 3.1 Type-A (5V @ 1.5W, thumbscrew holes) 1 x USB 3.1 Type-C (5V @ 15.0W, thumbscrew hole) 1 x USB 3.1 Type-C OTG (5V @ 15.0W) 1 x USB-C UART (USB 2.0, 5V, 900mA, debug console only) |
| Network Interface | 2 x RJ45 5 Gbps PoE+ (30W) 4 x QSFP28 25 Gbps (3x for T4000 variant) |
| Wireless | Wi-Fi 6 IEEE 802.11ax Bluetooth 5 |
| Cellular | 4G / 5G / LTE support |
| Antenna | 2 x Wi-Fi, 4 x Cellular, 1 x GPS (7 total) |
| Industrial Interface | 4 x CAN FD with opto-isolation 3KV (T5000 only) 2 x RS-485 / RS-422 / RS-232 with 15KV ESD protection 4 x GPIO (5V/1A, 3V/1A, GND) |
| Storage & Expansion | M.2 M-Key 2280 Gen5x4: NVMe SSD M.2 E-Key 2230: Wi-Fi 6, BT 5 (PCIe, USB2) M.2 B-Key 3042/3052: 4G/5G/LTE (PCIe, USB2) M.2 B-Key 2242: GSM module (USB2) Nano SIM socket (PCB bottom side) TPM 2.0 (SLB 9672XU2.0 FW16.xx) 3 x 3-pin fan headers |
| Cooling System | Active cooling with 3 x 3-pin fan headers Rugged industrial-grade chassis |
| Dimensions | 185 (W) x 170 (D) x 85 (H) mm 2.67 Litres |
| Operating Temperature | -20°C to +85°C (main unit) -25°C to +60°C (t.b.d.) |
| Cooling System | Fanless, industrial-grade metal chassis |
| Humidity | 5% - 95% |
| Power Input | DC 9V ~ 36V, barrel jack with screw lock Power consumption: 180W peak (t.b.c.) RTC backup: CR2450, 10-year lifetime |
| Certifications | EMC: CE, FCC Safety: IEC 62368-1 |
| Supported OS | Jetson Linux (L4T) — customized Ubuntu 24.04 LTS kernel 6.8 |
| AI Frameworks | NVIDIA CUDA 13.0 NVIDIA cuDNN 9.12 NVIDIA TensorRT 10.13 |
| Robotics Software | NVIDIA Isaac ROS — hardware-accelerated perception and navigation NVIDIA Isaac GR00T — humanoid task models NVIDIA Holoscan — multi-sensor low-latency processing NVIDIA Metropolis — visual agentic AI |
| Development Tools | VPI 4.0 (Vision Programming Interface) — PVA v3 offload NVIDIA Isaac Sim — virtual robot simulation Nsight Systems 2025.5 — system-wide performance profiling |
| Supply Longevity | T5000: Available through January 2036 T4000: Available through January 2035 |
ORDERING PART NUMBER (SKU)
| SKU | SoM | AI Performance | Memory | Storage | Thermal Mode |
|---|---|---|---|---|---|
| EMB-THOREDGE-DX1-T5 | NVIDIA Jetson T5000 | 2,070 TOPS (FP4) | 128GB LPDDR5X | 1 TB NVMe Gen4 SSD | 40W ~ 130W |
| EMB-THOREDGE-DX1-T4 | NVIDIA Jetson T4000 | 1,200 TOPS (FP4) | 64GB LPDDR5X | 1 TB NVMe Gen4 SSD | 40W ~ 70W |
Specifications are preliminary and subject to change. Please contact us for full part number with specific storage and accessory configuration requirements.
Descargas
Product Images 