NVIDIA IGX Thor Developer Kit - Enterprise Edge AI Platform
The NVIDIA IGX Thor Developer Kit is an enterprise-ready development platform designed to accelerate industrial and medical edge innovation with real-time sensor processing, AI reasoning, and functional safety. Powered by the cutting-edge NVIDIA Blackwell GPU architecture and delivering up to 5,581 FP4 TFLOPS of AI compute performance, this platform enables developers to effortlessly run multiple generative AI models at the edge for physical AI agents, autonomous robots, and mission-critical applications.
Unmatched AI Performance for Edge Computing
Built on the revolutionary NVIDIA Blackwell GPU architecture, the IGX Thor Developer Kit delivers server-class AI performance in an edge-optimised platform. The kit includes the NVIDIA IGX T7000 board with a 2,560-core Blackwell iGPU, 14-core Arm Neoverse V3AE CPU, and a pre-installed NVIDIA RTX PRO 6000 Blackwell dGPU. Compared to the previous-generation IGX Orin Developer Kit, it provides up to 8× higher AI compute on iGPU, 2.5× higher AI compute on dGPU, and 2× better connectivity, making it ideal for next-generation robotics, visual AI agents, and real-time medical imaging systems.
Key Features
-
Blackwell GPU Architecture: 2,560-core iGPU with fifth-generation Tensor Cores and Multi-Instance GPU (MIG) technology for robust workload isolation
-
Up to 5,581 FP4 TFLOPS: Powerful AI compute for running multiple generative AI models including Vision Language Action (VLA) models and large language models
-
14-core Arm Neoverse V3AE CPU: High-performance 64-bit processor optimised for low-latency, real-time applications
-
128 GB LPDDR5x Memory: 256-bit memory interface with 273 GB/s bandwidth for demanding AI workloads
-
Functional Safety Island (FSI): Integrated safety features with onboard Safety MCU for mission-critical industrial and medical applications
-
400 GbE Connectivity: NVIDIA ConnectX-7 SmartNIC with QSFP112 for ultra-high-speed networking and sensor fusion
-
10-Year Lifecycle Support: Enterprise-grade hardware with extended support for long-term deployments
-
NVIDIA AI Enterprise Software: Runs NVIDIA Isaac for robotics, Metropolis for visual AI, Holoscan for sensor processing, and Halos safety AI agents
Advanced Hardware Accelerators
The IGX Thor Developer Kit includes a comprehensive suite of dedicated accelerators designed to handle complex sensor processing and AI inference tasks. These include a third-generation Programmable Vision Accelerator (PVA), dual video encoders and decoders, an optical flow accelerator, and more. The platform leverages Transformer Engine and Multi-Instance GPU (MIG) support in the Blackwell architecture to deliver the performance required to run multiple AI models simultaneously with robust workload isolation.
Enterprise-Ready Functional Safety
Purpose-built for industrial and medical environments, the IGX Thor platform seamlessly integrates AI and functional safety into a single system. With the Functional Safety Island built into the Thor SoC and an onboard safety microcontroller, the developer kit enables advanced safety functions through the IGX software stack. This makes it ideal for safety-critical applications including collaborative robots, autonomous machines, and regulated medical devices requiring ISO 26262 and IEC 61508 compliance.
Extensive Connectivity for Sensor Fusion
Designed for high-speed sensor fusion and multi-sensor robotics applications, the developer kit offers extensive I/O options:
-
2 × 200 GbE QSFP ports for ultra-high-speed networking
- 2 × 1 GbE RJ45 Ethernet ports
- 4 × USB-A 3.0 ports
-
USB-C connectivity
-
DisplayPort 1.4a output
-
Wi-Fi and Bluetooth (Realtek module)
-
1 TB NVMe SSD for high-speed storage
Cloud-to-Edge AI Development
The IGX Thor Developer Kit delivers a seamless cloud-to-edge experience, running the NVIDIA AI Enterprise software stack and NVIDIA NIM for physical AI applications. Developers can leverage industry-leading frameworks including NVIDIA Isaac for robotics development, NVIDIA Metropolis for visual AI agents, NVIDIA Holoscan for real-time sensor processing, and the NVIDIA Halos Outside-in Safety Agent Blueprint for industrial safety AI applications.
Applications
- Autonomous mobile robots and humanoid robotics
- Collaborative industrial robots with safety AI
- Real-time medical imaging and surgical robotics
- Visual AI agents for smart manufacturing
- Edge AI for smart hospitals and healthcare
- Mission-critical industrial automation
What's in the Box
- NVIDIA IGX T7000 Board Kit (with 2,560-core Blackwell iGPU, 14-core Arm Neoverse V3AE CPU, BMC, and ConnectX-7 SmartNIC)
- NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition dGPU (pre-installed)
- Safety MCU
- 1 TB NVMe SSD
- Realtek Wi-Fi/BT Module
- Power Supply (uATX)
- Quick Start Guide
Full Specifications
Compute
| iGPU |
2,560-core NVIDIA Blackwell architecture GPU with fifth-generation Tensor Cores, Multi-Instance GPU with 10 TPCs, 1.57 GHz |
| dGPU |
NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition (included) |
| CPU |
14-core Arm Neoverse V3AE 64-bit CPU |
| AI Performance |
Up to 5,581 FP4 TFLOPS |
| Memory |
128 GB 256-bit LPDDR5x, 273 GB/s bandwidth |
Accelerators
| Vision Processing |
Third-generation Programmable Vision Accelerator (PVA) |
| Video |
Dual encoders and decoders |
| Optical Flow |
Dedicated optical flow accelerator |
Storage & Connectivity
| Storage |
1 TB NVMe SSD |
| Network |
NVIDIA ConnectX-7 SmartNIC with QSFP112 400 GbE (2 × 200 GbE ports), 2 × 1 GbE RJ45 |
| USB |
4 × USB-A 3.0, 1 × USB-C |
| Display |
DisplayPort 1.4a |
| Wireless |
Wi-Fi 802.11ax, Bluetooth (Realtek module) |
Safety & Management
| Functional Safety |
Functional Safety Island (FSI) on Thor SoC |
| Safety MCU |
Onboard safety microcontroller |
| Management |
NVIDIA BMC Module (Board Management Controller) |
Power & Environment
| TMP (Thermal Management Power) |
Up to 130 W (T7000 module) |
| Power Supply |
uATX power supply (included) |
| Lifecycle Support |
10 years |
Software
| Software Stack |
NVIDIA AI Enterprise, NVIDIA NIM |
| Frameworks |
NVIDIA Isaac (robotics), NVIDIA Metropolis (visual AI), NVIDIA Holoscan (sensor processing), NVIDIA Halos (safety AI) |
| AI Models |
Supports Vision Language Action (VLA) models, LLMs, VLMs including NVIDIA Isaac GR00T N and NVIDIA Cosmos Reason |
Specifications are sourced from the manufacturer product page and may be updated.