NVIDIA IGX T7000 Industrial Edge AI Board Kit with Blackwell GPU
The NVIDIA IGX T7000 Board Kit is an enterprise-grade edge AI computing platform designed for industrial, medical, and robotics applications requiring high-performance AI inference, real-time sensor processing, and functional safety. Built on the revolutionary NVIDIA Blackwell GPU architecture, it delivers up to 5,581 FP4 TFLOPS of AI compute when paired with an optional NVIDIA RTX PRO 6000 Blackwell discrete GPU, making it ideal for autonomous machines, collaborative robots, medical imaging systems, and sensor-intensive workloads.
Unmatched AI Performance at the Edge
At the heart of the IGX T7000 lies a 2,560-core NVIDIA Blackwell architecture iGPU delivering 2,070 FP4 TOPS of AI performance, paired with a powerful 14-core Arm Neoverse-V3AE 64-bit CPU. The platform can be expanded with an NVIDIA RTX PRO 6000 Blackwell Max-Q discrete GPU adding 3,511 FP4 TOPS, achieving 8x higher AI compute on iGPU and 2.5x higher AI compute on dGPU compared to NVIDIA IGX Orin. This makes the T7000 capable of running multiple generative AI models simultaneously at the edge with robust workload isolation using Multi-Instance GPU (MIG) technology.
Key Features
-
MicroATX form factor with flexible expandability for demanding edge AI workloads
-
Dual 200 GbE networking powered by NVIDIA ConnectX-7 SmartNIC with 400 GbE capability
-
Functional Safety Island (FSI) built into the Thor SoC for true hardware separation between safety and non-safety domains
-
ISO 26262 and IEC 61508 compliance targeting ASIL D/SC3 and ASIL/SIL 2 for safety-critical applications
-
Board Management Controller (BMC) for enterprise-grade remote management and monitoring
-
128 GB 256-bit LPDDR5x memory with 273 GB/s bandwidth for high-throughput data processing
-
10-year product lifecycle with long-term software support through NVIDIA AI Enterprise
High-Speed Connectivity for Sensor-Intensive Workloads
The IGX T7000 advances edge connectivity with dual 200 GbE networking, delivering 2x the bandwidth compared to IGX Orin. The integrated NVIDIA ConnectX-7 SmartNIC supports RDMA for low-latency, high-throughput data movement, enabling direct connection to cameras, LIDAR, radar, and medical imaging equipment without data bottlenecks. The system includes 32 lanes of PCIe Gen5 with an integrated PCIe switch for maximum I/O flexibility.
Enterprise-Grade Software Stack
The T7000 is powered by the IGX enterprise software stack, providing the performance, durability, security, and safety needed to run autonomous physical AI agents at the edge. Support for Transformer Engine accelerates generative AI workloads, while the onboard Safety MCU and Functional Safety Island enable advanced safety functions for mission-critical environments. The platform includes a suite of accelerators such as a third-generation Programmable Vision Accelerator (PVA), dual encoders and decoders, and an optical flow accelerator.
Ideal Applications
-
Autonomous mobile robots (AMRs) and collaborative robots (cobots) with real-time obstacle avoidance
-
Medical imaging systems requiring real-time AI inference and functional safety
-
Industrial automation with multi-sensor fusion and predictive maintenance
-
Smart manufacturing with vision-guided robotics and quality inspection
-
Autonomous vehicles and transportation systems with safety-critical compute
What's in the Box
- NVIDIA IGX T7000 Board Kit (MicroATX form factor)
- NVIDIA IGX T5000 Thor Module with Blackwell iGPU and 14-core CPU
- NVIDIA BMC Module for remote management
- Safety MCU
- NVIDIA ConnectX-7 SmartNIC with 400 GbE networking
- Realtek WiFi/Bluetooth Module
- 1000 GB NVMe SSD
- Optional: NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition dGPU (when purchased as Developer Kit)
Full Specifications
Compute
| GPU (iGPU) |
2,560-core NVIDIA Blackwell architecture with 5th-generation Tensor Cores |
| GPU Performance (iGPU) |
2,070 FP4 TOPS |
| Optional dGPU |
NVIDIA RTX PRO 6000 Blackwell Max-Q (24,064 CUDA cores, 3,511 FP4 TOPS) |
| Total AI Performance |
Up to 5,581 FP4 TFLOPS (with dGPU) |
| CPU |
14-core Arm Neoverse-V3AE 64-bit |
| Multi-Instance GPU |
10 TPCs |
| Accelerators |
3rd-gen Programmable Vision Accelerator (PVA), dual encoders/decoders, optical flow accelerator |
| TDP |
Up to 130 W (module) |
Memory & Storage
| System Memory |
128 GB 256-bit LPDDR5x |
| Memory Bandwidth |
273 GB/s |
| Storage |
1000 GB NVMe SSD |
Networking & Connectivity
| Ethernet |
NVIDIA ConnectX-7 SmartNIC: QSFP112 400 GbE (dual 200 GbE) |
| PCIe |
32 lanes PCIe Gen5 with integrated PCIe switch |
| Wireless |
Realtek WiFi/Bluetooth module |
Functional Safety
| Safety Features |
Functional Safety Island (FSI) in Thor SoC, dedicated Safety MCU |
| Safety Certifications |
ISO 26262, IEC 61508 (targeting ASIL D/SC3 and ASIL/SIL 2) |
Management & Software
| Management |
NVIDIA BMC Module for remote management |
| Software Stack |
NVIDIA AI Enterprise, IGX enterprise software stack, NVIDIA Halos OS support |
| Product Lifecycle |
10 years |
Physical
| Form Factor |
MicroATX |
| Weight |
Contact manufacturer for specifications |
Specifications are sourced from the official manufacturer product page and may be updated.