Specifications
Manufacturer
NVIDIA
Platform
IGX Thor (industrial-grade)
AI Performance (iGPU)
Up to 2,070 FP4 TFLOPS
AI Performance (dGPU)
Up to 5,581 FP4 TFLOPS (T7000)
iGPU
2,560-core NVIDIA Blackwell
5th-gen Tensor Cores · MIG
CPU
14-core Arm Neoverse-V3AE
@ 2.6 GHz
Memory
128GB 256-bit LPDDR5X
@ 273 GB/s (full ECC)
dGPU (T7000)
RTX PRO 6000 Blackwell Max-Q
RTX PRO 5000 Blackwell
Networking
4× up to 25 Gbps MGBE (T5000)
Dual 200GbE ConnectX-7 (T7000)
Storage
1TB M.2 NVMe PCIe Gen5
UFS
Functional Safety
Functional Safety Island (FSI)
+ Safety MCU
BMC
Yes (T7000 & Dev Kit)
Real-Time
Preemptible real-time Linux kernel
MIG
Multi-Instance GPU isolation
Software
IGX-SW 2.0 (GA, Mar 2026)
Holoscan · Isaac
Availability
GA — dev kits shipping
Overview
NVIDIA IGX Thor is the enterprise-ready sibling of Jetson Thor — an industrial-grade edge AI platform that pairs the same Blackwell compute with the hardware, safety, and manageability features regulated industries demand. It is purpose-built for physical AI in industrial automation, medical devices, and robotics, where functional safety, deterministic real-time behavior, and extended lifecycle support are non-negotiable.
The platform spans four members: the IGX T5000 System-on-Module (compact, safety-capable compute for custom carrier boards), the IGX T7000 Board Kit (microATX form factor that scales with a discrete RTX PRO 6000 Blackwell Max-Q GPU and enterprise networking), the full-featured IGX Thor Developer Kit, and the Developer Kit Mini (smaller footprint with an on-board safety module for mobile robots and compact industrial systems).
Versus the previous-generation IGX Orin, IGX Thor delivers 8× higher AI compute on the integrated GPU, 2.5× higher with discrete-GPU acceleration, and 2× the networking bandwidth via dual 200GbE ConnectX-7. Critically, IGX T5000 matches Jetson T5000 performance with full DRAM ECC enabled — industrial safety features impose no performance or memory penalty.
Key Benefits
<strong>Enterprise-Grade Reliability:</strong> Extended temperature range, shock/vibration compliance, and ECC throughout make IGX Thor suitable for mission-critical 24/7 deployments. <strong>Functional Safety:</strong> A dedicated Functional Safety Island (FSI) plus safety MCU enables certified safety applications in robotics and medical. <strong>Dual 200GbE RDMA:</strong> ConnectX-7 with GPU Direct RDMA lets sensor data bypass the CPU into GPU memory — the foundation of NVIDIA Holoscan Sensor Bridge for cameras, lidar, radar, and medical imaging. <strong>Multi-Instance GPU:</strong> Blackwell MIG partitions the iGPU into fully isolated instances with hard performance guarantees for mixed-criticality workloads. <strong>Real-Time Linux:</strong> A preemptible real-time kernel ships by default for deterministic control loops and sensor handling.
Request a Quote — NVIDIA IGX Thor
QS Compute — global B2B supply for NVIDIA edge AI and data center hardware. Enterprise procurement, direct manufacturer channels.
Get Your Quote →