NVIDIA IGX Thor
The next-generation, enterprise-ready AI platform for physical AI is now here.
NVIDIA IGX Thor
The next-generation, enterprise-ready AI platform for physical AI is now here.
NVIDIA IGX™ is an industrial-grade edge AI platform that combines enterprise-level hardware, software, and support services, specifically designed for industrial applications, robotic systems, and medical environments, offering powerful AI processing capacity, high-bandwidth sensor data processing, enterprise security, and functional safety features.
The platform also comes with NVIDIA AI Enterprise and up to 10 years of support services, allowing you to deliver AI solutions safely and securely to support human-machine collaboration.
NVIDIA IGX™ is an industrial-grade edge AI platform that combines enterprise-level hardware, software, and support services, specifically designed for industrial applications, robotic systems, and medical environments, offering powerful AI processing capacity, high-bandwidth sensor data processing, enterprise security, and functional safety features.
The platform also comes with NVIDIA AI Enterprise and up to 10 years of support services, allowing you to deliver AI solutions safely and securely to support human-machine collaboration.
The fastest way to develop and deploy enterprise-grade physical AI solutions at the edge
The fastest way to develop and deploy enterprise-grade physical AI solutions at the edge
Unmatched Performance
Discover high-performance, energy-efficient systems designed for low-latency, real-time applications.
Functional and AI Safety
Build robot safety AI agents for industrial robots with NVIDIA Halos, combining functional safety and proactive AI safety for autonomous machines.
Enterprise Support
NVIDIA AI Enterprise software includes 10 years of NVIDIA enterprise support, ensuring accelerated and secure application performance even under the most demanding conditions.
NVIDIA IGX Product Family
NVIDIA IGX Thor
NVIDIA IGX T7000
MicroATX board kit equipped with discrete GPU (dGPU) support, functional safety features, NVIDIA ConnectX-7 networking, and BMC (Baseboard Management Controller) support.
NVIDIA IGX T5000
An IGX Thor System-on-Module (SoM) solution for custom carrier board designs with optional functional safety features.
NVIDIA IGX Thor Developer Kit
A development kit featuring a pre-installed dGPU for all IGX Thor use cases.
NVIDIA IGX Thor Developer Kit Mini
A development kit designed for robotics safety.
NVIDIA IGX Orin
NVIDIA IGX Orin 700
A MicroATX board kit offering discrete GPU (dGPU) support, functional safety, NVIDIA ConnectX-7, and BMC support.
NVIDIA IGX Orin Developer Kit
A development kit for all IGX Orin use cases.
Specifications
Customers can select a complete development kit or a SoM for production-ready NVIDIA-Certified Systems, all of which ensure enterprise customers receive performance, reliability, and support from NVIDIA AI Enterprise software.
The NVIDIA IGX Thor™ Developer Kit accelerates industrial and medical edge innovation with real-time sensor processing and AI reasoning.
| IGX Thor Developer Kit | IGX Thor Developer Kit Mini | |
|---|---|---|
| dGPU | With NVIDIA RTX PRO™ 6000 Blackwell Max-Q Workstation Edition | - |
| AI Performance | Up to 5581 TFLOPS (FP4-Sparse) Storage: 64 GB eMMC | 2070 TFLOPS (FP4-Sparse) |
| iGPU | NVIDIA Blackwell architecture GPU with 2560 cores across 10 TPCs, fifth-generation Tensor Cores, Multi-Instance GPU (MIG) | NVIDIA Blackwell architecture GPU with 2560 cores across 10 TPCs, fifth-generation Tensor Cores, Multi-Instance GPU (MIG) |
| GPU Max Frequency | 1.57 GHz | 1.57 GHz |
| dGPU | NVIDIA Blackwell architecture GPU with 24,064 cores and fifth-generation Tensor Cores | - |
| CPU | 14-core Arm® Neoverse®-V3AE 64-bit CPU; 64 KB I-Cache, 64 KB D-Cache, and 1 MB L2 Cache per core, with 16 MB Shared System L3 Cache | 14-core Arm® Neoverse®-V3AE 64-bit CPU; 64 KB I-Cache, 64 KB D-Cache, and 1 MB L2 Cache per core, with 16 MB Shared System L3 Cache |
| CPU Max Frequency | 2.6 GHz | 2.6 GHz |
| FVision Accelerator | 1x PVA v3 | 1x PVA v3 |
NVIDIA IGX Thor™ Production Systems enable customers to easily bring their products to production alongside our OEM partners, while all NVIDIA-Certified Systems guarantee that enterprise customers receive exceptional performance, reliability, and support from NVIDIA AI Enterprise software.
| IGX T7000 | IGX T5000 | ||
|---|---|---|---|
| dGPU | Integrated with NVIDIA RTX PRO™ 6000 Blackwell Max-Q Workstation Edition. | Integrated with NVIDIA RTX PRO™ 5000 Blackwell. | - |
| AI Performance | Up to 5581 TFLOPS (FP4-Sparse) | Up to 4293 TFLOPS (FP4-Sparse) | Up to 2070 TFLOPS (FP4-Sparse) |
| iGPU | NVIDIA Blackwell architecture GPU with 2560 cores and 5th Gen Tensor Cores; features Multi-Instance GPU (MIG) support with 10 TPCs. | NVIDIA Blackwell architecture GPU with 2560 cores and 5th Gen Tensor Cores; features Multi-Instance GPU (MIG) support with 10 TPCs. | NVIDIA Blackwell architecture GPU with 2560 cores and 5th Gen Tensor Cores; features Multi-Instance GPU (MIG) support with 10 TPCs. |
| iGPU Max Frequency | 1.57 GHz | 1.57 GHz | 1.57 GHz |
| dGPU | NVIDIA Blackwell architecture GPU with 24,064 cores and 5th Gen Tensor Cores. | NVIDIA Blackwell architecture GPU with 14,080 cores and 5th Gen Tensor Cores. | - |
| CPU | 14-core Arm® Neoverse®-V3AE 64-bit CPU, 64 KB I-Cache, 64 KB D-Cache, 1 MB L2 Cache per core, 16 MB shared system L3 Cache. | 14-core Arm® Neoverse®-V3AE 64-bit CPU, 64 KB I-Cache, 64 KB D-Cache, 1 MB L2 Cache per core, 16 MB shared system L3 Cache. | 14-core Arm® Neoverse®-V3AE 64-bit CPU, 64 KB I-Cache, 64 KB D-Cache, 1 MB L2 Cache per core, 16 MB shared system L3 Cache. |
| CPU Max Frequency | 2.6 GHz | 2.6 GHz | 2.6 GHz |
| FVision Accelerator | 1x PVA v3 | 1x PVA v3 | 1x PVA v3 |
Accelerate the development of edge AI applications with the NVIDIA IGX Orin Developer Kit. Complete with an enclosure and power supply, this kit is ideal for high-performance, industrial-grade edge AI applications.
| IGX Orin Developer Kit | |
|---|---|
| AI Performance | Up to 1,705 TOPS of AI performance with the optional RTX 6000 Ada Generation GPU |
| SOM (System on Module) | GPU: NVIDIA Ampere architecture GPU with 2,048 CUDA cores and 64 Tensor Cores CPU: 12-core Arm® Cortex®-A78AE v8.2 CPU architecture |
| NVIDIA ConnectX-7 | NVIDIA ConnectX-7 networking Two 100 GbE Ethernet ports 32-lane PCIe Gen 5 switch (x8 upstream, x16 downstream, x8 downstream) |
| Safety MCU (sMCU) | Infineon Aurix TC397 |
| NVIDIA BMC (Baseboard Management Controller) Module | Aspeed AST2600 Microchip ERoT |
| GPU Max Frequency | 1.185 GHz |
| CPU Max Frequency | 1.971 GHz |
Customers can choose between a complete board kit or a System on Module (SoM) as part of NVIDIA Certified Systems designed for production. All NVIDIA Certified Systems include support for NVIDIA AI Enterprise software, providing enterprise customers with high performance, reliability, and comprehensive support.
| IGX Orin 700 | |
|---|---|
| Performance | Up to 1,705 TOPS |
| Technical Specifications | iGPU: NVIDIA Ampere GPU with 2,048 CUDA cores and 64 Tensor Cores CPU: 12-core Arm Cortex-A78AE CPU Memory: 64 GB LPDDR5 with ECC Storage: 64 GB eMMC |
| Power | Without dGPU: Up to 125 W power consumption With dGPU: 400 W power consumption |
| I/O Throughput | 2X100 Gb/s |
| Optional dGPU | ✓ |
| Integrated Dual 100 Gb/s NVIDIA ConnectX-7 | ✓ |
| BMC | ✓ |
| Functional Safety | ✓ |
| Carrier Board Customization | ✓ |
| Product Lifecycle and Enterprise Software Support | 10 years (through 2033) |
| Value Proposition | The same NVIDIA certification process is used to ensure enterprise-grade software support, AI safety, and functional safety. |
Comprehensive Software Platform for Enterprise Edge AI Systems
IGX delivers a production-ready solution by combining industrial-grade hardware with enterprise software and support services.
Comprehensive Software Platform for Enterprise Edge AI Systems
IGX combines industrial-grade hardware with enterprise software and support services to deliver a production-ready solution.
IGX Thor: Superior Performance for Industrial-Grade Systems
The NVIDIA IGX Thor platform enables real-time sensor processing and AI reasoning for physical AI agents through its industrial-grade hardware, enterprise software, and functional safety features. The platform is powered by an integrated GPU (iGPU) based on the NVIDIA Blackwell Architecture, with support for an optional discrete GPU (dGPU). This configuration delivers up to 5,581 FP4 TFLOPS of AI computing performance, making it possible to run multiple generative AI models simultaneously at the edge.
Compared to NVIDIA IGX Orin, IGX Thor:
- Up to 8× higher AI processing performance on the iGPU,
- Up to 2.5× higher AI processing performance on the dGPU,
- And 2× greater connectivity capacity.
Functional Safety for Dynamic Robotic Environments
NVIDIA IGX Thor combines high-performance AI computing with functional safety for robots operating in dynamic environments. The platform supports both an “inside-out” safety approach, which uses sensors on the vehicle or robot, and an “outside-in” safety approach, which leverages surrounding infrastructure sensors.
At the core of the platform is a dedicated Functional Safety Island that provides an independent safety processor, isolating safety-critical workloads from other system components. This architecture is designed to meet the ASIL D/SC3 and ASIL/SIL 2 safety levels of the ISO 26262 and IEC 61508 standards.
Enterprise-Grade Software for AI Applications Used in Mission-Critical Tasks
NVIDIA AI Enterprise software runs edge AI applications on NVIDIA IGX Thor, delivering mission-critical performance, security, and long-term enterprise support across all software and hardware layers.
The platform seamlessly runs generative AI applications, from VLA (Vision-Language-Action) models such as NVIDIA Isaac GR00T N to popular large language models (LLMs) and vision-language models (VLMs) such as NVIDIA Cosmos Reason. It also brings cloud-based NVIDIA AI Enterprise and NVIDIA NIM technologies to the edge environment.
Supported applications include:
- Robotics with NVIDIA Isaac,
- Visual AI with NVIDIA Metropolis,
- Sensor data processing with NVIDIA Holoscan,
- And NVIDIA Halos.
