NVIDIA Jetson AGX Thor Overview and Developer Resources

From RidgeRun Developer Wiki


Follow Us On Twitter LinkedIn Email Share this page






  1. REDIRECT development:NVIDIA Jetson AGX Thor/Main contents

Welcome to RidgeRun's guide to NVIDIA®Jetson AGX Thor™



NVIDIA® Jetson AGX Thor is a high-performance computing platform for advanced AI and robotics—particularly humanoid robots. It features a Blackwell-architecture GPU and 128 GB LPDDR5X, delivering up to 2070 FP4 TFLOPS of AI compute for high-speed sensor processing and real-time control.

NVIDIA Jetson AGX Thor


NVIDIA Jetson AGX Thor developer kit


The developer kit is the reference platform by NVIDIA: it exposes high-bandwidth I/O, power delivery, and thermal design so you can profile end-to-end workloads (LLMs, VLMs, sensor fusion) under realistic conditions.

NVIDIA Jetson AGX Thor SoM


Jetson AGX Thor System-on-Module (SoM)


At the heart of the developer kit is the Jetson AGX Thor SoM, a production-ready compute module that integrates CPU, GPU, memory, and accelerators. The SoM is designed to be embedded directly into final products—robots, autonomous machines, and edge servers—where size, power, and reliability constraints matter.

What is new in NVIDIA Jetson AGX Thor?

NVIDIA® Jetson AGX Thor is the latest flagship in the NVIDIA Jetson lineup, pushing edge AI computing to a new level of performance and scalability. Some of its main features are:


Feature Jetson AGX Thor (2025) Gain vs Jetson AGX Orin (64 GB)
CPU Complex 14-core Arm® Neoverse V3AE (64-bit) with SMP architecture +2 cores (≈16% more), newer Neoverse V3 vs Cortex-A78AE → higher IPC and efficiency
GPU Architecture Blackwell with 2560 CUDA Cores and 96 4th Gen Tensor Cores, new Transformer Engine, MIG support +512 CUDA cores (≈25% more) and +32 Tensor Cores (50% more), plus new Transformer Engine for LLMs
Memory 128 GB LPDDR5X 2x capacity
Memory Bandwidth 256-bit bus @ 4266 MHz → 273 GB/s ~34% more bandwidth (Orin ≈ 204 GB/s)
On-Edge AI Performance Up to 2070 TFLOPs (FP4) | 1035 TFLOPs (FP8) Massive jump vs Orin's ~275 TOPS INT8 (Thor delivers ~7.5x more FP8 throughput)
Holoscan Sensor Bridge Ethernet sensor-to-GPU direct, low latency, auto-integration Entirely new: Orin relied on PCIe/USB bridges
Multi-Instance GPU (MIG) Partition GPU into 2 MIG instances, each independent New feature absent on Orin
PVA 3.0 165 GFLOPs (FP32) | 320 GFLOPs (FP16) ≈2x performance of Orin's PVA 2.0


What are the potential applications of NVIDIA Jetson AGX Thor?

Jetson AGX Thor is purpose-built for advanced edge AI and robotics use cases, including:

  • Autonomous Machines – Industrial robots, collaborative robots, and autonomous mobile robots requiring real-time multimodal sensor fusion.
  • AI-Enhanced Healthcare – Medical imaging, surgical robotics, and Holoscan-enabled sensor-driven diagnostics.
  • Smart Cities & Infrastructure – Intelligent traffic systems, real-time video analytics, and large-scale sensor network management.
  • Next-Gen Vehicles – Autonomous driving, driver-assist systems, and software-defined vehicle platforms.
  • Generative AI on the Edge – Running large transformer-based LLMs, vision-language models, and generative applications in disconnected or low-latency environments.
  • Defense & Aerospace – Rugged AI deployments for UAVs, drones, and mission-critical sensor fusion workloads.

NVIDIA Jetson AGX Thor represents the convergence of HPC, AI, and sensor integration — all at the edge.

RidgeRun knows how important documentation is for your project and how a single source of information may decrease the time-to-market of your product. That's why we've decided to share our expertise with the developer community, creating a centralized wiki with technical documentation, tutorials, examples, and much more!
The following video from NVIDIA provides a nice overview of the NVIDIA Jetson AGX Thor.

RidgeRun Support

RidgeRun is an official NVIDIA Partner, and we have created this extensive set of documentation to support our joint customers. If you have any questions on the content, please contact us through our contact us page.

RidgeRun provides support for embedded Linux Camera Driver development for NVIDIA's platforms, specializing in the use of hardware accelerators in multimedia applications. RidgeRun's products take full advantage of the accelerators that NVIDIA exposes to perform transformations on the video streams, achieving great performance on complex processes.

This page contains detailed guides and information on how to get started with the NVIDIA Jetson AGX Thor and start using its full capabilities.

To get up-to-speed with your NVIDIA Jetson AGX Thor board, start by clicking below:



For direct inquiries, please refer to the contact information available on our Contact page. Alternatively, you may complete and submit the form provided at the same link. We will respond to your request at our earliest opportunity.


Links to RidgeRun Resources and RidgeRun Artificial Intelligence Solutions can be found in the footer below.




  Index Next: Introduction