Installation Guide Overview#
This guide provides complete instructions for installing and deploying TensorRT-RTX on supported platforms. Whether you’re setting up TensorRT-RTX for the first time or upgrading an existing installation, this guide will walk you through the process.
About TensorRT-RTX#
NVIDIA TensorRT for RTX (TensorRT-RTX) is a compact, high-performance inference library that brings TensorRT optimization to NVIDIA RTX GPUs across desktops, laptops, and workstations. It uses Just-In-Time (JIT) compilation to automatically optimize models for the end-user’s specific GPU — eliminating the need for ahead-of-time engine builds per device.
For a deeper look at how TensorRT-RTX works, its two-phase compilation pipeline, and how it relates to other TensorRT ecosystem libraries, refer to the Architecture Overview.
What’s in This Section#
This installation guide is organized into the following sections:
Prerequisites
Before installing TensorRT-RTX, review system requirements, supported platforms, and required dependencies.
Installing TensorRT-RTX
Step-by-step instructions for installing TensorRT-RTX using:
SDK zip file (Windows)
Tarball file (Linux)
Deploy Your First Model
Walk through the complete deployment pipeline — download an ONNX model, build a portable engine (AOT), and run inference with JIT compilation.
ONNX Conversion Guide
Export your trained model to ONNX from PyTorch, TensorFlow, or Hugging Face — or build ONNX models programmatically.
For programmatic inference using the C++ or Python API, refer to Using the Native Runtime API in the Inference Library section.