NVIDIA CUDA Installation Guide for Linux

Step‑by‑step guide for installing and verifying the NVIDIA CUDA Toolkit on supported Linux distributions.

Brand: Nvidia
Category: Application program
Document type: Installation Guide
Language: English
Pages: 33
File format: PDF
File size: 1.8 MB
Published: 05 October, 2026
Updated: 05 October, 2026
MD5 checksum: 73E70C42377865C2E60189491B56A629
PDF: NVIDIA CUDA Installation Guide for Linux PDF

About this manual

This guide explains how to install the NVIDIA CUDA Toolkit on Linux systems, covering pre‑installation checks, package‑manager and runfile installation methods, and post‑installation verification. It also includes a cross‑build environment for ARM and answers common installation questions.

The document is aimed at users familiar with the Linux command line and provides detailed instructions for setting environment variables, handling conflicting installations, and extracting installer contents. It concludes with additional considerations and links to further development resources.

What's inside

Introduction
System Requirements
Pre‑installation Actions
Package Manager Installation
Runfile Installation
Cross‑build Environment for ARM
Post‑installation Actions
Frequently Asked Questions
Additional Considerations

Specifications

Supported Linux distributions	Fedora 20, CentOS 6.x, Ubuntu 14.04, OpenSUSE 13.1, SLES 11 SP3, etc.
Required GPU	CUDA‑capable NVIDIA GPU
Required compiler	gcc compiler and toolchain

Frequently asked questions

Why do I see "nvcc: No such file or directory" when I try to build a CUDA application?

Your PATH environment variable is not set correctly; add the CUDA bin directory (e.g., /usr/local/cuda-6.5/bin) to PATH.

Why do I see "error while loading shared libraries" when I run a CUDA application?

Your LD_LIBRARY_PATH is missing the CUDA library directories; add /usr/local/cuda-6.5/lib or /usr/local/cuda-6.5/lib64 to LD_LIBRARY_PATH.

Download PDF (1.8 MB)