Abhigyan Chakraborty

Enabling GPU Computing in WSL2 and Linux with CUDA

Phase 3: High-Performance Computing — Part 1

(Part 8 of series Blueprint for a Modern Research Computing Environment)

Follow me :

LinkedIn    Website    Website

Quick Summary

This article installs and verifies NVIDIA drivers so your GPU is accessible from Ubuntu — one path for WSL2 users, one for native Linux users — and deliberately skips a system-wide CUDA Toolkit install to avoid conflicts with the pip-bundled CUDA libraries Part 10 will install. It closes with a heads-up on one extra step TensorFlow on WSL sometimes needs, so it doesn’t read as an error when it happens.

Objective

This part is only relevant if your computer has an NVIDIA GPU. If you have an AMD or Intel GPU, or no dedicated GPU, skip to Part 10 — GPU setup for those is a different process entirely.

By the end, you’ll have:

Content

💡 Getting Unstuck (Expand for AI Troubleshooting Prompts)

AI tools (like Claude, Grok, Gemini, or ChatGPT) give the best troubleshooting advice when they have the exact text of the tutorial. To ensure the AI understands what you are trying to build, we will give it the actual file.

1. Download the tutorial file

2. Upload it to your AI

3. Ask for Help Copy and paste this exact prompt into the chat along with your file:

I have attached the markdown file for the tutorial I am following. Please read it so you understand the specific environment I am trying to build.

I need help with the following:

Step [X]: [paste exact step text from the blog]

Command I ran: [paste exact command]

What happened: [paste full output/error — if the terminal was truly blank, say so explicitly]

Please help me troubleshoot and fix this error. You can use your general knowledge to solve the problem, but your solution MUST align with the architecture and tools taught in the attached file. Do not suggest alternative setups that contradict the tutorial. Once fixed, tell me what to do next in the article.

To go deeper on a step before you run a command:

I have attached the Part8.md file. Look at Step [X] and explain exactly what the command does and why we are doing it before I run it.

Think of this series as the roadmap and your AI assistant as your learning companion.

Why GPU Matters for Research and ML

GPUs run thousands of small operations in parallel, making them exceptionally powerful for machine learning and scientific computing. For large datasets or complex models, GPU access is a practical necessity.

Prerequisites

How GPU Access Works: WSL vs Native Linux

WSL: Your GPU is managed by Windows; you never install drivers inside Ubuntu on WSL. Native Linux: You install the NVIDIA driver on the Linux system itself.

Note on CUDA Toolkit: This article skips the system-wide CUDA Toolkit to avoid conflicts with pip-bundled libraries.

NVIDIA GPU drivers installation

For WSL2 Users

💡Expand for details

Step 1 — Install or Update NVIDIA Drivers on Windows

CUDA support on WSL is powered by your Windows NVIDIA driver, which acts as a bridge between the hardware and your Linux environment. If your driver is outdated, WSL may fail to detect your GPU.

1. Identify your GPU: Not sure which card you have? Open Task Manager (Ctrl+Shift+Esc), go to the Performance tab, and click GPU on the left. The name of your card is in the top-right corner.

2. Download the Driver: Visit the NVIDIA Driver Downloads page.

3. Installation: Run the downloaded file. Select Express Installation to let the installer handle the settings automatically.

4. Finalize: Restart your computer. This step is mandatory for the new driver to properly initialize the communication layer between Windows and WSL.

Step 2 — Verify GPU Access Inside WSL

Open your Ubuntu terminal and run:

  nvidia-smi

If you see a table showing your GPU name, driver, and CUDA version, you are set. If command not found, update your Windows driver.

Step 3 — Set the WSL GPU Library Path

WSL exposes GPU driver interfaces via /usr/lib/wsl/lib. Open your .bashrc (nano ~/.bashrc) and add:

  export LD_LIBRARY_PATH=/usr/lib/wsl/lib:$LD_LIBRARY_PATH

Save (Ctrl+X, Y, Enter) and run source ~/.bashrc.

Step 4 — A Heads-Up About Part 10

When installing TensorFlow in Part 10 (pip install tensorflow[and-cuda]), it is normal for it to occasionally fail to locate libraries on WSL. This is an issue with library discovery, not your driver. We will fix this in Part 10 with a scoped LD_LIBRARY_PATH export.

Native Linux Users

💡Expand for details

Step 1 — Install NVIDIA Drivers

Follow NVIDIA’s official installation guide. You only need the driver—the next part will handle the specific CUDA libraries.

Step 2 — Verify GPU Access

Run nvidia-smi. You should see your GPU name and version without errors.

Before You Proceed to Part 10

Run nvidia-smi and note the maximum supported CUDA Version (e.g., 12.x). Use this as an upper bound when installing PyTorch and TensorFlow.

Warning: Check ~/.bashrc for old LD_LIBRARY_PATH=/usr/local/cuda/lib64 entries and remove them; they conflict with pip-installed CUDA libraries.

What’s Next

What You’ve Done:

Next: Part 9 — Installing PyTorch and TensorFlow | Previous: Part 7 — Project Organization and Managing Scientific Data

All Blogs

This site is open source. Improve this page.