type
status
date
slug
summary
tags
category
icon
password

CUDA Driver Upgrade for Windows

Typically, the Windows computer you purchase comes with the graphics card driver pre-installed. Therefore, we can directly open the application that is already installed on the computer:
notion image
Follow the steps as described in the CSDN blog.

CUDA Upgrade for Windows with WSL Virtual Machine

If you are upgrading the cuda driver in wsl, first upgrade the relevant drivers on windows, and then proceed to upgrade the drivers in wsl.

Ensure System Updates

Add NVIDIA Official Driver PPA

Add PPA

<BASH>

Verify PPA Addition

If you see output similar to the following, the PPA has been successfully added:

Find Available NVIDIA Drivers

You will see output similar to the following:
Generally, selecting the latest stable version is the best choice. For example, if nvidia-driver-560 is the latest, you can choose to install it.

Install the Selected NVIDIA Driver

Disable Nouveau Driver (if not already disabled)

Nouveau is an open-source NVIDIA driver that may conflict with the official driver. Typically, Nouveau is automatically disabled when installing the official driver, but we can manually confirm.
Create or Edit the Blacklist File
Add the following content:
Save and exit (in nano, press Ctrl + O to save, Ctrl + X to exit).
Update initramfs

Reboot the System

Verify NVIDIA Driver Installation

Sample output (the effective output should show changes in NVIDIA-SMI and Driver version)
notion image

Download and Install CUDA 12.4

You can go to the NVIDIA CUDA Toolkit 12.4 Download Page to download and execute all the instructions provided by the official website!

List Installed CUDA Versions

Sample output:

Check if the Desired CUDA Version's nvcc is Correctly Installed

<BASH>
Expected output:

Configure Environment Variables to Use CUDA 12.4

Edit Shell Configuration File:
Add CUDA 12.4 Path:
Add the following at the end of the file:
Update the cuda Symbolic Link:
Apply Changes:
Verify Environment Variables:
Expected Output:
算法秋招面试-手撕神经网络代码python快排实现