Ubuntu Nvidia driver驱动安装（新）
<h1 id="前言">前言</h1><p>英伟达更新了安装驱动的方式，更新一下文档</p>
<p>旧文：Ubuntu Nvidia driver驱动安装及卸载</p>
<h1 id="下载官方驱动安装">下载官方驱动安装</h1>
<p>1.安装驱动前一定要更新软件列表和安装必要软件、依赖（必须）</p>
<pre><code>sudo apt-get update #更新软件列表
sudo apt-get install -y g++ gcc make
</code></pre>
<p>2.查看GPU型号（你自己知道显卡型号就不用查了，直接下一步；通用查看显卡型号命令：lspci | grep -i vga）</p>
<pre><code>lspci | grep -i nvidia
</code></pre>
<p>得到如下输出，其中 <code>Tesla T4</code> 就是型号</p>
<pre><code>00:06.0 3D controller: NVIDIA Corporation TU104GL (rev a1)
00:07.0 3D controller: NVIDIA Corporation TU104GL (rev a1)
</code></pre>
<p>3.官网下载对应驱动，注意要把 nvidia 驱动放在英文名文件夹下</p>
<p>https://www.nvidia.cn/drivers/lookup/</p>
<p><img src="https://img2024.cnblogs.com/blog/1303876/202409/1303876-20240904181146260-1039598493.png" alt="" loading="lazy"></p>
<p>4.卸载原有驱动，请看下方「正确卸载驱动以及查看是否卸载干净命令」</p>
<p>5.禁用nouveau(nouveau是通用的驱动程序)（必须）</p>
<blockquote>
<p>Ubuntu 系统集成的显卡驱动程序是 nouveau ，它是第三方为 NVIDIA 开发的开源驱动，我们需要先将其屏蔽再安装 NVIDIA 官方驱动，否则直接安全驱动的话往往会报以下错误：ERROR: The Nouveau kernel driver is currently in use by your system. This driver is incompatible with the NVIDIA driver</p>
</blockquote>
<p>将 <code>nouveau</code> 驱动加入黑名单，将如下内容加入 <code>/etc/modprobe.d/blacklist.conf</code> 文件中：</p>
<pre><code>sudo vim /etc/modprobe.d/blacklist.conf
# 添加如下内容
blacklist vga16fb
blacklist nouveau
blacklist rivafb
blacklist rivatv
blacklist nvidiafb
</code></pre>
<p>然后执行命令</p>
<pre><code>sudo update-initramfs -u
</code></pre>
<p>重启机器后（必须），执行如下命令，如果没有输出则证明禁用成功</p>
<pre><code>sudo lsmod | grep nouveau
</code></pre>
<p>6.开始安装 NVIDIA 官方驱动</p>
<p>执行 <code>dpkg -i</code> 将本地仓库包安装到系统中</p>
<pre><code>sudo dpkg -i nvidia-driver-local-repo-ubuntu2004-560.35.03_1.0-1_amd64.deb
</code></pre>
<p>输出</p>
<pre><code>sudo dpkg -i nvidia-driver-local-repo-ubuntu2004-560.35.03_1.0-1_amd64.deb
Selecting previously unselected package nvidia-driver-local-repo-ubuntu2004-560.35.03.
(Reading database ... 114186 files and directories currently installed.)
Preparing to unpack nvidia-driver-local-repo-ubuntu2004-560.35.03_1.0-1_amd64.deb ...
Unpacking nvidia-driver-local-repo-ubuntu2004-560.35.03 (1.0-1) ...
Setting up nvidia-driver-local-repo-ubuntu2004-560.35.03 (1.0-1) ...
The public nvidia-driver-local-repo-ubuntu2004-560.35.03 GPG key does not appear to be installed.
To install the key, run this command:
sudo cp /var/nvidia-driver-local-repo-ubuntu2004-560.35.03/nvidia-driver-local-63C60622-keyring.gpg /usr/share/keyrings/
</code></pre>
<p>此时复制授权密钥</p>
<pre><code>sudo cp /var/nvidia-driver-local-repo-ubuntu2004-560.35.03/nvidia-driver-local-63C60622-keyring.gpg /usr/share/keyrings/
</code></pre>
<p>再次执行</p>
<pre><code>sudo dpkg -i nvidia-driver-local-repo-ubuntu2004-560.35.03_1.0-1_amd64.deb
</code></pre>
<p>修复依赖关系（如果有问题）：</p>
<pre><code>sudo apt-get install -f
</code></pre>
<p>更新包列表</p>
<pre><code>sudo apt-get update
</code></pre>
<p>搜索驱动 nvidia-driver-560，是否存在</p>
<pre><code>sudo apt-cache search nvidia-driver-560
</code></pre>
<p>输出</p>
<pre><code>nvidia-driver-530-open - Transitional package for nvidia-driver-560-open
nvidia-driver-530 - Transitional package for nvidia-driver-560
nvidia-driver-560-open - NVIDIA driver (open kernel) metapackage
nvidia-driver-560 - NVIDIA driver metapackage
nvidia-headless-560-open - NVIDIA headless metapackage (open kernel module)
nvidia-headless-560 - NVIDIA headless metapackage
nvidia-headless-no-dkms-560-open - NVIDIA headless metapackage - no DKMS (open kernel module)
nvidia-headless-no-dkms-560 - NVIDIA headless metapackage - no DKMS
xserver-xorg-video-nvidia-560 - NVIDIA binary Xorg driver
</code></pre>
<p>开始安装驱动</p>
<pre><code>sudo apt-get install nvidia-driver-560
</code></pre>
<p>安装完成，最后执行 <code>nvidia-smi</code>，有输出，表示安装成功！！</p>
<pre><code>nvidia-smi
Wed Sep4 10:06:18 2024
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 560.35.03 Driver Version: 560.35.03 CUDA Version: 12.6 |
|-----------------------------------------+------------------------+----------------------+
| GPUName Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| FanTemp Perf Pwr:Usage/Cap | Memory-Usage | GPU-UtilCompute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0Tesla T4 Off | 00000000:00:06.0 Off | 0 |
| N/A 62C P0 29W / 70W | 1MiB /15360MiB | 5% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
| 1Tesla T4 Off | 00000000:00:07.0 Off | 0 |
| N/A 67C P0 31W / 70W | 1MiB /15360MiB | 5% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
|GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
|No running processes found |
+-----------------------------------------------------------------------------------------+
</code></pre>
<h1 id="正确卸载驱动以及查看是否卸载干净命令">正确卸载驱动以及查看是否卸载干净命令</h1>
<pre><code>如果安装完显卡驱动之后，发现有问题，需要卸载，用下面方法，一步步来。
方法一：
#先查看驱动以及版本安装情况，命令如下：
ls /usr/src | grep nvidia
#进入安装目录，用驱动自带卸载命令卸载，一般情况下能卸载干净
cd /usr/bin
ls nvidia-*
sudo nvidia-uninstall
#然后用上面的查看命令查看是否卸载干净，完成卸载
ls /usr/src | grep nvidia
#如果需要卸载干净所有英伟达驱动命令，如下
sudo apt-get remove --purge nvidia-*（/nvidia*)
sudo apt autoremove # 此命令在万般无奈之下在使用，可能会卸载无辜软件，影响稳定性。。。
方法二：（推荐）
#直接卸载驱动
sudo ./显卡驱动包名称 --uninstall
#卸载两件套
sudo apt-get purge nvidia*
sudo apt-get autoremove
sudo reboot
</code></pre><br><br>
来源：https://www.cnblogs.com/niuben/p/18397121
頁:
[1]