1. 卸载原驱动

sudo apt remove *cuda*
sudo apt remove *nvidia*
sudo /usr/bin/nvidia-uninstall
sudo dpkg -l | grep ^rc | cut -d' ' -f3 | sudo xargs dpkg --purge
sudo rm -rf ~/.cuda-license-*
sudo apt purge nvidia-cuda-toolkit
sudo apt remove nvidia-driver-*
sudo apt purge nvidia-*

2. 禁用nouveau驱动

修改/etc/modprobe.d/blacklist.conf，在最好添加以下代码

blacklist nouveau
options nouveau modeset=0

执行命令sudo update-initramfs -u
重启后执行 lsmod | grep nouveau，没有显示则禁用成功

3. 下载安装驱动

下载

sudo wget https://developer.download.nvidia.com/compute/cuda/11.7.0/local_installers/cuda_11.7.0_515.43.04_linux.run

安装

sudo sh cuda_11.7.0_515.43.04_linux.run
输入accept确认

选择Install安装

查看

nvcc -V

nvidia-smi

其他

docker运行指定runtime=nvidia报错
unknown or invalid runtime name: nvidia

可能原因

未安装nvidia-container-runtime

解决方案

执行安装nvidia-container-runtime
sudo apt install nvidia-container-runtime
修改/etc/docker/daemon.json

{
    "registry-mirrors": [
        "https://docker.mirrors.ustc.edu.cn",
        "https://hub-mirror.c.163.com",
        "https://registry.docker-cn.com"
    ],
    "runtimes": {
        "nvidia": {
            "args": [],
            "path": "nvidia-container-runtime"
        }
    }
}

重启docker

sudo systemctl daemon-reload
sudo systemctl restart docker