如何在UUbuntu 24.04 安装 NVIDIA L40 显卡驱动与 CUDA(12.8 / 13.1)

To install the NVIDIA L40 driver and cuda on Ubuntu 24.04
适用于 NVIDIA L40 / L40S 数据中心显卡,在 Ubuntu 24.04 LTS 系统下部署深度学习、AI 推理、CUDA 计算环境。

一、环境说明
操作系统:Ubuntu 24.04 LTS(x86_64 架构)
GPU 型号:NVIDIA L40(数据中心级显卡)
NVIDIA 驱动
nvidia-driver-570-server(推荐,生产环境稳定)
nvidia-driver-590-server(可选,用于支持 CUDA 13.1)
CUDA Toolkit
CUDA 12.8(长期稳定版本,推荐)
CUDA 13.1(新特性版本,测试/验证使用)
BIOS 启动模式:UEFI (不要开启EFI安全引导,Secure Boot 会阻止未签名或非官方签名的内核模块加载,而 NVIDIA 驱动、VFIO 及 DKMS 模块往往不满足 Secure Boot 的签名要求)
Above 4G Decoding:已开启(GPU 透传 / 大显存 BAR 必需)

二、打开终端并更新
使用sudo更新apt软件包列表并使用sudo升级apt软件包

sudo apt update && sudo apt upgrade -y

三、查看 Ubuntu 可用的 NVIDIA 驱动版本
使用sudo命令列出ubuntu驱动程序列表

sudo ubuntu-drivers list
......
nvidia-driver-565, (kernel modules provided by nvidia-dkms-565)
nvidia-driver-535, (kernel modules provided by linux-modules-nvidia-535-generic)
nvidia-driver-575, (kernel modules provided by nvidia-dkms-575)
nvidia-driver-555-open, (kernel modules provided by nvidia-dkms-555-open)
nvidia-driver-575-open, (kernel modules provided by nvidia-dkms-575-open)
nvidia-driver-570-server, (kernel modules provided by linux-modules-nvidia-570-server-generic)
nvidia-driver-590-server, (kernel modules provided by linux-modules-nvidia-590-server-generic)
nvidia-driver-555, (kernel modules provided by nvidia-dkms-555)
nvidia-driver-535-server-open, (kernel modules provided by linux-modules-nvidia-535-server-open-generic)
nvidia-driver-560, (kernel modules provided by nvidia-dkms-560)
nvidia-driver-570-server-open, (kernel modules provided by linux-modules-nvidia-570-server-open-generic)
nvidia-driver-570-open, (kernel modules provided by linux-modules-nvidia-570-open-generic)
nvidia-driver-535-server, (kernel modules provided by linux-modules-nvidia-535-server-generic)
nvidia-driver-560-open, (kernel modules provided by nvidia-dkms-560-open)
nvidia-driver-565-open, (kernel modules provided by nvidia-dkms-565-open)
nvidia-driver-550, (kernel modules provided by linux-modules-nvidia-550-generic)
nvidia-driver-550-open, (kernel modules provided by linux-modules-nvidia-550-open-generic)
nvidia-driver-570, (kernel modules provided by linux-modules-nvidia-570-generic)
nvidia-driver-535-open, (kernel modules provided by linux-modules-nvidia-535-open-generic)
open-vm-tools-desktop

四、安装驱动
L40 / A100 / H100 等数据中心卡,Server 驱动比普通驱动更适合 L40 / 多卡服务器 / 无显示环境

sudo apt install nvidia-driver-570-server    #数据中心卡(如L40)专用驱动,稳定性更好,支持 MIG、多用户多实例等特性

如果需要新版可以安装
sudo apt install nvidia-driver-590-server

五、重启系统

sudo reboot

六、验证驱动是否正常加载

root@su:~# nvidia-smi
Thu Jun 12 06:14:12 2025       
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.133.20             Driver Version: 570.133.20     CUDA Version: 12.8     |
|-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA L40                     Off |   00000000:13:00.0 Off |                    0 |
| N/A   29C    P0             79W /  300W |       0MiB /  46068MiB |      3%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+
                                                                                         
+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|  No running processes found                                                             |
+-----------------------------------------------------------------------------------------+

7、安装 CUDA
安装CUDA12.8

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/cuda-keyring_1.1-1_all.deb
sudo dpkg -i cuda-keyring_1.1-1_all.deb
sudo apt-get update
sudo apt-get -y install cuda-toolkit-12-8

用 nvcc --version 确认cuda的版本,如果显示Command nvcc not found,则编辑~/.bashrc,配置 CUDA 环境变量(解决 nvcc not found)

vim ~/.bashrc
export PATH=/usr/local/cuda-12.8/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-12.8/lib64:${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

#更新变量
source ~/.bashrc

root@su:~# nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2025 NVIDIA Corporation
Built on Fri_Feb_21_20:23:50_PST_2025
Cuda compilation tools, release 12.8, V12.8.93
Build cuda_12.8.r12.8/compiler.35583870_0

安装CUDA 13.1

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/cuda-ubuntu2404.pin
sudo mv cuda-ubuntu2404.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget https://developer.download.nvidia.com/compute/cuda/13.1.1/local_installers/cuda-repo-ubuntu2404-13-1-local_13.1.1-590.48.01-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu2404-13-1-local_13.1.1-590.48.01-1_amd64.deb
sudo cp /var/cuda-repo-ubuntu2404-13-1-local/cuda-*-keyring.gpg /usr/share/keyrings/
sudo apt-get update
sudo apt-get -y install cuda-toolkit-13-1


用 nvcc --version 确认cuda的版本,如果显示Command nvcc not found,则编辑~/.bashrc,配置 CUDA 环境变量(解决 nvcc not found)

vim ~/.bashrc
export PATH=/usr/local/cuda-13.1/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-13.1/lib64:${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

#更新变量
source ~/.bashrc

root@su:~# nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2025 NVIDIA Corporation
Built on Fri_Feb_21_20:23:50_PST_2025
Cuda compilation tools, release 12.8, V12.8.93
Build cuda_12.8.r12.8/compiler.35583870_0

八、锁定驱动版本防止升级冲突

sudo apt-mark hold nvidia-driver-570-server
sudo apt-mark hold cuda-toolkit-12-8



Q1:CUDA 版本一定要和 nvidia-smi 显示一致吗?
不需要完全一致,只要 驱动版本 ≥ CUDA 需求版本 即可。

Q2:可以同时安装多个 CUDA 版本吗?
可以,需手动切换 /usr/local/cuda 软链接或 PATH。

评论已关闭。