[email protected]安装NVIDIA A10驱动(rpm离线形式)

NVIDIA A10 是一款基于 Ampere 架构的 GPU,专为数据中心场景设计,主要面向AI 推理、图形渲染和虚拟桌面基础设施(VDI)等高性能计算需求。它在性能、能效和多用途支持方面表现出色,是 A100 的“轻量级”商业化兄弟产品


1.准备工作

1.启用 Extra Packages for Enterprise Linux (EPEL) 存储库

dnf install epel-release -y

2.安装开发工具可确保必要的构建依赖项

dnf groupinstall “Development Tools” -y

3.构建内核模块所需的头文件和工具

dnf install -y kernel-devel kernel-headers

⚠️:当前内核版本要和安装的保持一致

4.Dynamic Kernel Module Support (DKMS) 是一个用于自动重建内核模块

dnf install dkms -y

2.检查

1.检查当前gpu信息

lspci | grep -i nvidia

3.下载驱动

⚠️:Linux 64-bit xxx.run 通用安装包, Linux 64-bit RHEL 9 xxx.rpm

nvidia-driver-local-repo-rhel9-570.124.06-1.0-1.x86_64.rpm

4.安装驱动

1.安装离线rpm

rpm -ivh nvidia-driver-local-repo-rhel9-570.124.06-1.0-1.x86_64.rpm 
  /etc/yum.repos.d/nvidia-driver-local-rhel9-570.124.06.repo
  /var/nvidia-driver-local-repo-rhel9-570.124.06/

2.安装驱动

dnf module install nvidia-driver:latest-dkms -y

3.查询所安装的具体dkms包名

dkms status
ls /usr/src | grep nvidia

4.编译安装驱动

dkms install -m nvidia -v xxx

[root@vgt-rocky9-60 ~]# dkms install -m nvidia -v 570.124.06
Sign command: /lib/modules/5.14.0-503.38.1.el9_5.x86_64/build/scripts/sign-file
Signing key: /var/lib/dkms/mok.key
Public certificate (MOK): /var/lib/dkms/mok.pub
Certificate or key are missing, generating self signed certificate for MOK...

Building module(s)............................................................................... done.
Signing module /var/lib/dkms/nvidia/570.124.06/build/nvidia.ko
Signing module /var/lib/dkms/nvidia/570.124.06/build/nvidia-modeset.ko
Signing module /var/lib/dkms/nvidia/570.124.06/build/nvidia-drm.ko
Signing module /var/lib/dkms/nvidia/570.124.06/build/nvidia-uvm.ko
Signing module /var/lib/dkms/nvidia/570.124.06/build/nvidia-peermem.ko
Installing /lib/modules/5.14.0-503.38.1.el9_5.x86_64/extra/nvidia.ko.xz
Installing /lib/modules/5.14.0-503.38.1.el9_5.x86_64/extra/nvidia-modeset.ko.xz
Installing /lib/modules/5.14.0-503.38.1.el9_5.x86_64/extra/nvidia-drm.ko.xz
Installing /lib/modules/5.14.0-503.38.1.el9_5.x86_64/extra/nvidia-uvm.ko.xz
Installing /lib/modules/5.14.0-503.38.1.el9_5.x86_64/extra/nvidia-peermem.ko.xz
Adding linked weak modules...
Running depmod...... done.


[root@develop03 ~]# ls /usr/src/ |grep nvidia
nvidia-570.124.06

dkms install -m nvidia -v 570.124.06

5.开启Persistence-M属性(重启nivdia on)

systemctl enable --now  nvidia-persistenced
systemctl status nvidia-persistenced

NVIDIA 驱动中的 Persistence Mode(持久化模式)是用于控制 GPU 在空闲时是否卸载驱动并释放资源的一个选项,主要用于提升性能、缩短响应延迟以及避免频繁加载驱动带来的开销。

A daemon to maintain persistent software state in the NVIDIA driver。

6.重启

reboot

5.离线rpm关于nvidia驱动

[root@vgt-rocky9-60 ~]# dnf search nvidia
Last metadata expiration check: 0:02:48 ago on Tue 29 Apr 2025 06:17:09 AM UTC.
============================================================== Name & Summary Matched: nvidia ==============================================================
kmod-nvidia-open-dkms.noarch : NVIDIA driver open kernel module flavor
libnvidia-cfg.x86_64 : NVIDIA Config public interface (nvcfg)
libnvidia-fbc.i686 : NVIDIA OpenGL-based Framebuffer Capture libraries
libnvidia-fbc.x86_64 : NVIDIA OpenGL-based Framebuffer Capture libraries
libnvidia-ml.i686 : NVIDIA Management Library (NVML)
libnvidia-ml.x86_64 : NVIDIA Management Library (NVML)
libva-nvidia-driver.x86_64 : A VA-API implemention using NVIDIA's NVDEC
nvidia-driver.x86_64 : NVIDIA's proprietary display driver for NVIDIA graphic cards
nvidia-driver-assistant.noarch : Detect and install the best NVIDIA driver packages for the system
nvidia-driver-cuda.x86_64 : CUDA integration for nvidia-driver
nvidia-driver-cuda-libs.i686 : Libraries for nvidia-driver-cuda
nvidia-driver-cuda-libs.x86_64 : Libraries for nvidia-driver-cuda
nvidia-driver-libs.i686 : Libraries for nvidia-driver
nvidia-driver-libs.x86_64 : Libraries for nvidia-driver
nvidia-driver-local-repo-rhel9-570.124.06.x86_64 : nvidia-driver-local repository configuration files
nvidia-imex-570.x86_64 : NVIDIA LIBRARY
nvidia-kmod-common.noarch : Common file for NVIDIA's proprietary driver kernel modules
nvidia-modprobe.x86_64 : NVIDIA kernel module loader
nvidia-open.noarch : NVIDIA Driver meta-package
nvidia-persistenced.x86_64 : A daemon to maintain persistent software state in the NVIDIA driver
nvidia-settings.x86_64 : Configure the NVIDIA graphics driver
nvidia-xconfig.x86_64 : NVIDIA X configuration file editor
pcp-pmda-nvidia-gpu.x86_64 : Performance Co-Pilot (PCP) metrics for the Nvidia GPU
xorg-x11-nvidia.x86_64 : X.org X11 NVIDIA driver and extensions
=================================================================== Name Matched: nvidia ===================================================================
dnf-plugin-nvidia.noarch : DNF plugin needed to remove old kernel modules
libnvidia-nscq-570.x86_64 : NVSwitch Configuration and Query library
nvidia-fabric-manager.x86_64 : Fabric Manager for NVSwitch based systems
nvidia-fabric-manager-devel.x86_64 : Fabric Manager API headers and associated library
nvidia-libXNVCtrl.x86_64 : Library providing the NV-CONTROL API
nvidia-libXNVCtrl-devel.x86_64 : Development files for libXNVCtrl
================================================================= Summary Matched: nvidia ==================================================================
egl-gbm.i686 : Nvidia egl gbm libary
egl-gbm.x86_64 : Nvidia egl gbm libary
egl-x11.i686 : NVIDIA XLib and XCB EGL Platform Library
egl-x11.x86_64 : NVIDIA XLib and XCB EGL Platform Library
nv-codec-headers.noarch : FFmpeg version of Nvidia Codec SDK headers

Refer