A800 GPU服务器 麒麟V10操作系统安装CUDA

CUDA&Fabric Manager安装文档

摘要：驱动、Fabric Manager、CUDA组件的版本需完全一致（主版本和次版本），特别是nvidia-fabricmanager与驱动版本强一致，需要先查看nvidia-smi确认驱动版本后选择对应的下载安装。安装顺序如下：

1、CUDA

系统内uname –a 显示ky10.x86_64

#麒麟v10安装 CUDA，目前NV驱动版本是535，搭配cuda12.2使用

https://developer.download.nvidia.com/compute/cuda/12.2.0/local_installers/cuda-repo-kylin10-12-2-local-12.2.0_535.54.03-1.x86_64.rpmwget https://developer.download.nvidia.com/compute/cuda/12.2.0/local_installers/cuda-repo-kylin10-12-2-local-12.2.0_535.54.03-1.x86_64.rpm

#下载完成后执行

sudo rpm -i cuda-repo-kylin10-12-2-local-12.2.0_535.54.03-1.x86_64.rpm

sudo dnf clean all

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

sudo dnf -y install cuda

#全部执行完后显示complete！

#添加环境变量

export PATH=/usr/local/cuda/bin:$PATH

export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH

source ~/.bashrc

nvcc –version

输出正常，安装结束。

2、Fabric Manager

#查看本机的版本

rpm -qa | grep fabric-manager

#GPU驱动为535.54.03，版本不一致需要卸载

sudo yum remove nvidia-fabric-manager-535.129.03-1.x86_64

bash

# 配置NVIDIA仓库（适配麒麟V10的RHEL8路径）

sudo tee /etc/yum.repos.d/cuda.repo <<EOF

[cuda]

name=CUDA Repository

baseurl=https://developer.download.nvidia.cn/compute/cuda/repos/rhel8/x86_64/

enabled=1

gpgcheck=1

gpgkey=https://developer.download.nvidia.cn/compute/cuda/repos/rhel8/x86_64/7fa2af80.pub

EOF

# 清理缓存并安装匹配版本

sudo yum clean all

sudo yum install nvidia-fabric-manager-535.54.03

#安装依赖包

sudo rpm -ivh nvidia-fabric-manager-535.54.03-1.x86_64.rpm

#启动服务并验证

sudo systemctl start nvidia-fabricmanager

sudo systemctl status nvidia-fabricmanager  # 应显示 "active (running)"

附 NVIDIA CUDA仓库：Index of /compute/cuda/repos/rhel8/x86_64