Centos7 / RedHat7 install NVIDIA-TeslaP4 graphics driver manual
Linux systems are generally installed by default nouvea is open source graphics drivers, which conflicts with nvidia graphics driver, must first want to install nvidia ban nouvea!
Close secure boot (1 Pit) - modified by the system unit is completed, the default is set
- Restart the computer, enter the BIOS
- Find seure boot, usually in the boot options
- The "Enable" was changed to "Disable"
Detected NVIDIA drivers
Under the root user, run the following command to see if N normal identification card hardware
lspci|grep -i nvidia
-i parameter is ignored behind keyword search case
View the system kernel version
# uname -r
3.10.0-693.el7.x86_64 ; 不同操作系统的内核版本会不一样,记住它后面有用
# df ; 确认 boot 目录的空间不少于 300 MB
It should be [Special Note]: The company is the default of 957 kernel, you need to find a system unit will fall back to the kernel 693 **, or not installed on the N card driver
Note here to choose Linux-64-bit .
Disable the Nouveau driver (Pit 2)
-
Check whether nouveau run: lsmod | grep nouveau as just need to disable the run
-
Modify /lib/modprobe.d/dist-blacklist.conf file
The nvidiafb commented:
#blacklist nvidiafb
Add two lines
blacklist nouveau
options nouveau modeset=0
-
Modify two nvidia-installer-disable-nouveau.conf file
# vim /etc/modprobe.d/nvidia-installer-disable-nouveau.conf # vim /lib/modprobe.d/nvidia-installer-disable-nouveau.conf ... blacklist nouveau options nouveau modeset=0
-
Modify / etc / default / grub file
Add in the GRUB_CMDLINE_LINUX
rd.driver.blacklist nouveau nouveau.modeset=0
- Rebuild initramfs image
# mv /boot/initramfs-$(uname -r).img /boot/initramfs-$(uname -r).img.bak
# dracut /boot/initramfs-$(uname -r).img $(uname -r)
Restart (non-essential)
# systemctl set-default multi-user.target
# init 3
# reboot
Installation build environment (hole 3)
Build local yum source reference https://blog.csdn.net/Sunny_Future/article/details/78420508 or own search network yum source, due to the company's internal and external network isolation, yum source from system configuration section , which are not otherwise gcc installed successfully.
Run the following command
yum install gcc kernel-devel kernerl-hearders -y
It should be noted that the version of the kernel-devel yum install successful, to be consistent with the previous system kernel, if still high version of the 957 kernel, you need to download the installation package kernel-devel-3.10.0-693.el7.x86_64.rpm performed rpm installation, reference http://rpm.pbone.net/index.php3/stat/4/idpl/37924679/dir/scientific_linux_7/com/kernel-devel-3.10.0-693.el7.x86_64.rpm .html
Download and install.
You must first remove the high version of the 957 kernel-devel, and then install the 693 version, execution
#rpm -qa|grep kernel-devel
#rpm -e kernel-devel-3.XX.X-957.el7.x86_64.rpm
#rpm -ivh kernel-devel-3.10.0-693.el7.x86_64.rpm
Install NVIDIA drivers
#chmod +x NVIDIA-Linux-x86_64-440.33.01.run
#./NVIDIA-Linux-x86_64-440.33.01.run --kernel-source-path=/usr/src/kernels/3.10.0-693.el7.x86_64 -k $(uname -r)
OK all the way and ACCEPT, Install NVIDIA's 32-bit compatibillty libraries? Select Yes
N card driver final verification is performed
# nvidia-smi
postscript:
1. Before downloading the driver package is the official website of nvidia-driver-local-repo-rhel7-440.33.01-1.0-1.x86_64.rpm, seemingly wrong.
2, Cuda installation package Download: https: //developer.nvidia.com/cuda-downloads target_os = Linux & target_arch = x86_64 & target_distro = RHEL?
There are two types, namely runfile (.run) file and the file rpm recommended .run package, if you install only CUDA, do not select the package in which the driver installation options to install.
sh cuda_10.2.89_440.33.01_linux.run
rpm -ivh cuda-repo-rhel7-10-2-local-10.2.89-440.33.01-1.0-1.x86_64.rpm
reference:
1、https://blog.csdn.net/sunny_future/article/details/83500788
2、https://note.youdao.com/ynoteshare1/index.html?id=83c6d9db77126025afd1ef93b2076ced&type=note
3 https: //www.cnblogs.com/yangsichao/p/8026234.html