Wednesday, May 20, 2020

How to install Cuda 10.x in CentOS 7



Hardware configuration
B7119
BIOS v2.02.B10
BMC v7.0
CPU: Intel Xeon Platinum 8268 2.9GHz x2
RAM: DDR4 32GB x24 (Micron 36ASF4G72PZ-2G6D1)
GPU: Leadtek RTX-2080ti (Device ID 1E07)
OS: CentOS 7.7




Process
1. Install CentOS 7.7
     Hint: install Server with GUI with Compatibility Libraries and Development Tools.
2. Install OpenSSH (option)
     #yum install openssh*
3. Disable Nouveau
     a. #vi /usr/lib/modprobe.d/dist-blacklist.conf
     Add below
     blacklist nouveau
     options nouveau modeset=0
     
     b. #mv /boot/intramfs-$(uname –r).img /boot/initramfs-$(uname –r).img.bak
     c. #dracut /boot/initramfs=$(uname –r).img $(uname –r)
    
     d. #reboot
4. Boot to text mode
     a. #systemctl get-default

     b. #systemctl set-default multi-user.target

     c. #reboot
5. Install cuda
     a. Download and copy Cuda to Linux
     b. #./cuda_10.2.89_440.33.01
c. Type Accept. Press “Enter”.

d. Move light bar to Install. Press “Enter”

e. Installation process finished.

f. Export Library

6. Type nvidia-smi to check installed GPU cards.

7. Type nvidia-smi –q to query Device ID and Link Width



No comments:

Post a Comment

How to fix gpu_burn compiler failure issue

System Environment: Ubuntu 22.04 LTS Server CUDA v12.0 GPU: RTX-4080 (driver 525.85.05) AP: GPU_Burn v1.1 Symptom: met error in make gpu_bur...