Comment
Author: Admin | 2025-04-27
This does not impact any other operation of the system.Remote loginBoot the system, login on the terminal, and note the IP address (“inet”) of the LAN interface:ip addrNext, we use our terminal from MacOS and ssh to login and configure the system remotely. (Delete the line with the IP address of the new system from the ~/.ssh/known_hosts file if the IP address was used before with another server and LAN interface.)CUDA 9.1 InstallationCUDA 9.1 (9.1.85-1) with driver version 387.26 (which does not seem to display consumed power anymore though for 1050Ti cards) is available and it seems to work with the other software we plan to use so we can stop using CUDA 8.0 (we used 8.0.61-1 before). For additional reference, NVIDIA instructions for CentOS/RHEL are here. The archive with all version is here.(You can view available software version with sudo yum –showduplicates list and specifically install a particular version by adding version to package name.)[ ] Download RPM(local) version of CUDA9.1 from NVIDIA here to your remote PC[Linux][x86_64][CentOS][7][rpm (local)][Download][ ] Copy the file from your remote PC to the server, typing on the remote PC (MacOS)scp cuda-repo-rhel7-9-1-local-9.1.85-1.x86_64.rpm @:/home//[ ] Login to your server (assuming your user id on remote and server are the same)ssh [ ] Install rpmsudo rpm -i cuda-repo-rhel7-9-1-local-9.1.85-1.x86_64.rpm[ ] Install epel-release repository (tested version 7-9) and dkms (version 2.4.0-1)sudo yum install epel-release -ysudo yum install dkms -y[ ] Install kernel headers and dev packages (our kernel was 3.10.0-693.el7.x86_64)sudo yum install kernel-devel-$(uname -r) kernel-headers-$(uname -r) -y[ ] Install cuda-9.1.85-1 (or just cuda if you only have one version available)sudo yum install cuda-9.1.85-1 -yBacklisting Open Source driversNext we will have to blacklist the open source Nouveau driver.[ ] Edit /etc/modprobe.d/blacklist-nouveau.confsudo vi /etc/modprobe.d/blacklist-nouveau.conf[ ] Add to the file so that it includes the followingblacklist nouveaublacklist lbm-nouveauoptions nouveau modeset=0alias nouveau offalias lbm-nouveau off[ ] Save and exit[ ] Backup the current initramfssudo mv /boot/initramfs-$(uname -r).img /boot/initramfs-$(uname -r)-nouveau.img[ ] Create the new initramfs (in Ubuntu it is sudo update-initramfs -u)sudo dracut /boot/initramfs-$(uname -r).img $(uname -r)[ ] Rebootsudo reboot[ ] Login and try if this lists the status of the GPU(s)nvidia-smiUseful UtilitiesScreenInstall
Add Comment