Step 1: Installing the latest NVidia Driver Upon your next reboot, your system should be up to date in everything.
If you run that on a brand new install, the upgrade may take a while, and download /install a lot of packages.
#NVIDIA CUDA DRIVER UBUNTU UPDATE#
Once you’re root: update, upgrade, and reboot: apt -y update (from now on, I’ll omit the and only list the commands, because that’s easier to copy-n-paste) I usually chose “normal install” with “install 3rd party drivers” enabled, but that shouldn’t matter much.įirst, let’s make sure your machine is up to date: Log into your machine, open a terminal, and become root:
#NVIDIA CUDA DRIVER UBUNTU INSTALL#
If not, feel free to grab a USB image from, put it on a USB stick, and install it. I assume you already have a working Ubuntu system.
Step 0: Make sure your Ubuntu box is up-to-date. However, the steps in this post should – with small modifications (such as using yum instead of apt when using CentOS) – also work on other distros.
This post started as a google doc with notes to myself (yes, I did install rather a lot of different machines, recently 🙂 ) then it evolved into documenting this procedure for others and finally led to the thought of “let’s just post it publicly, to whomever it may be useful”… so here we are.īefore saying anything else, one note on Linux versions: I personally prefer Ubuntu, and unless you’ve got a strong preference for another distro can only recommend Ubuntu 18, if only because it’s the only distro with which I managed to get two of my “integrated graphics plus nvidia mobile GPU” based laptops to work without having to use things like Bumblebee (shudder). In this article I’ll provide some step-by-step instructions to get all of the software stack required for CUDA/OptiX installed on a Ubuntu box, hopefully serving as a copy-n-paste blueprint for whoever else wants to do that, too …. Having recently switched to NVidia I now – rather obviously – spend a lot more of my time coding in CUDA and OptiX and one of the first things I noted is that getting the right CUDA/OptiX software stack on linux isn’t always as automatic as one would have hoped for.
0 kommentar(er)
