Open vs. Proprietary NVIDIA Drivers

I’m looking some feedback as to why one would choose one NVIDIA driver set—i.e., open or proprietary—over the other; a quick search suggests back in the Fedora 28 days it could make a difference, but I can’t seem to locate any more recent discussion.

Thanks, all!

1 Like

There is a substantial performance difference if you are doing anything gpu intensive.


If you use the proprietary drivers from nVidia, you have to remember to install them again every time the kernel is updated. If you use the Open Source ones from rpmfusion, it’s “fire and forget” because the akmod service will take care of doing what’s needed for the new kernels to use them the first time you reboot after the kernel gets updated.

1 Like


There are three sources for NVIDIA graphics card drivers.
As sideburns communicated, drivers are available from NVIDIA corporation. Closed source drivers from “rpmfusion” repository are linked to fedora’s software repository and automatically update with fedora software update procedures. “Closed” or “non-free” means if NVIDIA has protected the intellectual property used to create the driver software, the copy you have is restricted by license terms you must legally comply with.

A third team has been writing “free” Open versions of drivers, this is nouvoeau group also available through package management. (I may have used creativity in spelling nouvaeu :()

If you only install one driver, you will not be confused which one is actually being used.

In my case, with Nouveau my system can’t poweroff. It’s weird since I can suspend and reboot. But I guest there must be something unique/peculiar on my old laptop (or even broken?) that not work well with Nouveau.

For some users, mostly with desktop GPUs the Open drivers “work”. But for better utilization of hardware and performance, power management, thermal management etc. the proprietary drivers work. For developers who need CUDA for their work, proprietary drivers are the only solution. Basically as of now Open drivers are just drivers used by people to boot up into a working system to install Non-free drivers.

For fedora, RPM Fusion automates the process of installing the nvidia non-free drivers which is often a bit outdated because they test and ensure it works which is same for most distributions. If user installs drivers manually from the nvidia’s .run package, then the user will have to manually do it every update.

1 Like

Thanks for all the insights, folks; it’s much appreciated.

Besides gaming, what would y’all qualify as “GPU-intensive?” That is, would videoconferencing count?

1 Like

Video conferencing is not very GPU-intensive, but crypto mining (e.g. bitcoin mining) is and miners use the GPU to do it.


GPU intensive could be considered the realm of high performance rendering of 3d images or scenes for example. Digital currency mining doesn’t have to be the use case, but it sure drove the cost of GPU cards up, so theres that.


video-conferencing should not be gpu intensive.

Anything which uses cuda, opencl or does 3d rendering would be though.

So, gpu accelerated photo editing for example.

Ultimately, if the open source drivers are meeting your needs, I wouldn’t worry about it. If you start having performance or compatibility problems, you can switch to the proprietary drivers.


Model training with CUDA, Blender rendering with CUDA or Other 3D accelerated applications like games. Generally for Nvidia, having non-free driver is better. For AMD, having Free drivers are better as long as you don’t have any professional workload requirements.


Thanks everyone for the input… I have a relatively capable laptop for the first time ever because I always used to take the “old” ones that struggled with Windows and gave them new life with Linux, so I’m used to trying to eke out every little performance enhancement. It sounds like I don’t really need to do that right now. :wink:


I do a lot of photo editing so it make sit faster and more responsive. The light gaming I do I havent seen a difference. The main thing is I have the dreaded Nvidia/Intel Optimus laptop with external monitor setup. I have noticed running the novuea/intel driver set causes the cpu to run about 10 degrees celsius hotter. The Nvidia/Nvidia driver setup corrects this.

This comment has never made sense to me. I have the dual intel/nvidia config and from the very beginning it has worked perfectly for me.

I do make a habit of researching before I leap in, and made it a point to follow the recommendations from rpmfusion for the configs, including setting the nvidia card as primary so it runs both the internal and external monitors. That involved copying the file /usr/share/X11/xorg.conf.d/nvidia.conf to /etc/X11/xorg.conf.d/nvidia.conf and editing the latter file to make the nvidia card primary.

I try to avoid anything from Nvidia, but if i had an active Nvidia card, i would use Nouveau again. Why? Ideology, security, distro integration, likely in that order. I could ignore the performance difference until the card was replaced with AMD and FOSS driver.

1 Like

He is comparing Open vs Nouveau. Nouveau is bad for Optimus.

Agreed it works out of the box on Fedora with nouveau drivers… And the Nvidia driver setup is very easy. But 10deg celsius temp increase is the big issue for me. Thats why I use the proprietary drivers.

We’ve mixed something up here, no? The akmods from RPM Fusion are also the proprietary drivers from nVidia, just packaged up nicely to work well with Fedora kernels. They’re not Open Source (i.e., they’re not nouveau which lives happily in the Fedora repositories).

Nouveau generally lags the proprietary driver because nVidia provides no documentation on it’s GPUs at all. So the nouveau developers reverse engineer to obtain the required data from devices. I don’t know if this has changed recently. Wikipedia says this is still the case:


Open source drivers may lack some features depending on which card you have:

1 Like

No. The drivers are closed source, but the akmods that customize them for the new kernel aren’t, and that’s what I was referring to.

1 Like