Yep, its a subset of rpmfusion’s repo. It’s exactly the same driver.
The article itself is more beginner-friendly and explains several things for new users. On the other hand, rpmfusion’s how-to contains additional things not covered by the article. They complement each other quite nicely: seasoned users will take commands they need from how-to, new ones can read detailed instructions with additional background info from
They explain this in the article and point the way to install driver using Gnome Software with screenshots.
They were not errors per say, some unnecessary actions that can confuse new users. The most obvious ones were pointed out in the comments, maybe even I added my two cents
You’re right, of course. If remember correctly and judging from my own experience with Nvidia cards/drivers I consider the instructions given in this exact article release-agnostic enough for them to work well. And as I said authors give some background information which is useful for understanding of the process regardless of Fedora release.
I’m not very qualified to talk about this topic, my understanding is neither very deep nor very extensive. With this caveat this is how I understand the topic.
The X.Org server is very old, was modified and extended heavily. It’s a huge codebase that’s hard to maintain and to develop further. It contains quite a number of features that were included at some point and aren’t used anymore (but can be hard to remove). Also due to it’s age it wasn’t designed with modern security considerations in place. And you do understand that trying to add security features aftewards is not the same as including them in the original design.
Wayland aims to resolve all this issues, it’s modern, designed with security in mind from the very beginning, with simpler and cleaner codebase. From what I heard pretty much everyone in the industry agrees that it’s the way forward.
It’s still not ready for the prime time, but it’s getting there. It still doesn’t implement all the features the developers want. It is default in Fedora for two to four releases now (I don’t remember which Fedora was the first to make it default, it’s easy to find). I’m using it without any major issue.
Still one more thing to note is that majority of GUI applications (especially older ones which aren’t actively developed) are still made for X.Org only and don’t support Wayland natively. XWayland – sort of compatibility layer – is used to run them under Wayland now. It all works pretty well, but you can encounter some unexpected behavior, especially when using some older or niche/specialized software that can be untested with Wayland.
Generally speaking imaginary “ordinary user” won’t see the difference between X.Org session or Wayland session. But the technology underneath is very different indeed )
Also I think I need to say that X.Org can be more stable, Wayland is still sort of experimental. Most distributions don’t use it still. But in Fedora it works quite well, and it’s interesting for me as the tech of the future)
@biosharkdev I agree that reducing the number of third party repos is a good idea. What specific issues/concerns are there with negativo17? One thing I like about it is the ease of installing CUDA. Otherwise, I think sticking to RPMFusion might be good if there are no issues.
Unfortunately, one big downside with nouveau, aside from good 3D support and graphics glitches, is the lack of re-clocking support on recent hardware:
there will still be the issue like with Maxwell / Pascal / Volta of only running at the boot clock frequencies without any re-clocking support for being able to drive the hardware at its optimal clock frequencies. For overcoming that challenge, additional firmware support or workarounds need to be devised around the PMU handling. Until that happens, the Nouveau performance past the GeForce GTX 700 series remains very slow.
I’ve uninstalled my Nvidia driver and installed nouveau. And I’m on Wayland.
[xxx@wasteland ~]$ loginctl show-session 2 -p Type
[xxx@wasteland ~]$ lsmod | grep nouveau
nouveau 2248704 12
i2c_algo_bit 16384 1 nouveau
drm_kms_helper 212992 1 nouveau
mxm_wmi 16384 1 nouveau
ttm 114688 1 nouveau
drm 491520 13 drm_kms_helper,ttm,nouveau
wmi 36864 3 intel_wmi_thunderbolt,mxm_wmi,nouveau
video 49152 1 nouveau
So far everything works ok except 2 things:
dark background pages in Firefox seem to flicker vaguely (one of this pages is ask.fedora.org because I use the dark theme).
my terminal which had a bit of transparency had the vague flickering as well, however by decreasing some of the transparency from the preferences seemed to remove the problem.
Do I need to install firefox-wayland package? It’s strange to think about having 2 types of firefox
If I install the wayland version should I uninstall the default one?
Thanks for the info, that’s interesting.
However, if one doesn’t use 3D applications that would require heavy graphics processing, why would this matter? Unless the slow performance is felt in the day to day use of Gnome shell.
At any rate, for now I’ll give nouveau a try. If I’ll decide to go back to the proprietary drivers I’ll install them from RPMFusion. By the way, they also have CUDA suport: dnf install xorg-x11-drv-nvidia-cuda
Without re-clocking, either you run the hardware at low frequency which might falter at higher loads, or force running higher frequency all the time, which generates more heat and possibly noise. It’s a pretty terrible choice either way.
If you don’t have much need for graphics acceleration, the low frequency option might be good enough, but it makes having that hardware rather pointless.
It’s confusing because this page mentions that CUDA package, but this page instructs to install repo rpms from nvidia’s site.
It’s no secret that nvidia has unfortunately been hostile to open source. Except for the least demanding use-cases, the proprietary driver seems to be the only realistic option.
This is a good summary:
The main blocker for making the open-source NVIDIA driver viable for Linux desktop users and gamers though is re-clocking support for newer generations of hardware… With the GeForce GTX 900 Maxwell series and newer, there isn’t yet any re-clocking support so graphics cards are stuck to operating at their boot frequencies, which generally is quite low compared to their rated base/boost clock frequencies. Until NVIDIA releases the signed PMU firmware or the Nouveau developers achieve a workaround, any GPUs newer than the GTX 600/700 Kepler or GTX 750 Maxwell series is a no-go if you want decent performance. It’s not known if/when a solution will be in place for better supporting these newer generations of NVIDIA GPUs.
Thus for now the best Nouveau open-source driver support remains with the GTX 600/700 Kepler series since at least there the graphics card can be manually re-clocked by writing a value to DebugFS… Sitll no automatic/dynamic re-clocking, but at least users can force their Kepler (and GTX 750 Maxwell1) parts to the rated frequencies.
Exactly. Nouveau’s performance is enough to run modern Linux desktop. It isn’t good for gaming, that’s a given.
You’re right here. Many Linux developers and so to say more hardcore users buy laptops / computers with integrated Intel graphics as it’s open-source support is second to none.
Also they say that AMD GPUs in the last several years have made huge progress with their open-source drivers.
All the troubles/questions with NVidia are mainly from the people who already has/had hardware with NVidia GPUs, mainly migrating from Windows. Or it’s work-provided and user don’t have the choice ) Seasoned Linux users usually are usually aware about these problems and if they use NVidia GPUs they do it by choice with understanding all the concerns. I think
This is what I mean, essentially this is my case too ))
To summarize the topic a bit
my humble opinion is this:
If you already have NVidia hardware and need one or more of the following:
3D performance for gaming or rendering
accelerated video encoding/decoding
CUDA for hardcore scientific computations
this list can be expanded but nothing concrete comes to mind right now
– then you have no option but to use proprietary NVidia drivers. In most cases you’re better off using packaged version from RPMFusion (though you have some other options available as we’ve discussed).
If you don’t need anything from that list, you need just stable desktop workloads, and if nouveau works well on your GPU (it works better on some GPUs than on others) – then you’re better off using nouveau.
Nouveau’s performance is absolutely totally not comparable to proprietary driver, there’s no question about this. Keep in mind that it’s totally open source and developed by community without much support from (evil!) NVidia. This is also why it can’t support all the numerous NVidia’s GPUs equally well.
On the other hand proprietary driver is widely known to cause wide variety of issues, NVidia’s developers don’t want to include support for many new features required by modern Linux system/desktop. At least with some of these issues nouveau’s support is better – resulting in more smooth and stable Linux desktop experience. Again, at least on the GPUs that work well with nouveau.
Ah, I see. Then I guess I’ll stick with negativo for now. Even if not developing CUDA code, applications often require compilation.
They have, and I hope ROCm (AMD’s answer to CUDA) becomes a great success, but sadly CUDA is the standard right now for AI/ML. ROCm installation seems complicated, hardware support is limited, and performance lags behind CUDA, although the progress has been good to see. With Intel jumping into discrete graphics, one can hope that the open source driver situation will improve or at least kick nvidia into collaborating better (or at all). Maybe it would even help nouveau’s efforts.
Also linux gaming is improving and you have several options, whether natively or WINE or VMs. But of course it’s back to the proprietary driver for that.