Fedora 36 Planning to Run Wayland by Default with Nvidia's Proprietary Driver

https://www.phoronix.com/scan.php?page=news_item&px=Fedora-36-NVIDIA-Wayland

by rbanffy • 5 years ago

84 61 5 years ago

axegon_2• 5 years ago

I've stated multiple time that I'm not biblical about open source software and I'll use the proprietary solutions if there is no good open source alternative. However nvidia is a special case(as usual). My laptop has a low end mobile nvidia gpu which is anything but special or demanding(mx250 to be exact). The laptop itself is dual screen(the Asus zenbook duo series) and nouveau worked out of the box, no issues whatsoever. I mean there were some but they were unrelated to the gpu driver. In essence I never really needed to deal with the proprietary drivers. My workstation is a bit different in that regard. The GPU was the one thing in it that I cheaped out on - an oldish quadro model from 2017-18. But again the nouveau drivers worked perfectly and all 3 monitors worked perfectly fine. I did however plan on playing with CUDA a bit and decided to install the proprietary drivers. Big mistake. It took me a whole weekend to recover from the tons and tons of problems that came with them. So in that regard, I think that if a distro offers them out of the box, the chances of wasting so much time in fixing and getting things to work may be a good thing. That said, I'm not entirely sure that forcing people to use a proprietary driver by default in a linux distro is a wise idea. A much better option would be for the users to decide on their own which one they want to use.

ZekeSulastin• 5 years ago

What problems did you even run into that would cause you to lose a whole weekend? If we’re dealing with anecdata, I never had an experience that bad even when having to use bumblebee/primusrun/etc on a laptop; my last desktop Linux install was with an RTX 2060 and it was pretty much painless using X.

Also, it’s not defaulting to installing the driver anyways - as described in the article and its sources, what Fedora is actually planning to do is to change GDM’s udev rules to select the Wayland session by default if the proprietary drivers are running.

axegon_2• 5 years ago

Pff lost track of it all. Probably 15 things went wrong along the way, some obvious and some very subtle problems, iirc I even recompiled the kernel at one stage. I've read that the newer ones(2060 being among them), the process is far less painful. I just checked and I'm on a Quadro M5000 which is from 2015(I found it dirt cheap so I went with it).

rwmj2• 5 years ago

The headline is confusing. The proposal is if you are already using the NVIDIA driver, to default to Wayland instead of Xorg. Previously Xorg was the default with the NVIDIA driver. Non-proprietary drivers should still be the preferred choice for Fedora users, avoiding NVIDIA hardware if at all possible.

The actual change proposal is: https://fedoraproject.org/wiki/Changes/WaylandByDefaultOnNVI...

Discussion: https://lists.fedoraproject.org/archives/list/devel@lists.fe...

brnt5• 5 years ago

Tried Fedora 34 with Intel graphics (multiple machines), Wayland stil reliably caused lockups a few minutes in.

I aint touching that with a 10 foot pole anytime soon.

dralley• 5 years ago

Wayland still reliably caused lockups a few minutes in.

How do you know this (that it was wayland)? I have never had a lockup caused by anything other than swapping, across multiple machines with Intel graphics.

rwmj2• 5 years ago

I use Intel graphics + Xorg + Fedora 35 and it's all rock-solid. (I doubt Wayland would have been any different in that regard, but I need remote X despite certain people claiming that remote X is a useless appendix that no one still uses).

saurik• 5 years ago

Well, maybe the idea is that it is more stable than Xorg... when used with NVIDIA's proprietary driver.

technofiend• 5 years ago

When the pandemic started, Zoom only worked on x.org. They've since added Wayland support but googling for it shows the client is still behind on screen sharing. That may just be a harder problem to solve in Wayland without zoom itself acting as a compositor?

https://support.zoom.us/hc/en-us/articles/201362153-Sharing-...

Linux sessions utilizing Wayland can only share an entire desktop or whiteboard. In order to share just a specific application, you will need to launch your Linux session with Xorg instead.

rurban• 5 years ago

Working since FC32 with an AMDGPU on Wayland fine. Much better than my previous Nvidia/Debian desktop.

I wont touch any nvidia nor x.org with a 10ft pole anymore.

1_player3• 5 years ago

I'm on F35, RX 6800 XT @ 4k and Wayland hasn't caused any issue yet. Gaming works great as well. I love how smooth it feels.

bravetraveler3• 5 years ago

I was very worried that this 'only free' principle was being tossed out the window, thank you for clarifying

pxc2• 5 years ago

Of course, yeah. :)

Are you a Fedora dev? Do you have an impression that NVIDIA is getting to be a bit more cooperative with the desktop Linux development community these days, like with the changes related to GBM that made this decision possible for Fedora?

Conan_Kudo3• 5 years ago

Yeah, they definitely are. Though they're still not cooperative enough to make it so NVIDIA hardware has a good basic level of performance with the OSS drivers. :(

While this Change currently describes the effort for GNOME, KDE Plasma is likely to also be ready for Fedora Linux 36 for this. There a couple of other blockers to deal with, but hopefully those are squared away in time...

sho_hn2• 5 years ago

I am not so sure that GBM support was motivated primarily by the Linux desktop community. It's rather a symptom of industry-wide trends.

I work in the automotive industry, where we buy high-end, high-margin SoC SKUs - not a lot of them (compared to phones, primarily), but enough to have some sway. Within the industry, most OEMs are moving toward inhouse platform-shaped development and have a desire to make their code increasingly reusable and hardware-agnostic, as well as lower integration costs for third-party app content. This drives the adoption of open source commodity stack technologies and standards like Wayland, and means hardware vendors now have to demonstrate or offer support for them - with a single code path. It then replaces vendor-proprietary stuff like nVidia's EGLStreams previously used for similar cross-process compositing use cases as Wayland is now. In the past, vendors were asked to solve the problem for the OEM as they saw fit on their hardware.

I've personally requested vendors to provide GBM support on the Orin SoCs by writing it into requirements. I know nVidia previously started incorporating limited GBM support in their GPGPU product line, probably due to similar customer asks in that industry, but I don't know it well.

For the Linux desktop community the takeway is that it has a massive impact beyond just the Linux desktop - by being a leader in the commodification of technologies everyone just ends up adopting, especially as embedded development trends toward a similar complexity level as desktops and phones with similar feature requirements. Accordingly it deserves support from those who have these technologies in their supply chain.

neverminder• 5 years ago

I've heard a rumor the main reason for Nvidia's reluctance to open source their driver is that they use it to enable/disable features. So the same board can be GeForce or Quadro, depending on what the driver tells it to be. Is this true?

trzeci• 5 years ago

It's an unconfirmed rumor.

Also I'd be skeptical to some extend on this. Maybe back in time when a chip fail ratio was lower was a case. But now, there is a quite measurable amount of chips not passing requirements for a high tier cards, but can be used in the lower models (with limited amount of cores etc or worse thermal efficiency).

tyingq2• 5 years ago

It seems plausible given stuff like this:

"A developer driver inadvertently included code used for internal development which removes the hash rate limiter on RTX 3060 in some configurations," the company said in a statement. "The driver has been removed."[1]

So it's hard to tell when it's binning, and when it's hobbling.

[1] https://www.theverge.com/2021/3/16/22333544/nvidia-rtx-3060-...

bravetraveler3• 5 years ago

Probably fairly true. Nvidia tries to differentiate their offerings a LOT with software driven features

I feel a lot of this comes down to them using closed source software that they either don't have the code to, or aren't privileged to distribute. Libraries are a common spot for this to appear

bserge• 5 years ago

Yeah, so GeForce/Quadro used to be a software only difference. Later on (around Fermi/Kepler) it became a hardware difference but cards could still be modded by means of board straps and package traces.

Kepler could literally have all its cores enabled with a simple bridge on one laser cut. That's fun but very hit or miss, the more cores are disabled the higher the likelihood they're actually defective so it was binned lower.

Maxwell can have their Hwid modded but performance will be lower than actual Quadros.

Newer cards have an even bigger performance difference, more hardware differences I think.

Tomte• 5 years ago

Wouldn't it be fun if NVidia could sell a "firmware upgrade" for your GPU, in this time where GPU prices are through the roof and they cannot supply the market adequately?

Of course, I'd hate that in normal times, but if they now made me an offer to "upgrade" my 1060 (to moderately higher performance, but still lower end), I'd probably take it.

bitwize• 5 years ago

Maybe they should take a cue from Intel and implement on-chip DLC?

Const-me2• 5 years ago

Very likely.

AFAIK there’re only 4 differences between GeForce and Tesla/Grid: more and faster VRAM, more FP64 GFlops on most GPUs, unlimited count of nvenc encoding streams versus 2 on GeForce, and support for virtualization. Only the first two are hardware differences, the other 2 are software.

Another thing about virtualization, it sounds ridiculous but nVidia charges non-trivial per-user fee to allow people to use the hardware they already own. Specifically, for virtual workstation they want $450 per concurrent user. That’s on top of the $3000-$12000 purchase price of these GPUs.

moondev• 5 years ago

Wow I did not know this. I've been thinking about picking up an Nvidia card to play with vgpu on vsphere - so it sounds like this would be impossible without an additional license fee?! What a bummer.

rbanffy• 5 years ago

fee to allow people to use the hardware they already own.

IBM does that with mainframes, but, at least, it’s hardware they own that they shipped in the box for you to rent out when needed or to buy and enable.

xg15• 5 years ago

Dumb question: Closed-source doesn't mean encrypted. What stops people from disassembling and reverse-engineering the blob?

Of course, distributing a "cracked" driver would breach licensing agreements and violate copyright (and possibly other laws), so it's out of the question for any kind of professional use. However, if Nvidia uses it to differentiate services, wouldn't there be a significant risk of private consumers (gamers, hobbyists, etc) using such a driver to "upgrade" their cards for personal use?

NtGuy25• 5 years ago

They have. There are driver patches that allow unlimited encoding streams and other "enterprise" features.

bravetraveler3• 5 years ago

That's actually how we got nouveau, as I understand things...

The firmware itself truly is signed and so on

1_player3• 5 years ago

AFAIK it is both closed source and encrypted. Otherwise we would already have nouveau with black market firmware extracted from the official driver than unlocks power and clock management (something that nouveau can't do, reason why it's so slow to be useless on modern cards)

invokestatic• 5 years ago

I’ve been “daily driving” F35 with Wayland on Nvidia for the last couple weeks and it’s been an… OK experience. The only real show stopper for me now is that GNOME night light doesn’t work (Nvidia issue) and screen sharing in some select applications. I’ve worked around the former by setting the color temperature on the monitor itself.

Another big deal is there’s no hardware video decode support in Firefox. I don’t have integrated graphics, so I rely on Nvidia’s VDPAU implementation that doesn’t work on Wayland. This really stinks!

I’m really on the fence about it and I keep thinking of switching back to Xorg.

my1233• 5 years ago

so I rely on Nvidia’s VDPAU implementation that doesn’t work on Wayland

NVIDIA’s VDPAU successor seems to be Vulkan Video. So those issues are going away at some point in the future at least.

elFarto• 5 years ago

Regarding the video decode support in Firefox, I've got a working solution to that (a VA-API implementation backed by NVDEC). Just waiting for some Firefox patches to land to release it.

No sure it works on wayland though, not tested it there.

BoysenberryPi• 5 years ago

Tried Wayland recently on Fedora 35 with an AMD GPU and found that it causes issues with a lot of Proton games that I don't have on X.org. So, Wayland is still a "no thanks" for me.

2OEH8eoCRo0• 5 years ago

Same. I had weird issues when playing Diablo II: Resurrected with menu items not being clickable or other jank that was not present with xorg. It's almost there though. I'm excited when features like this are pushed as defaults because it means more feedback, more pressure to fix remaining issues.

1_player3• 5 years ago

I have been gaming with AMD on F35 with some demanding stuff, no issues here. Proton works great.

derekjdanserl• 5 years ago

Anyone know if Nouveau can operate in integrated-only mode? And if so how can that be confirmed? The only reason I use Nvidia drivers is to use integrated-only graphics mode.

btdmaster• 5 years ago

Something like[1]?

[1] https://wiki.archlinux.org/title/NVIDIA_Optimus#Use_Intel_gr...

Fedora 36 Planning to Run Wayland by Default with Nvidia's Proprietary Driver

Related Stories