Support us on Patreon to keep GamingOnLinux alive. This ensures all of our main content remains free for everyone. Just good, fresh content! Alternatively, you can donate through PayPal. You can also buy games using our partner links for GOG and Humble Store.
We do often include affiliate links to earn us some pennies. See more here.

Jason Ekstrand of Collabora just announced NVK, a brand new open source Mesa Vulkan driver for NVIDIA GPUs. No you're not dreaming but it's also not from NVIDIA directly — as usual the open source community is doing the work.

It's still early-ish days for NVK but what they've been able to do with it so far sounds reasonably promising. The current nouveau drivers for NVIDIA in Mesa aren't going to be removed but they don't exactly work well, and a lot of it was done coding in the dark with reverse engineering.

Things have come a long way though, with NVIDIA gradually releasing more documentation and recently their open source Linux GPU kernel modules.

NVK itself was written "almost entirely from scratch using the new official headers from NVIDIA" over the last several months targeting Turing+. Ekstrand has some pretty high goals for it, like wanting it to become "the new reference Vulkan driver within Mesa" and so they're "building NVK with all the best practices we've developed for Vulkan drivers over the last 7.5 years and trying to keep the code-base clean and well-organized".

This is a serious effort to get a good open source driver for NVIDIA GPUs and that sounds awesome. Eventually it could become "what RADV is to AMD hardware" but there's a lot of work to get there.

As for OpenGL support, they might even end up using the Zink driver (an OpenGL implementation on top of Vulkan), instead of writing a whole new OpenGL driver to go along with it. Zink has come a long way too so it could be a pretty good plan.

See the blog post for all the info.

Article taken from GamingOnLinux.com.
41 Likes
About the author -
author picture
I am the owner of GamingOnLinux. After discovering Linux back in the days of Mandrake in 2003, I constantly checked on the progress of Linux until Ubuntu appeared on the scene and it helped me to really love it. You can reach me easily by emailing GamingOnLinux directly. You can also follow my personal adventures on Bluesky.
See more from me
The comments on this article are closed.
All posts need to follow our rules. For users logged in: please hit the Report Flag icon on any post that breaks the rules or contains illegal / harmful content. Guest readers can email us for any issues.
17 comments

Corben Oct 4, 2022
Woah
Well, one step further with better implementations for nVidia users. If it manages to get on par feature-wise (e.g. supporting nvenc, async reprojection for VR, etc.) with the proprietary driver, heck yeah! If nVidia would only go the same way AMD did with their support for the open source drivers and community...
fagnerln Oct 4, 2022
This is nice! I hope that they reach an usable state soon™.

But what I find interesting in this post is about the usage of Zink. Zink is amazing dude, it guarantee that we won't have a bad support to OGL in newer cards if they decide to not support old techs on it.
Shmerl Oct 4, 2022
It still needs the kernel driver with reclocking though, right? That didn't get upstream yet.
dpanter Oct 4, 2022
Classic Nvidia "support". Are they going to help out at all this time?
mr-victory Oct 4, 2022
It still needs the kernel driver with reclocking though, right? That didn't get upstream yet.
It needs the GSP (GPU system processor) firmware, which is provided by Nvidia since open kernel modules announcement. So we have reclock support in this driver as well. Open kernel modules aren't upstreamed.
Shmerl Oct 4, 2022
It needs the GSP (GPU system processor) firmware, which is provided by Nvidia since open kernel modules announcement. So we have reclock support in this driver as well. Open kernel modules aren't upstreamed.

Do you mean this firmware already works with current nouveau kernel module or it needs that custom (not upstreamed) open kernel module from Nvidia still?

This is about just the Vulkan implementation, not an announcement of an upstream kernel module? Unless I'm missing something.


Last edited by Shmerl on 4 October 2022 at 6:48 pm UTC
mr-victory Oct 4, 2022
Do you mean this firmware already works with current nouveau kernel module or it needs that custom (not upstreamed) open kernel module from Nvidia still?
Actually, the new firmware is intended to be used by nouveau but instead of that we got a new driver using the firmware, because apparently starting from scratch is easier than refactoring nouveau.
The "kernel module" part is used as reference, not directly.
At least this is what I understood.


Last edited by mr-victory on 4 October 2022 at 6:53 pm UTC
dziadulewicz Oct 4, 2022
Classic Nvidia "support". Are they going to help out at all this time?
No. They are busy with selling $2000 GPUs.

That's exactly a part of the problem. They are busy with it. AMD should get more attention in the market. Well, Intel is there too now with their first gaming GPU efforts.
hummer010 Oct 4, 2022
Turing+ ... this should be pretty mature by the time I can afford a supported GPU!
sarmad Oct 4, 2022
Classic Nvidia "support". Are they going to help out at all this time?

I think it depends on the progress of this one and how promising it is. If it progresses well nVidia will most likely adopt it and start helping, since that will be cheaper and easier for them than maintaining their own closed driver. However, this depends on whether this new driver is able to meet the same performance levels as the proprietary one.
ElectricPrism Oct 4, 2022
These are confusing times.
STiAT Oct 4, 2022
Open source driver based on vulkan using the zink opengl front would be really lovely.

Will be a long shot though, but we knew after nvidia open sourced it we're talking years not month for a proper open source driver. Depends on the backing, if there are companies backing the devs fulltime, that will be a way faster journey (Valve? RedHat?)

I still hope long run Nvidia goes the way of AMD and contributes / invests there too. But they took a long time even opening up a bit (most work seems still behind locked doors, must be a documentation and organizational hell keeping the issues and request and internal development synced)

It's a start, let's see what time brings. I appreciate the effort too.

Nouveau have done a good job actually getting something working, and we should all appreciate that. But that game has changed.


Last edited by STiAT on 4 October 2022 at 10:52 pm UTC
okasion Oct 5, 2022
Sounds too good to be true, but I believe it because competition; Intel really open sourced their drivers for their ARC video cards just 2 days ago, and with all the bad PR Nvidia has been getting lately, I can see they being pressured to do "the right thing" for once if they want to stay competitive.
okasion Oct 5, 2022
Turing+ ... this should be pretty mature by the time I can afford a supported GPU!

But it still includes Turing right? I mean, what will they do, do not support a Geforce GT1080? because that's Turing.
ripper81358 Oct 5, 2022
I never thought that this would happen. However i wonder why such a driver is needed. The most compelling reasons to use an Nvidia GPU are features like NVENC,DLSS and computing support via Cuda or Opti-X. To get accsess to these technolgies the proprietary drivers (at least the userspace parts) are needed anyway.
CyborgZeta Oct 5, 2022
I dislike NVIDIA, and will never buy any of their GPUs.

However, I do think this is good news.
Grogan Oct 5, 2022
I dislike NVIDIA, and will never buy any of their GPUs.

Me too, not only did I hate the hassles (I always build latest kernels from kernel.org) and getting in over my head and having to go try to find third party patches, but every single Nvidia card I've ever bought died sooner than it should have. (even before the following I had an expensive GEForce 6600 GT die just out of warranty)

I remember thinking Charlie Demerjian (writing for The Inquirer at the time) was full of crap, until I started to see the trend for myself both on customer systems and personally. I believed Nvidia over anything that came out of the rectum of The Inquirer (sorry Charlie). It took longer on discrete GPUs with cooling fans etc. but the chipsets were all affected. To make a long story short, the GPUs were experiencing cracks between the connector "bumps" on the chips and the underlying substrate.

I had an expensive dual GPU Nvidia card that was nothing but trouble in games, and it died in less than a year (and BFG went out of business so I burned for it). So I put in another good Nvidia card that spent most of its life in the closet (I was unsatisfied and bought a new one a few months later). I decided it was time for a new build, so I sold it in that PC in good faith, and didn't it fail in about 3 months. The guy wasn't even a gamer. I was pissed off, but I bought him an entry level video card (Radeon 5450 or something like that... $60) out of my pocket and installed it for free. He didn't know the difference with his usage.

That soured me on Nvidia forever. I don't buy anything that has any of their hardware in it, for anyone.

I'm uglier than Linus, so I won't send you a pic of me flipping them the bird :-)


Last edited by Grogan on 5 October 2022 at 7:42 pm UTC
While you're here, please consider supporting GamingOnLinux on:

Reward Tiers: Patreon. Plain Donations: PayPal.

This ensures all of our main content remains totally free for everyone! Patreon supporters can also remove all adverts and sponsors! Supporting us helps bring good, fresh content. Without your continued support, we simply could not continue!

You can find even more ways to support us on this dedicated page any time. If you already are, thank you!
The comments on this article are closed.