Fedora actually seems to be the sweet-spot for near rolling-release packages and stability these days.

I recently switched from Manjaro to Fedora and so far I see no reason to go back to either Arch based or Debian based distros.

Last edited by Julius on 1 Mar 2023 at 8:43 pm UTC

Quoting: ShmerlI use Debian testing. But there are a few things I build or maintain outside of Debian repos:

* Kernel itself (I usually build newest release, testing gets kernels with quite some delay).
* Firmware from upstream kernel repo (for AMD GPU). This one is updated in Debian very rarely.
* Mesa main for gaming (not replacing system Mesa).
* Wine from WineHQ Debian repo.
* vkd3d-proton / dxvk master from upstream.
* Firefox beta from Mozilla Linux build.

Other than that, I use what's in Debian repos.

Sadly the first half of these are why I stick with nvidia.

But I also tend to use Debian Sid.

Garuda Linux is also really cool if you like Arch, but want something overly psychedelic.

Manjaro

I've been using Manjaro and Arch as my main system since 2013, and any problem I've had it would have been as bad or worse on a "stable" distro.

Last edited by lucinos on 2 Mar 2023 at 3:35 am UTC

Quoting: slaapliedje

Quoting: ShmerlI use Debian testing. But there are a few things I build or maintain outside of Debian repos:

* Kernel itself (I usually build newest release, testing gets kernels with quite some delay).
* Firmware from upstream kernel repo (for AMD GPU). This one is updated in Debian very rarely.
* Mesa main for gaming (not replacing system Mesa).
* Wine from WineHQ Debian repo.
* vkd3d-proton / dxvk master from upstream.
* Firefox beta from Mozilla Linux build.

Other than that, I use what's in Debian repos.

Sadly the first half of these are why I stick with nvidia.

But I also tend to use Debian Sid.

Garuda Linux is also really cool if you like Arch, but want something overly psychedelic.

I don't see an issue with using kernel and Mesa that way. Nvidia isn't making it better, in fact it's worse because they don't follow kernel releases strictly. You may end up in situation when your new (usually very recent) hardware requires kernel X and Nvidia blob only supports kernel Y which is older. With AMD that's impossible.

Last edited by Shmerl on 2 Mar 2023 at 5:17 am UTC

Quoting: Shmerl

Quoting: slaapliedje

Quoting: ShmerlI use Debian testing. But there are a few things I build or maintain outside of Debian repos:

* Kernel itself (I usually build newest release, testing gets kernels with quite some delay).
* Firmware from upstream kernel repo (for AMD GPU). This one is updated in Debian very rarely.
* Mesa main for gaming (not replacing system Mesa).
* Wine from WineHQ Debian repo.
* vkd3d-proton / dxvk master from upstream.
* Firefox beta from Mozilla Linux build.

Other than that, I use what's in Debian repos.

Sadly the first half of these are why I stick with nvidia.

But I also tend to use Debian Sid.

Garuda Linux is also really cool if you like Arch, but want something overly psychedelic.

I don't see an issue with using kernel and Mesa that way. Nvidia isn't making it better, in fact it's worse because they don't follow kernel releases strictly. You may end up in situation when your new (usually very recent) hardware requires kernel X and Nvidia blob only supports kernel Y which is older. With AMD that's impossible.

I haven't had an issue last longer than a few days with new hardware needing a different kernel, historically Nvidia has been WAY better at that then AMD. I still firmly believe AMD open sourced their stuff because they couldn't ever figure out the cadence of getting a driver out in time for a new kernel / xorg version on time. There were months when it'd be broken when a new version of Xorg would release, or some slight thing made it so it wouldn't compile for a new kernel. And I'm not even talking 'new kernel that I downloaded from kernel.org' or compiled xorg. Just stuff that would be in rolling release distributions.

It has gotten better with AMD after they opened stuff so it's all in mesa/kernel. But then you're in the position with what you deal with. You know what I deal with? 1) the bleeding edge nvidia cards definitely are supported quick in the drivers, and they get put into experimental at the very least, usually the drop into sid (which I'm using anyhow) or if you're running stable, they'll fall into backports. They haven't broken kernel compilation in ages. It's probably been 6 or so years since I remember having to download the .run driver install from their site to get a new video card to work.

Really what it comes down to is both have their pluses and minuses. A lot of it depends on what distribution you use as well. Debian makes installing the nvidia driver dead simple.

Quoting: lucinosManjaro

I've been using Manjaro and Arch as my main system since 2013, and any problem I've had it would have been as bad or worse on a "stable" distro.

The BEST thing about Arch is the documentation! It is legitimately the best out of any Linux distribution. So when they do break stuff, you can just go to archlinux.org and they'll almost 100% have already figured out it's broken and post on there how to fix it. If you've been using it that long, you'll likely remember when Ghostscript was borked (if I recall it was a break with the upstream project, but Arch of course builds things fast, so they were the first distro to get hit by it that I know of). The solution was of course to downgrade it to the not broken version. But the fun side effect was that you'd try to print stuff and get black bars wherever there was text!

Quoting: slaapliedjeThey haven't broken kernel compilation in ages.

I think I remember some news not so long ago that it took a while for Nvidia to support newest kernel. Whether you are affected by it or not totally would depend on hardware that needs such kernel. If your hardware isn't affected - you won't notice the problem. Point is that for Nvidia it's possible with their current approach.

Also, if you really like this out of kernel tree approach for whatever reason - AMD provides it all the same (with their dkms). So in that sense there are more options with them than with Nvidia, since you can choose upstream kernel or out of tree dkms, unlike only the latter with Nvidia.

But releases sync is really a minor issue for me vs the blob itself. I was glad to get rid of it and its integration problems once I switched to AMD.

Last edited by Shmerl on 3 Mar 2023 at 4:28 am UTC

Quoting: Shmerl

Quoting: slaapliedjeThey haven't broken kernel compilation in ages.

I think I remember some news not so long ago that it took a while for Nvidia to support newest kernel. Whether you are affected by it or not totally would depend on hardware that needs such kernel. If your hardware isn't affected - you won't notice the problem. Point is that for Nvidia it's possible with their current approach.

Also, if you really like this out of kernel tree approach for whatever reason - AMD provides it all the same (with their dkms). So in that sense there are more options with them than with Nvidia, since you can choose upstream kernel or out of tree dkms, unlike only the latter with Nvidia.

But releases sync is really a minor issue for me vs the blob itself. I was glad to get rid of it and its integration problems once I switched to AMD.

Yeah, as long as I've been using Linux, the best / easiest to just get going was the pre-Parhelia Matrox cards, granted that's discounting the Intel ones, which still kind of have crap performance. I had high hopes for ARC to become a competitor, because I've never been happy with either solution from AMD/Nvidia, they both seem like they have too many advantages or disadvantages between the two. Remember when there used to be a ton of different GPUs to choose from? Those were good days. I think even these days Matrox uses nvidia chips.

Better competition would surely help at least to keep prices on normal levels. Intel so far didn't start competing fully.

Quoting: ShmerlBetter competition would surely help at least to keep prices on normal levels. Intel so far didn't start competing fully.

Yeah, I'm not even sure prices are where I am complaining against, I'd just like some sane driver model, and some stability. The problem, as I see it, with modern day computers is new video cards have to give a reason for their existence these days, faster and faster frame rates on games just doesn't cut it. Nvidia seems to always be the one who innovates new stuff, for AMD to then follow. But then Nvidia releases stuff in a 'this generation it's just a beta, and slow, next iteration will be faster.' way. AMD just waits long enough and tries to make a standard way of doing the same thing...

The way it really should go is that the innovation should be in the API (Vulkan in this case) then the GPUs try to support those new features. It actually used to be this way, way back with OpenGL, then at some point the GPU makers started adding in the extensions to it.

Quoting: slaapliedjeThe way it really should go is that the innovation should be in the API (Vulkan in this case) then the GPUs try to support those new features. It actually used to be this way, way back with OpenGL, then at some point the GPU makers started adding in the extensions to it.

That's a recipe for disaster. You can't just sit down and design the perfect API and have hardware find a way to do it. Issues always get found in implementation, that's why software and hardware "iterate". What you are suggesting is the waterfall development model; people tried it, didn't work out.

Quoting: slaapliedjeYeah, I'm not even sure prices are where I am complaining against

Prices surely have been creeping up quite steadily. Remember when high end gaming GPUs were around $400+? Now they are above $1000.

Quoting: Shmerl

Quoting: slaapliedjeYeah, I'm not even sure prices are where I am complaining against

Prices surely have been creeping up quite steadily. Remember when high end gaming GPUs were around $400+? Now they are above $1000.

For sure... though honestly I've always bought the x8xx versions, never the x9xx or even Ti editions. I think the most I've spent on any of those was 800. Still crappy. That's why I'm just skipping whatever is coming out now. My current PC doesn't need to be upgraded in the slightest. Ha, I feel Liam having his survey popping up asking me to update my PC specs is egging me on to upgrade. But I really don't have a real reason to at this point (unless I upgrade my VR headset to something with a massive amount of more pixels, I see no need to get a new GPU).

Quoting: ShabbyXThat's a recipe for disaster. You can't just sit down and design the perfect API and have hardware find a way to do it. Issues always get found in implementation, that's why software and hardware "iterate". What you are suggesting is the waterfall development model; people tried it, didn't work out.

The problem is that makes it where we are now. Nvidia creating new features, AMD playing catch up, and then having two different implementations out there, and game devs have to target both features on pick which one they're going to support (see most Linux native games targeting nvidia and having issues running on AMD GPUs).