XWayland Gets Initial Support For EGLStreams To Support NVIDIA's Driver

[RussianNeuroMancer]

Uhhh.... I very much remember using bumblebee with Steam didn't work with the launcher and required the launch command in the games setting to be changed to include it. I also had to include it with Wine or virtualbox/vmware, for any DCC program etc.

You mistake with assumption that bumblebee issues could be in any way related to DRI_PRIME usage experience.

The 512MB RAM resource constraint example wasn't about the 7 series nvidia GPUs, it was about hardware over time not always being enough for software regardless of driver support, I thought the RAM with Gnome was a good example.

It isn't, because for half year or so I used Gnome Shell on 7 inch tablet (DEXP Ursus 7W) with 1GB RAM. (Now I writing this message from 2GB RAM version of Dell 5855, not because I switched to get more RAM, but just because Ursus 7W died.) With couple of tweaks Gnome is pretty much fine for such hardware (I get as low as ~300MB after boot) and there is no other choice anyway, because Win10 consume twice more RAM, Win8 phased out and Android x86 7.1 is stuck at Linux 4.9 (which makes it barely usable on Intel BayTrail/CherryTrail) until they release Android x86 8.1 in a couple of years.

If you upgrade the RAM you're opting to spend more money, in which case you could replace that GPU

Not an option (due to incompatibility, working GPU unavailability, etc.) and as I said there is no other reasons to replace this laptops besides nvidia legacy driver issues.

It's not nvidia being a dick to it's customers? You bought the laptop with nvidia GPU say, did it come with Linux or something like Windows OS installed? If you choose to change the OS, how is nvidia being a dick?

Unrelated as there is no such issue with other vendors and nvidia is only vendor who stands out and not support it's own customers. One exception I can remember is here, and another is PowerVR-based GPUs by Intel, but it's not about preinstalled OS, as you can see (in case of PowerVR-based GPUs support was dropped for Windows too, at the same time).

There isn't any major nvidia bugs that I'm aware of, for the customer base that matters to them, anything serious will get resolved.

It's good idea to tell desktop (no proper Wayland support) and laptop (no proper Optimus support while fglrx had supported PowerXpress on Linux since March of 2011 year) Linux users beforehand that they doesn't matter to nvidia. This is actually good advice which I will keep in mind, thank you and LIKE

By the way, reminded me that before this Wayland story we had Xrandr circus with nvidia proprietary driver for years.

Remind me how long Intel has not released a stable version of their drivers again, despite being open source?

You don't need to, fortunately if there is enough interest people could stand up and fix it themselves, like for example Hans de Goede from Red Hat did with Intel BayTrail/CherryTrail. Yeah-yeah, I don't need to remind you how it's impossible with closed drivers.

Not sure why you're hating.

Well, let's take bad support of desktop and laptop Linux users as an example. Because, you know, many people around here is desktop and/or laptop owners who also happened to run Linux.

> Transition period for this HW generation takes some time and was completed couple of years ago.
Eh? It held xorg version back for 6-12 month on some distros even though nvidia users were not the reason, AMD GPU users required it to be held back.

You read it right.

Not sure how that was considered completed years ago, prior to vega you mean?

Well, as was said above, Vega situation was exactly as with all nvidia hardware (go on, tell me that requirement of proprietary driver installation (hybrid in AMD case) is something bad and stuff...) so I obviously talking about transition of non-GCN GPUs from fglrx to FOSS stack.

[cybertraveler]

ARCH SUPPORT

LEGACY HARDWARE SUPPORT

I get the sense that you're trying to not understand me. I really don't understand how you can not realize the numerous problems with what you're saying. I'll point some of them out to you:

1. "you're generally expected to have technical competency that you understand there will be more effort involved and knowledge required" <- nvidia is more effort and more complications. Even if someone is competent and skilled, it's likely that they will prefer a system that requires less effort. With AMD and Intel this will be the case in most circumstances.
2. "Your more than likely to have issues with things like proprietary wireless driver support, and this isn't something that only plagues Arch." <- I have no idea why you're even mentioning this. My entire post can pretty be summed up as 'I don't recommend nvidia hardware because they mostly ship proprietary drivers and they make it hard for third parties to implement Open Source drivers.' It should be obvious to you that I do not support, desire or promote other hardware manufacturers which have the same practises as nvidia. Perhaps you've categorized me in your mind as being some kind of die-hard AMD/Intel fan that hates nvidia because I had a bad experience once when I was 5. That's not the case. The moment nvidia change their stance on GNU/Linux to be more like Intel and AMD is the moment I start supporting nvidia. Until that times comes, I wont purchase or promote nvidia or any other company that has similar practises. Please note: I'm only talking about people switching to GNU/Linux. If someone loves Windows and only wants to use Windows I'd likely promote nvidia over AMD as the price/performance ratio is generally better for nvidia than it is AMD.
3. "I'm pretty sure I've seen an open-source one that was good until a few kernel releases ago, then something changed and it's no good but the vendor isn't too interested or something in addressing it." <- I'm not going to cover many more points of yours after this one. It honestly baffles me that you can say most of this stuff without realizing for yourself what the problems are with what you are saying. It almost feels embarrassing to have to point this out: if a hardware manufacturer creates or supports an Open Source driver for their hardware and bugs emerge in that driver, anyone with the ability to hack on driver code can fix the issues. Of course, Open Source drivers will get bugs and regressions. I'd never claim otherwise. The major difference between Open Source drivers and proprietary drivers is that in the case of the Open Source drivers, we can actually, practically do something about it if we want to. OH MY GOD!! Please. Think about it
4. "Geforce 7 is pre 2008" [snip] "You can argue that the support may be removed for other cards in future" [I would argue that] ", but realistically, how long ago did that support get removed since that GPU was released? Are your users highly likely to still be using the system with that GPU when that time comes and have an issue with it impacting software support?" <- I don't know. I've got a bunch of really old systems that I still use today. My gaming machine has a whole bunch of 8 year old parts in it. I dislike seeing perfectly working hardware get scrapped because of avoidable issues like proprietary drivers no longer being updated. Thankfully all of my systems rely only on Open Source drivers, so I can and do tend to use them until they physically fail. When I recommend distros and hardware to other people, I want to give them the same great experience I have. I want them to have systems that they can reasonably expect to get use out of for years to come.

I'm not going to respond to the other points you made. It's not because I'm 'conveniently ignoring' ignoring them. I just find it mentally taxing explaining things that are abundantly obvious to me. I don't know why you're not able to figure it out for yourself.

[brrrrttttt]

"Nvidia makes huge code contribution to qt with New Qt 3d Studio application." https://www.phoronix.com/scan.php?pa...A-Qt-3D-Studio

Nvidia unloads a large maintenance burden to Qt. This software is basically for commercial customers anyhow?

[polarathene]

I get the sense that you're trying to not understand me. I really don't understand how you can not realize the numerous problems with what you're saying. I'll point some of them out to you:

I'm not going to respond to the other points you made. It's not because I'm 'conveniently ignoring' ignoring them. I just find it mentally taxing explaining things that are abundantly obvious to me. I don't know why you're not able to figure it out for yourself.

I honestly don't see much effort required to run nvidia, sure it's a bit more effort than Intel/AMD in most cases(albeit for some users they may have issues with Intel drivers and have to learn to use modesetting, or for AMD the extra work to currently use compute(ROCm) like CUDA), but it's consistent/reliable effort. On popular distros like Manjaro you legit just select when installing that you're using nvidia hardware, no extra efforts there. Arch, you'll likely have to read up on some things, I agree less time/effort required to is better, but that isn't always the case as I've been trying to point out, even with Intel/AMD. "Why can't I use my AMD Vega GPU?" and "Why can't I use newer xorg while nvidia/intel users can", while past problems still happened and weren't that short lived.

Yeah it's great point about open-source drivers, but nothing had been done for several kernel releases(despite numerous users on the internet bringing it up and I believe bug reports), I did quite a journey myself spanning days understanding network side of linux considerably, trying different tools, trying to figure out how to get the NIC working. I wasn't quite at the point that I could solve it myself even with the code available, it didn't help that it was well known that linux networking tools were unable to adjust the power state, that it could only be corrected by booting into Windows and changing a setting there. For that experience it didn't make much of a difference that code was open or not, I wanted/needed the hardware to work and despite my efforts(which is far more than would be invested on nvidia hardware), I was unsuccessful. Had an issue with proprietary firmware for my motherboard, TSC was failing on resume from suspend, it was falling back to some other timer(HPET I think) which caused perf problems.. I e-mail ASRock, the mb vendor and get a prompt response and BIOS update to fix it.

If you were in one of the situations where your GPU wasn't going to work for months on end, or was holding you back from software updates you'd like, while the proprietary alternative is working soundly, what do you prefer to use? In some cases you could perhaps put in some extra work to resolve those, but it'd be more than the work/effort involved with the nvidia drivers working, a point you try to make as a solid reason not to go with nvidia. It's great that this isn't common, just bad luck. At least with nvidia you're far more likely to get a consistent experience/path where you don't get that issue? (consequently it does have it's drawbacks due to the points you raise too, where a wider span of GPU models is affected because of proprietary code that the community cannot fix even if they wanted to, that it prevents Wayland support and the like...not so much because nvidia isn't supporting it but because the community itself doesn't want to assist due to the proprietary stance, fair enough).

It's all good if you don't want to go further with this discussion. It's basically the same points being echoed, and we both feel strongly that we have obvious points that we're failing to communicate to one another. I'm not against open-source in anyway, I contribute open-source code when I can myself, I'd be all for nvidia having decent nouveau driver like AMD has managed to migrate from their binary driver to a better performing open-source driver/community. I just don't think nvidia having a proprietary driver equates to experiences that are worse than AMD/Intel, I've used all 3 through the years and they all have had pros/cons.

If in the future nvidia start releasing more documentation and promptly provide legal access to signed firmware/microcode then I will promptly reconsider my stance on them. If nvidia go so far as actually developing open source code for GNU/Linux, then I'll buy and promote their products.

They have been doing that(though not to the amount everyone would prefer) to assist the nouveau project? They haven't left that project in the dark completely. It's definitely an area they could improve on considerably.

Nvidia does go so far as actually developing open source code for GNU/Linux....? So start buying/promoting their products I guess?

I don't think nvidia are a bad company. I just think their GNU/Linux support is massively inferior to that of AMD and Intel. As such, I don't support them and I encourage GNU/Linux noobies to avoid them too.

I agree with you and think that's a perfectly fine thing to do. Just that it's not always going to turn out to be rosey advice of AMD is better than nvidia on Linux, which I've been trying to communicate. AMD is typically the better option to go with these days, I just find some of the reasons as to "why not nvidia" to be a bit confusing or exaggerated. Nvidia might not have the best driver suport on Linux, but they do contribute to bettering Linux, they're not entirely proprietary and bad like people seem to think on here.

[polarathene]

You mistake with assumption that bumblebee issues could be in any way related to DRI_PRIME usage experience.

Uhh... I thought we're talking about nvidia here, specifically their Optimus support with Intel iGPU and dGPU on laptops? Bumblebee is pretty much well known for being the approach to handling that? I think these days DRI_PRIME support might be better? I thought that was more for the open-source drivers though than nvidia's proprietary driver? Sorry for the wrong assumption?

It isn't, because for half year or so I used Gnome Shell on 7 inch tablet (DEXP Ursus 7W) with 1GB RAM. (Now I writing this message from 2GB RAM version of Dell 5855, not because I switched to get more RAM, but just because Ursus 7W died.) With couple of tweaks Gnome is pretty much fine for such hardware (I get as low as ~300MB after boot) and there is no other choice anyway, because Win10 consume twice more RAM, Win8 phased out and Android x86 7.1 is stuck at Linux 4.9 (which makes it barely usable on Intel BayTrail/CherryTrail) until they release Android x86 8.1 in a couple of years.

I guess I failed to get my point across here? My bad for mentioning specifics, althouogh I remember someone saying their Gnome Shell experience was booting to 1.1GB RAM and I thought that'd be a good example to contrast to Gnome in earlier days with GTK2 and 512MB RAM? You kind of provide examples that work instead with Windows and Android over time becoming problematic(not sure how much of Android x86 is open source, but I'm sure there is a full open-source example that could be used....as long as you get what I was trying to say.)

Not an option (due to incompatibility, working GPU unavailability, etc.) and as I said there is no other reasons to replace this laptops besides nvidia legacy driver issues.

Ok great, I tried to use an inability to replace/upgrade RAM in a system as an alternative example of this and that in time that it could require you to get a new laptop or make new choices/sacrifices in software/DE decisions. I get that in your particularly case that may not apply....but you get the point? It's true that with AMD your GPU driver issue wouldn't be there and you could have a better experience on the hardware without changes....just try understand my point with the RAM, that proprietary isn't the only cause of the issue, open-source software/drivers could still lead to desired experience not working out.

Unrelated as there is no such issue with other vendors and nvidia is only vendor who stands out and not support it's own customers. One exception I can remember is here, and another is PowerVR-based GPUs by Intel, but it's not about preinstalled OS, as you can see (in case of PowerVR-based GPUs support was dropped for Windows too, at the same time).

If you buy hardware in a product that comes with a different OS and is intended that was as the product, as in that is specifically what the vendor is supporting officially(if they endorse Linux support specifically in that actual product not just their hardware, different story)... Then I don't get how you're complaining is justified, it just so happens your GPU works with their proprietary driver on Linux up to a certain point. Did the support end on Windows too for the nvidia hardware? Had the warranty expired since? Is nvidia legally going against the terms of support they have with that product to you as the customer? I'm going to say no. Not sure how that's treating the customer bad? Plenty of businesses operate that way, if I was expected to offer support to all my clients, unpaid time for an eternity I don't think that's going to work out well for me? You bought the product with fixed terms of support(if any).

The fact that it works better with another vendor like AMD, is a positive for those vendors. All the more reason to go with them over those whom don't. But it doesn't mean the other vendors that don't have as great support are bad if they're not required to.

It's good idea to tell desktop (no proper Wayland support) and laptop (no proper Optimus support while fglrx had supported PowerXpress on Linux since March of 2011 year) Linux users beforehand that they doesn't matter to nvidia. This is actually good advice which I will keep in mind, thank you and LIKE

It's not that nvidia hasn't tried to make Wayland support a possibility. It's just other developers are not interested in supporting their approach. That's cool, don't support it, leave that to nvidia to work on resolving. Afaik, they are working on resolving that with current open-source projects, actively trying to engage with the community for feedback and work with the GBM devs for the better of everyone.. How's that bad?

Is Wayland in a good state that having nvidia driver support would be good? From what I've seen Wayland is still rough around the edges and I wouldn't personally use it yet. Gnome supposedly has support for nvidia eglstreams, not too good I hear but if they want to support it good for them(and their customers who matter to them regardless of their hardware preference, they're choosing to be good to their clientbase not alienate or judge them....which seems to be an issue you have with nvidia for being proprietary). Nothing wrong with using X11 atm still, it's actively being developed, can do things that Wayland presently cannot that matter to certain users, has a wider amount of support, etc.. Wayland will be in a more relevant position year by year as it progresses, but it's not for everyone atm.

Proper Optimus support? Gnome supports this to my knowledge, even without Gnome, you can at least still say what GPU to run apps on which shouldn't be too many? The ability to do that works, there are programs that allow for listing what is relevant to use the dGPU with and have that work seamlessly for you like I assume you want, so bit confused with what you're going on about here. Probably not as good as Intel/AMD if that's what you're trying to convey? Nvidia has actively been working on improving all this, for some of the optimus support they've been blocked by Intel having to have new code to support sharing of the framebuffer, nvidia wrote alot of the code for that to my knowlege but collaborated with Intel. The work they continue to do benefits other vendors too, not just nvidia. For example eGPU support over ThunderBolt 3.0 is able to benefit as it too requires the Intel iGPU support(something about fencing in framebuffer), this is regardless if AMD or nvidia eGPU afaik.

By the way, reminded me that before this Wayland story we had Xrandr circus with nvidia proprietary driver for years.

I'm not familiar with it Perhaps it was before my time on Linux when AMD GPU wasn't in too good of a state as it is today?

Well, let's take bad support of desktop and laptop Linux users as an example. Because, you know, many people around here is desktop and/or laptop owners who also happened to run Linux.

Isn't that on the user for choosing hardware that isn't as well suited for most on Linux? Just like with other hardware with proprietary drivers(wireless is a good candidate here). AFAIK nvidia is more interested in the professional market on linux then it is desktop users? That's where their real customer base is? Compute workloads, embedded products like in car displays, workstations for digital content creation(where AMD GPU was notoriously bad in the past), etc. Quite a few of DCC software only has CUDA support, this isn't something the users have much control over, I believe this will improve as AMD is only getting better and more attractive with their drivers, it'll just take time before the professional software supports things like ROCm.

Just to be clear, I'm not a big fan of nvidia proprietary driver and would love an equivalent open one to the quality AMD has these days. That I myself prefer to purchase AMD GPU in future, provided I can use it for what I need to do as a replacement to nvidia's hardware(not possible at present). nvidia doesn't work great on KDE with some things, but it otherwise does work well enough, on laptops I do remember it being a headache(prior to Manjaro), and that Fedora required extra work(with some confusing history for users new to fedora with different approaches), that openSUSE had poor support too(not a fun time especially if you wanted to use TumbleWeed, kind of had to stick with Leap) although that's changed recently.

[cybertraveler]

polarathene: there are exceptions to the rule. Nvidia do provide some Open Source contributions, but they don't do what AMD and Intel do. They don't produce awesome, near-complete, Open Source drivers for their consumer graphics cards. They often don't even provide the firmware/microcode required for someone else to do it for them. Again: you must be deliberately trying to not see my point.

you said: "I did quite a journey myself spanning days understanding network side of linux considerably, trying different tools, trying to figure out how to get the NIC working"

In my original post I made it very clear that I advise users to pick hardware that has Open Source drivers *and* is known to work on GNU/Linux. These people that I help don't have the problem you have.

you said: "I wasn't quite at the point that I could solve it myself even with the code available". That's not relevant to my point. I'm not out there telling users to "fix the code". Just having Open Source drivers is often enough. Other people who can fix the code often will. Open Source helps us find a way even if the original hardware OEM is non-compliant.

Your comment about ASRock providing a bios fix can be described as anecdotal at absolute best. If you're suggesting that manufacturers that only have proprietary drivers/firmware are generally better at shipping fixes than Open Source, then we must be living on different planets. I find that almost every hardware manufacturer will stop supporting their products after a small number of years and then you are stuck with whatever you've currently got. With Open Source drivers you and can and do get updates and improvements for many years after the product stops being sold. Sometimes over a decade. As new distros and kernel updates are released, all the old hardware drivers are dragged along with these releases. With proprietary drivers this is not the case at all. I have a bunch of old hardware with Windows XP drivers that I can't use now on Windows 7. Besides: we have a near universal problem of proprietary firmware right now. It's one of the next big mile stones for people into Open Source to deal with. If you're out there promoting or encouraging that GNU/Linux users purchase hardware like Nvidia then you're actually working against us. You're trying to get us to move backwards. It's taken a long time to get to this point where we can have fast modern systems full utilizing their hardware with pure Open Source drivers. AMD & Intel have played an important role in that, that people like myself are incredibly grateful for. If Nvidia don't want to come along for the ride then we (the people that care about software freedom) will leave them and you behind us. The next step is sorting out open hardware and open firmware.

The Vega situation is again an exception to the rule. If a user buys any other new AMD graphics card except for Vega they will have an amazing experience. If a user buys an RX 580 it will work out of the box on the popular distros and perform well and stable with the Open Source drivers. Even if they buy Vega, it's no big deal. They still have the proprietary driver (same as nvidia) and in ~3 months they'll be able to install Ubuntu 18.04 and get access to the Open Source driver. In 10 years from now, if their Vega card is still physically working, they will likely still be able to use it with the latest Linux distros with minimal or no effort. I wonder what their experience would be 10 years from now if they bought a 1080ti. And again: I really did in my opening post state very clearly that I advise users to pick hardware that has Open Source drivers and is known to work on GNU/Linux. As such I wouldn't advise AMD Vega and I wouldn't advise any new nvidia graphics cards. Not 1. I repeat from above: Nvidia often don't even provide legal access to the signed firmware/microcode so other developers can make their hardware integrate better with the Open Source ecosystem.

If you really don't care about computational freedom and Open Source, and you really don't understand and experience the benefits of it over proprietary code, then I'm wondering why you're even using GNU/Linux. You can run KDE on Windows apparently. You can run lots of other popular Open Source apps on Windows. You can take the amazing stuff produced by the Open Source crowd and run it in your backdoored, slave-user, permissioned, proprietary wonderland. There sure as hell are draw backs to running GNU/Linux instead of Windows. All those draw backs will be gone and you can have the full, ultimate proprietary driver experience on Windows. Go for it! Live your dream! Nvidia hardware works faster on Windows too. Nvidia care more about their Windows users.

[Khrundel]

But they want others (who don't profit by it) to do their work, that is the problem.

I wouldn't say that Nvidia wants anything from wayland and others. They were happy with X11. Even after wayland will replace xorg, they will still need to support X11 because of different UNIX-like OSes and old linux distros (enterprise apps frequently support only some ancient version of Red Hat). Wayland is just a different headache for Nvidia. They are willing to support Wayland, but with as little efforts as possible. Rewriting driver with GBM/DRI/MESA is too much for Wayland, or even for Linux itself.

On the other side AMD worked hard to get their stuff mainline merged (so that others could profit by it if they want)- that is a huge difference.

They have started their opensource initiative after they bought ATI. AMD lacked money, they are just trying to drop their own drivers development. For years opensource features and performance were poor, until Valve came to linux and even hired developers to enhance AMD card drivers. But Valve came because there was excellent NVidia driver. So, if you need to thank someone for growth of Linux games and evolution of opensource drivers start with NVidia.

[Khrundel]

You definitely should have been on the open source driver. Basically native and ported games have worked for a long time if you do the gl override. Maybe it's an inconvenience at most, but at least you would have had a perfectly stable and usable configuration. Sure AMD deserves some grief for releasing such a horrible proprietary driver as that, but at that same time they already had a far superior driver working extremely well.

I've tried that (have to... there was no working fglrx some time). Games performance was worse. Unigine valley showed better FPS, but without half of effects

[cybertraveler]

They have started their opensource initiative after they bought ATI. AMD lacked money, they are just trying to drop their own drivers development. For years opensource features and performance were poor, until Valve came to linux and even hired developers to enhance AMD card drivers. But Valve came because there was excellent NVidia driver. So, if you need to thank someone for growth of Linux games and evolution of opensource drivers start with NVidia.

I think there may be some kind of psychological phenomenon occurring among some of the people here. I'm not sure what it is exactly or what it comes from, but I can barely believe some of the stories I'm hearing. What you just said is amazing. With no hint of sarcasm, you have suggested that we should be thanking Nvidia for their help evolving Open Source drivers. Obviously if you just came out and said that statement with no further comments or context, people would consider you bat shit crazy. But alas! You have provided us a story to explain your seemingly absurd claim. All we have to do to believe your claim is accept the following premises:

• Valve came to Linux because there was excellent NVidia driver.
• AMD lack money to pay for developers.
• AMD are planning to drop their own driver development.

Are you serious?

Props to you for the creativity. I've got to say. That's pretty damn impressive. If someone asked me to come up with a somewhat plausible argument to back the claim that NVidia are champions of evolving Open Source drivers that we should be thanking, I'd probably tell them "it can't be done". There's just no way I could frame that proprietary software yielding, firmware withholding, occasional obfuscated code dumping, corporation as some kind of champion of Open Source. But you did it! Well done. You literally have me laughing right now.

Assuming you can clean up well and are comfortable donning a suit, I think you'd do great in politics.

[duby229]

I've tried that (have to... there was no working fglrx some time). Games performance was worse. Unigine valley showed better FPS, but without half of effects

With r600 that was kinda the trade off, lower performance in games but overall a way better experience. And every single Unigine demo released to date has numerous OpenGL specification violations. Just look at your .drirc, there are tons of workarounds in there for broken binaries. AMD and nVidia both are guilty of fixing broken game binaries with driver workarounds. That's a problem mesa deals with, but at unfair disadvantages. "The Way it's Meant to be Played" really means something more like, "Worked around to make the Best of it". It's unfortunate as hell. But still the better driver is the open source stack, for sure.