Gordon's Thoughts On Open-Source GPU Drivers

[pvtcupcakes]

Citations please. What do you really know about "typical end-users" and where do you know it from? How do you know how many are running binary drivers? Since you have such accurate data, go ahead and exclude anything from your figures that doesn't really belong in a discussion about the needs/wants of desktop users such as server machines.

Secondly, if it's good enough for most users, why all the discussion? Why all the complaining, fudding, and carrying on in every forum, blog, bug report, etc, etc. Certainly this has to be one of the most vocal minorities ever.

Third, if things are as you say, "90%" and "typical", how can you be sure that people are happy with the open drives and haven't just given in and accepted that FLOSS will never offer the extras such as gaming and just turn to something else (ps, wii, xbox, windows box, etc). How much feedback bias is going on where fact that people can't influences that people don't.

Lets say tomorrow, that someone waves a magic wand and all the binary drivers were open, pattens were abolished, Epic came through with UT3, Steam etc... all happy rainbows and shit. Are you saying that the typical 90% won't give a shit because Supertux, KPatience, and Firefox have all been "fine" for some time?

(Oh wait scratch Firefox, they're not fine with most FLOSS drivers either )

The 90% I'm referring to (obviously I don't have statistics, I just mean most users) are people like grandmothers, business people, or people who only browse the web.

These are the people that might only use their computer for a few minutes a day to check their email, go on facebook. They just don't have any interest in running modern games like Battlefield. Of course the open driver isn't perfect, but it's capable of displaying Facebook within Firefox. There isn't really a reason for these people to use the closed drivers.

For people who watch HD video, then I will admit that the Nvidia blob with vdpau is better than nouveau currently. But on AMD, both drivers are equally bad at accelerating video.

I'm not deluding myself that everyone else who uses a computer is like me and is interested in hardcore gaming.

[deanjo]

The 90% I'm referring to (obviously I don't have statistics, I just mean most users) are people like grandmothers, business people, or people who only browse the web.

Ya, those people are really big linux users now aren't they? Those people could use an iPad, smartphone, internet tv probably and be just as happy (if not happier).

[cynical]

To me the motivation behind wanting open or free software is so similar that I don't understand why people want to argue so much about the differences. Both philosophies are about wanting freedom. Open just goes far enough to see the code, and libre goes far enough to make sure modifications are possible. It's like the difference between knowing where/how a bike was made and being able to change and share your changes to it.

People talk about free/proprietary here and look at using closed drivers as being pragmatic, but I use open drivers for the same reason.

1. I don't need to worry about reinstalling them with kernel updates.
2. KMS will make the boot process look more polished/stable
3. There will be fewer graphical glitches than with ATI's closed driver
4. Better integration with Linux
5. Quicker changes due to its open nature, which is why I use Linux in the first place!

[deanjo]

People talk about free/proprietary here and look at using closed drivers as being pragmatic, but I use open drivers for the same reason.

1. I don't need to worry about reinstalling them with kernel updates.

dkms takes care of that, if not then usually when a update to the kernel is put out there are prepackaged blobs built against it available.

2. KMS will make the boot process look more polished/stable

True but it doesn't really do much for real user functionality. Not sure how it makes it look "stable" however.

3. There will be fewer graphical glitches than with ATI's closed driver
4. Better integration with Linux
5. Quicker changes due to its open nature, which is why I use Linux in the first place!

That is very vendor specific. I wouldn't necessarily say "quicker changes" as well as it depends what you are looking for as a change. openGL/gpgpu/decoding acceleration/ etc support is agonizingly slow to implement.

[psycho_driver]

@Gordon,
You know what's REALLY dangerous to the Linux desktop? Drivers that are a nightmare to install, update, remove. Ones that don't reliably support basic desktop needs; multi monitor, KMS, sleep/suspend, etc. Ones that break with kernel upgrades. Ones that won't even work with new releases of Xorg!! I'm looking at you proprietary drivers.

I'll stick with the drivers that come with my distro ( Fedora ) thankyou. Lets walk before we run.

Sorry, but in the case of nvidia at least, that's more your distro's fault than that of the binary drivers. I know, because I run gentoo on all of my machines but my wife's, which is Fedora. It has given me some grief before with the nvidia driver compatibility.

They always just work, and well, on the gentoo systems.

[Ex-Cyber]

The cleaner solution is to open up the source to drivers we already know work, and have everyone--hardware vendors, distros, individuals--working in the same source tree. Every one would win.

I have no idea how that would be cleaner. Vendor drivers are virtually guaranteed to have vendor-specific interfaces and optimizations all over the place (because they can, and because it gets them that extra 5 fps in Crysis). Something like Mesa+Gallium3D is the only way we realistically end up with everyone working in the same source tree without massive duplication of effort.

[cynical]

dkms takes care of that, if not then usually when a update to the kernel is put out there are prepackaged blobs built against it available.

Assuming dkms works well that is interesting but it isn't quite the same as not having to do anything.

True but it doesn't really do much for real user functionality. Not sure how it makes it look "stable" however.

Screen flickering doesn't exactly inspire confidence in one's operating system. It looks far more professional and I don't think you can argue that looks aren't important.

That is very vendor specific. I wouldn't necessarily say "quicker changes" as well as it depends what you are looking for as a change. openGL/gpgpu/decoding acceleration/ etc support is agonizingly slow to implement.

I mean more rapid development/releases, or at least bug fixes. Having to wait months for ATI to release a new driver to fix a graphical corruption bug isn't fun. I'm not saying they couldn't do a good job because they are proprietary, just that in my experience things happen more quickly on open source code.

[deanjo]

Assuming dkms works well that is interesting but it isn't quite the same as not having to do anything.

Actually it is. New kernel gets installed, drivers get automatically built against the kernel.

Screen flickering doesn't exactly inspire confidence in one's operating system. It looks far more professional and I don't think you can argue that looks aren't important.

To tell you the truth, I set my boot to verbose to look for potential issues that would otherwise be masked until you search looking at logs. Even then if I do choose to do a silent boot there is one "flicker" when going to the desktop and it is nearly unnoticeable on a LCD screen. In the old days you used to hear your CRT flip modes but that isn't the case anymore.

I mean more rapid development/releases, or at least bug fixes. Having to wait months for ATI to release a new driver to fix a graphical corruption bug isn't fun. I'm not saying they couldn't do a good job because they are proprietary, just that in my experience things happen more quickly on open source code.

That maybe the case with ATI's blobs but Nvidia does a considerably better job of it. Even with free drivers there are issues that have been around for years and there is no guarantee that it will get fixed any quicker.

[Kano]

The most annoying thing with nvidia drivers is the lack of xrandr functionality. Also the nvidia-settings app can not be used on commandline to configure a 2nd display. I would really like to write just a tiny script that switches to hdmi output (and disabled internal on a laptop) and configures audio to use external hdmi but thats impossible with nvidia. If you want to use xrandr on commandline with fglrx it works in theory but a Virtual entry might be needed. That should be fixed as the oss drivers do not need that. Do you think that nvidia-settings is better to config a multi monitor layout on the fly without restarting the xserver than xrandr can do?

[deanjo]

The most annoying thing with nvidia drivers is the lack of xrandr functionality. Also the nvidia-settings app can not be used on commandline to configure a 2nd display. I would really like to write just a tiny script that switches to hdmi output (and disabled internal on a laptop) and configures audio to use external hdmi but thats impossible with nvidia. If you want to use xrandr on commandline with fglrx it works in theory but a Virtual entry might be needed. That should be fixed as the oss drivers do not need that. Do you think that nvidia-settings is better to config a multi monitor layout on the fly without restarting the xserver than xrandr can do?

Yes the lack of nvidia-settings being able to set the monitor setup from cli is annoying but there is however a solution coming up for that. As far as multi monitor layouts goes I don't have an issue using nvidia settings for my display setup and I have to say using mosaic mode on the quadros @ work has been a pleasure to use on my quad display there.