Nouveau Driver Remains Much Slower Than NVIDIA's Official Driver

[Brane215]

No it doesn't. Maintaining and ensuring ABI compatibility does not mean one has to sacrifice a project. It just means greater care and measures have to be taken to ensure backwards compatibility.

If whole point of the project is freedom ofsource, then fixed ABI limits that freedom. If project developers are OK with that limit, fine. If not, then the limit is too restrictive and forking might be solution.

With fixed ABI, how would things like Wayland ( or even mroe radical) be possible ?

[deanjo]

If whole point of the project is freedom ofsource, then fixed ABI limits that freedom. If project developers are OK with that limit, fine. If not, then the limit is too restrictive and forking might be solution.

With fixed ABI, how would things like Wayland ( or even mroe radical) be possible ?

There are plenty of ways but it requires effort to do so and means a lot more maintenance then the developers may wish to put in. The beautiful part of software development is that virtually anything can be done, it just requires the effort and know how to do it.

[bug77]

You have EOL hardware from official manufacturer standpoint.

That's because he has an ATI/AMD card which is 5 years old. If you go nvidia, their evil blobs still provide support for FX5200 - which is 10 years old. Ok, it's the last update for that nvidia card, but surely you can see the difference.

[cbamber85]

It's a developers OS for developers. If they ever want to change that and have mass acceptance they then have to start putting what non-developer end users want and concentrate their efforts there. Right now I don't see a lot of effort trying to change that. It remains as it always has been, developers are #1, end users at the very bottom.

I agree with your statement, but I think it's a sad reflection of the state of the Linux desktop. I am a developer, who chooses Linux as my primary platform because it is much nicer to develop on than the competition. But I don't develop for the kernel, Gnome/KDE, etc. I am an end-user for those components.

Developer and end-user are not mutually exclusive roles.

[deanjo]

Actually Canonical focuses on the end users and are the most popular distro.

Canonical doesn't do much development at all. They market to the end user.

And something tells me that the rest tries to follow suit but when they fail they say that they don't care about the end user but the developer.

There are few distro's that are community ran and also let the community dictate the direction of the distro. For example:

https://features.opensuse.org/

I mean there are plenty of distros that focus on the end user because they want linux to gain mass acceptance (which is also the goal of linus torvalds from what I heard).

A lot of distros don't really do much in development but simply repackage. Most of your development work for linux comes from commercial entities that are not interested in the desktop market. For example: http://www.tuxradar.com/content/red-...-desktop-linux Instead they concentrate their development on where the money is which is big enterprise.

The reason devs are #1 is because of people like you who think that somehow linux doesn't want to be on all computers and use it as an excuse to not improve their software.

The number one reason is because the desktop market is far less profitable then the mobile/embedded/enterprise market and given their limited resources that is what they concentrate on for their development work.

[deanjo]

I agree with your statement, but I think it's a sad reflection of the state of the Linux desktop. I am a developer, who chooses Linux as my primary platform because it is much nicer to develop on than the competition. But I don't develop for the kernel, Gnome/KDE, etc. I am an end-user for those components.

Developer and end-user are not mutually exclusive roles.

They are not mutually exclusive that is correct. But being a developer you are more then likely willing to go through a lot more hoops to work with your system then the non developer crowd who don't care how it is done but just that it gets done.

[crazycheese]

Amen brother! I will use your quote everytime one of these open source zealots tries to tell me that it's my fault for the OS not working and that I have to write the drivers myself.

If you can write drivers same speed you can troll, go ahead.

That all really depends on the target audience for the operating system. Linux and most open source efforts are about making the developers life easier and letting other contribute to the code not about how to make the end users life easier. It's a developers OS for developers. If they ever want to change that and have mass acceptance they then have to start putting what non-developer end users want and concentrate their efforts there. Right now I don't see a lot of effort trying to change that. It remains as it always has been, developers are #1, end users at the very bottom.

They are not mutually exclusive that is correct. But being a developer you are more then likely willing to go through a lot more hoops to work with your system then the non developer crowd who don't care how it is done but just that it gets done.

Canonical doesn't do much development at all. They market to the end user.

Here you go, you answered own question. The solution is: distributions with matching philosophy for different standpoints.
Canonical did introduce many end-user interesting projects: software center (much more comfortable to find and discuss software), jockey (to autofetch drivers), automatic Xorg restarts (to eliminate "CLI suprise" in case of bad drivers), upstart (before systemd was born), networkmanager (this is clearly userfriendly) early adoption of PulseAudio (with funny effects).

They do develop much less inkernel, but they do work on integration and desktop consistency. And they have not misused their lead/semi-lead position to prevent other distributions to profit from their efforts, as of today.

[deanjo]

Canonical did introduce many end-user interesting projects: software center (much more comfortable to find and discuss software), jockey (to autofetch drivers), automatic Xorg restarts (to eliminate "CLI suprise" in case of bad drivers), upstart (before systemd was born), networkmanager (this is clearly userfriendly) early adoption of PulseAudio (with funny effects).

software center --> their version of click and run nothing special I believe Mandriva also had there version as well. Networkmanager was Redhat and Pulse again was not developed by Ubuntu and when it was introduced brought A LOT of heartache.

[hoohoo]

It doesn't seem to be "much slower", rather "reasonably fast, but slower then proprietary". Regarding trolls above, stable OSS driver will be great, even if it is 10% slower (most people wouldn't notice) because it will be much more stable (tested by millions of people) which makes difference especially for older cards, that companies tend to ignore. Also, security can be reasonably checked, rechecked and fixed by experience kernel devs, closed binary code is simply potential security and maintenance problem. And last but not least, having ALL drivers in Kernel, eventually, will make life of developers lot easier, especially regarding refactoring of various APIs.

Some people said they've had issues with closed vendor driver. Fair enough, me too: an AMD c50 based netbook I bought was not properly supported for some time. However my succession of card for desktops (8800GT -> GTX260 -> GTX560 and much longer string of AMD up to an HD7970) were supported well and quickly by the vendors.

Last time I looked into it (maybe 6 months ago) the vendor's drivers were still untouchable for OpenGL feature support and execution speed compared to the open source drivers.

Even on 2D operations like redraw window contents during window resize, the open source drivers drag.

Using open source drivers I find that Firefox may not use hardware acceleration and is thus dog slow. Never had that problem with proprietary.

Having all drivers in kernel / refactoring: before I buy a new video card I check for vendor support for the new card for Linux. If the kernel devs refactor or otherwise improve the APIs or kernel internals then it is AMD or NV responsibility to support the new kernel. In any case, how often does a major kernel subsystem like graphics (or SCSI layer or network stack) get overhauled? Every five years? Longer?


Other people's comments:

(paraphrase) "too stupid to uninstall a driver": I am perfectly capable of removing Nouveau after an install. But it's a pain in the neck: the VESA driver will just not get used if I install proprietary driver, so why should I have to mess with blacklisting the nouveau module? It's as if someone does not want me to disable it. It's not as bad now as it was when it first showed up, I guess some zealots at e.g. Fedora got half the message.

FPS & game consoles: I use PCs & Linux professionally, I play shoot 'em up games on Windows. I do not think consoles offer a compelling value proposition.

"trolls": I reject you out of hand. My comments about open source AMD/NV video drivers are perfectly valid.

[deanjo]

(paraphrase) "too stupid to uninstall a driver": I am perfectly capable of removing Nouveau after an install. But it's a pain in the neck: the VESA driver will just not get used if I install proprietary driver, so why should I have to mess with blacklisting the nouveau module? It's as if someone does not want me to disable it. It's not as bad now as it was when it first showed up, I guess some zealots at e.g. Fedora got half the message.

Someone tried to make that all easier with a trivial patch but some developers got their panties all bunched up.

http://www.phoronix.com/scan.php?page=news_item&px=NjYwMw