Arch Linux Revolts Against ATI Catalyst Driver

[phoronix]

Phoronix: Arch Linux Revolts Against ATI Catalyst Driver

While AMD continues to improve the ATI Catalyst Linux driver from where they were at years ago by introducing new features like CrossFire and OpenGL 3.0 support while addressing outstanding bugs, no Linux graphics driver is yet in a perfect state. As a result from our post yesterday we have read many driver complaints for both ATI and NVIDIA on Linux...

http://www.phoronix.com/vr.php?view=NzEwMg

[JGC_]

I would like to give some more explanation about the situation with Archlinux and Catalyst.
First of all, our current maintainer, Andreas Radke, is subscribed to AMDs beta mailinglist. He receives information and updates to the drivers before they go public. The mailinglist is very low traffic and almost no useful information is given. The driver updates that are announced on this list lately don't show much improvements in the driver itself, but rather "packaging improvements".
Without ugly workarounds that involves symlinking or editing a binary blob using perl or sed, it's not possible to make the current catalyst drivers working on a non-multilib distribution that doesn't ship /usr/lib64. Symlinking /usr/lib to /usr/lib64 is ugly, while the other solution involving editing the binary driver, violates the license agreement we have with AMD (we're not allowed to redistribute modified drivers). Setting LIBGL_DRIVERS_PATH or LIBGL_DRIVERS_DIR environment variables as we did before no longer works with these drivers somehow, the DRI drivers are always loaded from a path where they shouldn't be installed.

Another point is the upcoming release of xorg-server-1.6. As X.Org maintainer on our distribution, I'm preparing the new release today, which means all previous drivers will stop working. All opensource drivers can be rebuilt, Nvidia's drivers can be updated to the latest versions for xorg-server 1.6 support, but AMDs catalyst has no support. There's also no word on upcoming support for xorg-server 1.6, so I guess we'll have to wait for Ubuntu's release again to get a leaked driver that has xorg-server 1.6 support just like when the previous version of Ubuntu was released.

These issues make maintaining a binary, non-free driver a task nobody wants to pick up. Andreas maintains this driver because nobody else in the development team wants to do it, or has the hardware to do it. Now that he's not willing to maintain it anymore, we want to hand it over to the people who have to use it: the community. This can be trusted users who can put binary packages in the community repository, or regular community users who can place this package in AUR as PKGBUILD.

We don't complain about the quality of the driver itself. As your article says, this driver has improved a lot over time. We're complaining about the information given to us as distribution.

[tulcod]

I think this is a difficult decision to take: it is good to make a statement against AMD, but then again, Arch isn't the biggest distro at all (although I have to admit they're certainly growing). And in the end, some user in the community will post packages for the binary driver anyway.

Let's see if AMD will respond to this.

[Redeeman]

I think this is a difficult decision to take: it is good to make a statement against AMD, but then again, Arch isn't the biggest distro at all (although I have to admit they're certainly growing). And in the end, some user in the community will post packages for the binary driver anyway.

Let's see if AMD will respond to this.

they might respond, but will they DO anything? i hardly think so... they cant even take the time to alter some broken text on the driver download site, despite requests for them to do so...

AMD/ATI hardware is simply useless until the free drivers supports what you need, simple as that.

[Vadi]

Gooooood jooooob.

Making the already pitiful driver even harder to use (less quality maintenance). Clearly, they aren't quite caring to the compositing and gaming users.

Only shooting themselves in the foot imho.

[susikala]

they might respond, but will they DO anything? i hardly think so... they cant even take the time to alter some broken text on the driver download site, despite requests for them to do so...

AMD/ATI hardware is simply useless until the free drivers supports what you need, simple as that.

It would be nice if people just stopped using every opportunity to take a bash at AMD/ATI; this shows of a really Windows-user mindset, i.e. "it should just work", regardless of how ugly or buggy the solution is. If you want to things to "just work" and that is your main or only concern, you are advised to use Windows -- most Linux users, I think, would (albeit reluctantly) admit that on this point, Windows excels. Everything works. And what doesn't work you _can_ usually get to operate using some messy solution. It's this monotonic, puerile view that makes Windows (for me at least) so boring to use. Everything works, and every other aspect is pure shit.

But if you're interested in those other aspects too, like performance, security, portability (eg KMS in FreeBSD?), the ability to read and learn every last line of source code that runs on your computer, and change it or fix it to fit your needs, then you would show great appreciation for AMD/ATI's work in the field of graphics. The FOSS drivers are getting better astoundingly fast, and the new Linux graphics infrastructure will allow us to run our systems even more securely and get rid of lots of bugs and slowdowns.

So I think, in general, while Arch Linux's decision to drop Catalyst is, as appears from JGC's post, quite justified and in place, this doesn't subtract from the fact AMD/ATI is currently one of the companies investing the most in open source graphics.

So again, big kudos to AMD/ATI for the hard work -- and sorry for relatively offtopicing.

[muep]

I really think the Arch devs made a good decision while handling this.
Especially considering that they made a public statement that clearly explains
the reasons, and maybe also makes another request for better drivers to AMD.

As a Fedora user, I have been having the same kind of problems as Arch users
lately. When Fedora 9 shipped Xorg-server 1.5 (an RC version, but still one
that had a stable and published ABI, identical to the final 1.5), it took over
half a year to get working drivers for it. Only the release of
Ubuntu 8.10 which had the same Xserver version seemed to make Catalyst support
Xserver 1.5, and even then it took a couple of months (The special beta driver
for Ubuntu didn't either work for everyone).

I know Fedora and Arch Linux aren't officially supported distros for the driver
as Ubuntu is, but I'd really want AMD to start supporting at least the latest
stable releases of X.org, Linux and other key components of the Linux
graphis stack which they depend on, and release their driver in a form that can
easily be packaged for any distribution. The issues which Arch Linux, Fedora,
and probably many other smaller distributions are facing are completely
unnecessary, and they make many distribution developers very busy fixing stuff
at their distribution, when the work could be done just once at AMD's end.

Even after fixing the packaging and component version support issues, there are
some issues in the actual driver that would be very nice to have fixed.
However, they are completely moot, if one can't actually use the driver at all
due to incompatibilities.

Please, AMD, now hurry with Xserver 1.6 support already! I'd like to keep using
your products with Fedora 11, as I currently happen to be able to with Fedora
10.

[sundown]

It's not something new. For instance you can't install the driver on a sidux kernel, although for a differnet reason :)

[Lattyware]

I would like to give some more explanation about the situation with Archlinux and Catalyst.
First of all, our current maintainer, Andreas Radke, is subscribed to AMDs beta mailinglist. He receives information and updates to the drivers before they go public. The mailinglist is very low traffic and almost no useful information is given. The driver updates that are announced on this list lately don't show much improvements in the driver itself, but rather "packaging improvements".
Without ugly workarounds that involves symlinking or editing a binary blob using perl or sed, it's not possible to make the current catalyst drivers working on a non-multilib distribution that doesn't ship /usr/lib64. Symlinking /usr/lib to /usr/lib64 is ugly, while the other solution involving editing the binary driver, violates the license agreement we have with AMD (we're not allowed to redistribute modified drivers). Setting LIBGL_DRIVERS_PATH or LIBGL_DRIVERS_DIR environment variables as we did before no longer works with these drivers somehow, the DRI drivers are always loaded from a path where they shouldn't be installed.

Another point is the upcoming release of xorg-server-1.6. As X.Org maintainer on our distribution, I'm preparing the new release today, which means all previous drivers will stop working. All opensource drivers can be rebuilt, Nvidia's drivers can be updated to the latest versions for xorg-server 1.6 support, but AMDs catalyst has no support. There's also no word on upcoming support for xorg-server 1.6, so I guess we'll have to wait for Ubuntu's release again to get a leaked driver that has xorg-server 1.6 support just like when the previous version of Ubuntu was released.

These issues make maintaining a binary, non-free driver a task nobody wants to pick up. Andreas maintains this driver because nobody else in the development team wants to do it, or has the hardware to do it. Now that he's not willing to maintain it anymore, we want to hand it over to the people who have to use it: the community. This can be trusted users who can put binary packages in the community repository, or regular community users who can place this package in AUR as PKGBUILD.

We don't complain about the quality of the driver itself. As your article says, this driver has improved a lot over time. We're complaining about the information given to us as distribution.

Until I read this post, I thought this seemed like a rash move at best, but I think this is Phoronix's writeup. The writeup appears to imply this was done as the quality of the driver is bad, which seems odd considering how much ATI have put into the driver recently (it would seem odd to remove it *after* they start improving it, not to mention when nVidia don't do much better).

Shame they apparently have been keeping it in bad shape. That said, any closed software is never going to gel perfectly with any Linux distro, really.

Disclaimer: I am an Arch x64 user with an nVidia card.

[mutlu_inek]

It would be nice if JGC's official statement in this thread would be elevated to an addendum of the official news item to make the situation more clear. It is great that phoronix uses its visibility to give these issues a platform that might be heard by companies and commercial developers. Thus, I think it is important that the position of Arch Linux is made as clear as possible.

[BlackStar]

As an Arch user using fglrx, I cannot but applaud the decision. We cannot have a single, badly written binary driver holding back the whole distribution.

Full disclaimer: I will continue using the blob, as I need OpenGL 2.0+ for my work, but I don't expect the distro team to maintain the blob for me. The portion of the userbase that requires fglrx will be able to publish custom PKGBUILDs that get the job done.

Now, the real question is "how much time before we see GLSL support in open drivers?" Because in all other regards, they are ahead of fglrx.

[SyXbiT]

I've been using Arch linux exclusively for about a year now. If you know your way around linux, it's fantastic.

Before using Arch I used Debian (for about 2 years) My first laptop with debian was a Dell witih an x1400 ATI card. The driver was so atrocious (I couldn't even play a DVD. there were screen refreshing issues. and this was before compiz came out). Because of ATI, I sold the whole laptop and bought a new one with an nvidia card. Since then, I've made a very concious effort to ONLY buy laptops with nvidia.
In the last 4 years I've had 5 laptops (with just the first using ATI)

I then built a desktop. The decision was obvious, I bought an nvidia 8800GTS
Everything was great. Compositing/compiz worked flawlessly, DVD playback was great etc..

I'd kept hearing about ATI improving, and catching up, and phoronix would often imply they they were on feature parity with nvidia.
(note to self, features like crossfire etc.. don't matter. we care about Xserver support, and compositing)

So, I saw a good deal for an ATI 4870, and thought "hm... I'll give them another chance"

It was the biggest mistake in my computing life.
Immediately I went from a perfectly working desktop, to no compiz, crashes, and false promises.

I remember seeing how the Arch Linux devs delayed XServer 1.5 JUST because they were waiting for a catalyst driver that supported it.
This is unacceptable. The whole distro had to wait for them.

Then, phoronix started talking about catalyst h264 hardware acceleration, and how it was great (and about to be released). But guess what, yet again, it never happened. Who knows if it ever will.

All I know is nvidia delivered. No, they didn't promise, they just delivered.

Needless to say I got burned twice, and it won't happen again.
I sold the 4870 on ebay, and bought a GeForce GTX 260.

It's really pathetic that my previous $300 card couldn't even do video playback under compiz (and compiz has been around for ages!)

So, now I'm an Arch Linux x64 nvidia user. And I'm really, really glad.

I've sent Michael a couple of PMs saying that he should stop doing al these gaming benchmarks comparing ATI/Nvidia.

Most of us don't care if ATI gets 3fps more in some game.

A decent graphics card review would cover the IMPORTANT THINGS. namely compiz support, compositing, 2D performance,vdpau support, Power managing, ease of install, Xorg tweaks, support in x64, delays in support of XServer, DRI2 etc..

Those are the deciding factors when buying a graphics card for use in Linux

Most of us don't game in linux. We use linux for other reasons.

ATI, are you listening?

[Zhick]

All I know is nvidia delivered. No, they didn't promise, they just delivered.
AMD/ATi also never promised anything. It's realy Phoronix who try to catch a sensation by making things up (maybe that's a little bit to harsh) and predicting features etc only by guessing. It's worked good for them so far, they've got the attention of many ATi/fglrx users.
It would be nice if JGC's official statement in this thread would be elevated to an addendum of the official news item to make the situation more clear. It is great that phoronix uses its visibility to give these issues a platform that might be heard by companies and commercial developers. Thus, I think it is important that the position of Arch Linux is made as clear as possible.
HAHAHA Michael admitting a mistake/making a correction. Yeah sure, that's gonna happen.HAHAHA
It's sad but Phoronix realy is the THG of Linux-News-sites. I'm not saying everything Phoronix does is bad, for example their coverage of the foss-(ati)-drivers is very good, but there also are lots of very inaccurate etc articles.

[Dragonlord]

Uhm... I'm running 4870+fglrx+amd64... no problems here. What's the deal? It's a binary blob. Binary blobs never work consistent across different distros since every distro has their own idea about file system layout ( yep, many shit on FHS or do their own variation ) and kernel patching and what not else. They can't make it free. So wait until the free drivers catch up and stop shitting on them. It's NOT GOING TO HELP ANYBODY ( except harming Linux in general ). I like people with lack of foresight...

[SyXbiT]

My comments were not just to rag on ATI.
I understand they are making an effort on the OSS front.
BUT, I will not buy them until they fix the things I mentioned.
So, it's more like a helping hand

ATI, do you want more business?
Do you want to sell more graphics card to linux users?

Then fix compiz, VDPAU, and start supporting XServers at release ( or very close to)

If you do that, you'll have my business. If not, you won't

It's that simple

[bridgman]

Michael admitting a mistake/making a correction. Yeah sure, that's gonna happen

It's probably worth reading the mailing list referenced by the article before picking on the writing. JGC's post was very clear and level-headed but it came *after* the article (and the article reads pretty much like the mailing list entry).

I personally try to avoid symlinks (probably because I keep getting in trouble with them ;)) but right now I think the driver requires them for certain distro configurations. Much as I sympathize, I hope the Arch folks would not stop packaging fglrx just because of the requirement for symlinking.

That said, if they're moving to server 1.6 and we don't support 1.6 yet then packaging the driver for upcoming Arch versions doesn't make a lot of sense. Not sure if it's still worth packaging for users running older versions or if most Arch users run the latest bits anyways (ie there would be too few people running pre-1.6 to be worth the effort).

The only thing I didn't really understand was that the first part of each post basically said "we've stopped packaging because of technical issues" while the second half of each post said "we hope someone else will take over the packaging". I think that translates into "it's a pain in the butt so someone else should take it over", which is fair.

[kraftman]

All I know is nvidia delivered. No, they didn't promise, they just delivered.

You must be kidding. Just look here:

http://www.nvnews.net/vbulletin/showthread.php?t=115916

You sound funny now.


I'm really happy to hear that Arch Linux devs abandon support for catalyst. Should we blame them or AMD? I don't think so. AMD supports only few distros and Arch Linux devs wouldn't change their philosophy just, because binary blob doesn't work. If someone wants working fglrx just choose Kubuntu. I hope we'll get the new X server earlier than before now :)

[SyXbiT]

you cannot compare 2D performance issues to what ATI has done
I'm thrilled with my nvidia card.
ATI, after 2 years of compiz being around, still can't support it.
Don't have VPDAU, and still have MAJOR tearing when playing videos (even in metacity)

you're the one who's making ME laugh

obviously all drivers have bugs. Bug you can't compare

[kraftman]

you cannot compare 2D performance issues to what ATI has done
I'm thrilled with my nvidia card.
ATI, after 2 years of compiz being around, still can't support it.
Don't have VPDAU, and still have MAJOR tearing when playing videos (even in metacity)

you're the one who's making ME laugh

obviously all drivers have bugs. Bug you can't compare

I used nvidia cards for years and I can say their drivers sucks a lot. I'm a lot happier with Xorg Radeon driver. What makes you laugh exactly? I commented what makes me laugh in your post. What AMD promised? It looks YOU missed a point.

[mutlu_inek]

Michael admitting a mistake/making a correction. Yeah sure, that's gonna happen.

This is not about admitting any sort of "mistake," but using Phoronix as a platform to talk back to the big silent wall that is AMD... which, I believe, is what Michael is trying to do.

[SyXbiT]

I used nvidia cards for years and I can say their drivers sucks a lot. I'm a lot happier with Xorg Radeon driver. What makes you laugh exactly? I commented what makes me laugh in your post. What AMD promised? It looks YOU missed a point.

I agree with you. The OSS ATI drivers are pretty good, and if they do all you need, they're the best bet.
I, however, wanted Compiz, and openGL etc..
This requires me to use a binary blob.

I wasn't comparing radeonHD to nvidia, I was comparing catalyst to nvidia, and I believe nvidia has less problems

[kraftman]

I agree with you. The OSS ATI drivers are pretty good, and if they do all you need, they're the best bet.
I, however, wanted Compiz, and openGL etc..
This requires me to use a binary blob.

I wasn't comparing radeonHD to nvidia, I was comparing catalyst to nvidia, and I believe nvidia has less problems

Yes, I agree. I just wanted to point that nvidia sometimes promises too and does nothing. If I say fglrx is better than nvidia binary blob it will be really funny ;) Sorry for being rude before.

[MostAwesomeDude]

I find it very strange that radeonhd is being recommended, not radeon. Surely the Arch maintainers haven't forgotten that radeon supports *all* Radeons and not just the new ones, and that it's also the one that's actually build-tested?

As to the query near the beginning of the topic regarding GLSL, the state of it is that GLSL will probably show up in classic Mesa when somebody actually cares enough to write it, and in Gallium when I start supporting programmable shaders.

[Kano]

The problem is, even when you begin to ignore fglrx, then users will still demand it because of better 3d performane (or they have to tell em to buy a game console instead). So they have to use the ati installer directly, if thats better or not depends on point of view.

[Michael]

HAHAHA Michael admitting a mistake/making a correction. Yeah sure, that's gonna happen.HAHAHA

Actually, an addendum was made to that news post to reflect the mentioned forum post.

I have no problems fixing mistakes when they are discovered.

It's sad but Phoronix realy is the THG of Linux-News-sites. I'm not saying everything Phoronix does is bad, for example their coverage of the foss-(ati)-drivers is very good, but there also are lots of very inaccurate etc articles.

When you find an error or something that is inaccurate within an article, feel free to let me know and usually I'll work to resolve the situation. It's more about a lack of manpower than anything else. When problems are discovered, I do my best to resolve them. Pardon if not all articles and news posts are absolutely perfect. On top of writing 300+ articles and 700+ news posts per year by myself basically, and everything else related to Phoronix (not to even count all of the Phoronix Test Suite development time), I also have a full-time position at another company... As a result, sometimes I am just exhausted when writing.

[Michael]

Then, phoronix started talking about catalyst h264 hardware acceleration, and how it was great (and about to be released). But guess what, yet again, it never happened. Who knows if it ever will.

You're not the only one frustrated over the AMD XvBA situation. The fact of the matter is that it was delayed internally by AMD from its Q4'08 target.

[Dragonlord]

you cannot compare 2D performance issues to what ATI has done
I'm thrilled with my nvidia card.
ATI, after 2 years of compiz being around, still can't support it.
Don't have VPDAU, and still have MAJOR tearing when playing videos (even in metacity)

you're the one who's making ME laugh

obviously all drivers have bugs. Bug you can't compare
Hm... I must be special I guess.
1) I play games full screen with OpenGL and other stuff... works
2) I play videos ( fullscreen too )... works
3) I develop using GLSL a game... works
4) I blend with blender fullscreen ( pure OpenGL app )... works

[JGC_]

First of all, let's state that I'm a bit disappointed that a discussion on an open development list hits Phoronix as a news article without asking anyone from the distribution about comments.

Second, when this discussion started, Andy was already pissed off for some months with the way packagers are supported and how the driver works. He has been telling me "I'm giving AMD two more months and then I won't maintain this driver anymore if things don't improve" a while ago. Two months have passed and Andy decided to post on the public development mailinglist that he doesn't like to maintain the AMD Catalyst driver anymore and that we should look for community people to invest their time in this driver.
Note that the situation is not just AMD-specific. Nvidia also has this problem more or less, but they have the advantage that 50% of our developers have an Nvidia card and their drivers are in a better state and are not limited to X.Org architecture.
Remember that none of us gets paid to do Archlinux development. We're all spending spare time in this distribution. We maintain this distribution because we like to do so. When we lose interest in a package and don't want to maintain it anymore, it's better to drop the package to a group of people who are forced to use it, or who do want to maintain it.

As for xorg-server-1.6: This is a side-effect of the problems we face when packaging AMD's catalyst for our distribution. When our catalyst maintainer posts something on the AMD secret beta list, all he gets is silence. There's no information about what's coming up, when what will be supported and when what improvements will be made to the driver.
As it looks now, we're dropping the driver completely now because of compatibility issues with server-1.6, but won't add it back to the main distribution when it receives server-1.6 support. The community can pick up this task if they really want this driver, as there's no maintainer willing to maintain this driver anymore.

As for the radeonhd driver mentioned on the mailinglist as alternative: these people don't maintain the drivers related to X.Org and don't know which driver belongs to which chip exactly. For some newer chips the radeonhd driver works, but for most others the ati driver is the driver they need. Making fun out of someone on a public news site because he inofficially states that the radeonhd driver is an alternative is not a nice thing to do.

Then again, arch-dev-public is a development discussion list, which is public readable. It's not an announcement mailinglist. Articles like these really make me think whether I want to post something to arch-dev-public or just the private arch-dev list.

For your information, my draft for the xorg-server-1.6 announcement went to the private list and won't show up on the public list until all work on xorg-server-1.6 and its drivers are done.

[yoshi314]

http://www.archlinux.org/pipermail/arch-dev-public/2009-February/010398.html
My reason for it to go to community is so a TU can dedicate attention to it, Andreas does not want to waste time on fixing this drivers as they are of no particular interest to him. Maybe a TU in community can apply patches to make the x86_64 9.2 release to work. But well, I'm fine with moving them to AUR, there are some distributions that do not provide catalyst at all in any form, comes to mind paldo GNU/Linux.i think this sums up what i wanted to write anyway.

i haven't been using fglrx somewhere since 8.35. i tried it few times after it improved opengl performance, but i didn't like it.

[susikala]

First of all, let's state that I'm a bit disappointed that a discussion on an open development list hits Phoronix as a news article without asking anyone from the distribution about comments.

Well, I don't see a problem posting it as a news item on Phoronix. Your statement is a bit contradictive, first you talk about 'a discussion on an open development list', and then you say a maintainer decided to drop the package. Either you say it's not decided yet and still being discussed, and then this news item is indeed a bit overhasty, or you stand behind the (justified) decision of the maintainer. But I don't see how you can hold both points.

Also, this is quite relevant information for some Arch Linux users and I'm not sure everyone of them reads that list, so that news item is certainly in order.

[MostAwesomeDude]

As for the radeonhd driver mentioned on the mailinglist as alternative: these people don't maintain the drivers related to X.Org and don't know which driver belongs to which chip exactly. For some newer chips the radeonhd driver works, but for most others the ati driver is the driver they need. Making fun out of someone on a public news site because he inofficially states that the radeonhd driver is an alternative is not a nice thing to do.

I'm the only person that's mentioned it so far, and I've hardly been making fun. After all, if Arch chose to force radeonhd above radeon, they wouldn't be the first; Ubuntu does that as well. I'm just saying that I find it odd.

[RealNC]

It's also odd that there are two drivers. There should be only one.

[jonnycat26]

I'd kept hearing about ATI improving, and catching up, and phoronix would often imply they they were on feature parity with nvidia.
(note to self, features like crossfire etc.. don't matter. we care about Xserver support, and compositing)



I went down the same path as you. I was a longtime Linux + NVidia user, and when I needed to upgrade my video card I fell for the Phoronix propaganda about ATI.

Not that the card itself is bad, for a 70 dollar card (HD4650). it runs rings around any NVidia at a similar price point. It runs cool, and the power draw is low. But none of that matters if the drivers don't work, and fglrx just doesn't.

I'm back with NVidia. Maybe I'll pull the ATI out of the closet at some point, when ATI puts out a few stable fglrx releases in a row, but who knows how far ahead NV will be at that point.


I've sent Michael a couple of PMs saying that he should stop doing al these gaming benchmarks comparing ATI/Nvidia.


I don't think Michael has to stop, but the fact that the results don't mention that fglrx is... stability challenged... is kind of misleading.

[BlackStar]

As to the query near the beginning of the topic regarding GLSL, the state of it is that GLSL will probably show up in classic Mesa when somebody actually cares enough to write it, and in Gallium when I start supporting programmable shaders.

Ah great, I know whom to bug now :D

I wouldn't expect Mesa to pick GLSL support now that Gallium has been merged, the question was more in the spirit of "when will we have a prototype to test?" (AKA "are we there yet?") Annoying, I know, but this stuff will really make the FOSS drivers viable for all but the most specialiazed workstation users (so thanks for your work!)

A common OpenGL state tracker and GLSL compiler for all FOSS drivers - now that's something to look forward to.

...and sorry for derailing the thread.

[bugmenot]

I think it is good that they seem to take radeonhd and not radeon, not only because of the audio over hdmi support.

[RagingDragon]

[...]
I've sent Michael a couple of PMs saying that he should stop doing al these gaming benchmarks comparing ATI/Nvidia.

Most of us don't care if ATI gets 3fps more in some game.

A decent graphics card review would cover the IMPORTANT THINGS. namely compiz support, compositing, 2D performance,vdpau support, Power managing, ease of install, Xorg tweaks, support in x64, delays in support of XServer, DRI2 etc..

Those are the deciding factors when buying a graphics card for use in Linux

Most of us don't game in linux. We use linux for other reasons.

ATI, are you listening?

Personally I'd like the Phoronix reviews to cover both gaming and other issues. I like to play games (though my time is more limited now :(), but I also use my computer for other things.

I'm running Arch64 and have an AMD 4850 - at the time I bought it the AMD/ATI drivers sucked less than the Nvidia ones. However, today Nvidia's GT200's are very tempting...

[kensai]

Well, the news were well laid out by phoronix, is true I started ranting about the quality of this drivers. Since I can, because I own an nvidia and an ati card. In both machines I use Linux, and have never had trouble with nvidia, on the contrary with ati, even if the driver has become better, it is still too buggy. But, the things that were omitted were when I said, I do cared about the drivers:
Feb 25
PS: I do use this drivers, so I do care about them, but this is a choice that had to be made.
Feb 26
My reason for it to go to community is so a TU can dedicate attention to it, Andreas does not want to waste time on fixing this drivers as theyare of no particular interest to him.
OK, I'm in a bit of a puzzle right now. As I have just installed the 9.2 catalyst in arch64 and tested them all running fine.
Here, the symlinking problem was the only thing still bugging me.
After thinking it all over, I don't want Arch users to feel
uncomfortable or affected by our decisions on this, I am all in for lets make things work so the user benefit at the end. So I am up for
reconsideration of this.
Mar 1
Yeah, I tried to find a solution of how to work this out and keep catalyst at least in community, but it seems, ATI/AMD just aren't doing a thing to improve the situation. Yet, this package is so important for some users of Arch Linux that a TU should be the one in charge of them.

You see, I have tried, but in the end there is nothing I could do, but to start all over again and thing, this should be removed from the official repositories.

- Cheers

Eduardo "kensai" Romero

[iVistux]

What Arch is doing is not that spectacular, see Fedora, etc. refusing to let the driver settle in their repositories. The driver was always available by 3rd-party-repositories or other sources, now Arch created the same situation.

Unfortunately the proprietary driver is not good for everybody, but its stand is still better then years ago, he's made for people buying Linux-Notebooks from Dell or users using a FireGL/Pro. Fedora/Arch/Gentoo are often too bleeding edge for this driver.

[kensai]

What Arch is doing is not that spectacular, see Fedora, etc. refusing to let the driver settle in their repositories. The driver was always available by 3rd-party-repositories or other sources, now Arch created the same situation.

Of course is spectacular, Fedora has always rejected closed source applications. We have never done so, a lot pf us use closer source applications when they are better than the open source implementation. I do not use Windows, because it is not better than Linux, but I do use nvidia driver and ati driver because they are better than the open source version. Still in the case of ATI, they are a pain to deal with, and haven't been working on non-multilib distributions since 8.12 that is two releases behind of what is currently used. So this is not a matter of politics, about we taking a stance against closed source like fedora, is taking a stance aginst closed source applications that does not want to play fair with Linux.

[remm]

you're the one who's making ME laugh

Apparently, you are happy using a proprietary binary driver that never actually worked that well for me. So that does not work for me, I want a good free driver.

Besides, there's a limit to the level of arrogance I can tolerate from people, if you see what I mean ...

[Kano]

spectacular would be a fglrx binary driver that works as expected ;)

[Redeeman]

spectacular would be a fglrx binary driver that works as expected ;)
Who knows.. you might find someone that bets on such things.. i'd say you should get about the same odds as that bet about elvis falling down on the empire state building.. :)

[jeffro-tull]

So, if Mark Shuttleworth sneezes, Phoronix does an article on Ubuntu. For Arch to get an article, they need to do something controversial like treating Catalyst like a second-class citizen?

Anyways, I can't say that I blame them. I haven't used Catalyst in over a year (wait, no. I haven't use Catalyst period. I haven't used the fglrx kernel module in over a year), but I remember quite vividly jumping through hoops and/or waiting a couple months and/or downgrading distributions to get it to work. If the situation hasn't changed (regardless of whatever work they've been doing), then I can't blame the Arch dev's for not wanting to deal with it.

[alec]

At current stage, I don't care much about 3d performance. So the moment radeonhd has working powerplay, I'm switching.

Arch64+fglrx, compositing SLOW

[susikala]

So, if Mark Shuttleworth sneezes, Phoronix does an article on Ubuntu. For Arch to get an article, they need to do something controversial like treating Catalyst like a second-class citizen?



Last I checked, Ubuntu had over 2000 hits per days on Distrowatch. Arch Linux is apparently barely at 600, that's the difference for you.

[grantek]

Ubuntu is "Linux for human beings", Arch is "Linux for computers and people who aren't afraid of them". Having said that, I use Ubuntu because I'm a long-time Debian/dpkg fan and I haven't had a reason to install a new OS on my computer since I first loaded up Ubuntu :)

[Kano]

I don't get whats so great with Ubuntu. The kernel is good, I help em personally to support some specific kinds of hardware better in case something is broken because I reuse the kernel for my own distro. But besides the kernel there is not much difference when you would compare Debian 5.0 default GNOME install with Ubuntu 8.10. GNOME looks as ugly as usual on both. Ok, I prefer KDE 3.5 series, fast, no problems compared to new KDE 4 series with some gfx drivers. From points of support U always loses against pure Debian. Their main repository is really small, maybe 1000 vs 23000 packages in the rest. And full support you get only for the main one. Even if Debian would fix security issues with other packages (outside main) they do not get into Ubuntu, maybe some with universe security updates.

[RealNC]

Ubuntu is great for keeping the "z0mg it doesn't work 1 d3mand y00 fixor it NOW or else" crowd out of Debian :D

[deanjo]

Last I checked, Ubuntu had over 2000 hits per days on Distrowatch. Arch Linux is apparently barely at 600, that's the difference for you.

If articles and reporting are done purely on popularity then there should be far more articles done on openSUSE then Fedora.

[kraftman]

Ubuntu is great for keeping the "z0mg it doesn't work 1 d3mand y00 fixor it NOW or else" crowd out of Debian :D

Yes, great picture :D But many advanced Linux users are using Kubuntu too of course.

@Kano

Is Debian such noobie resistant as Ubuntu is? :> In my opinion some trivial simplifications make a lot of difference in this case. Btw. the worst thing in Ubuntu is they chose gnome as primary desktop environment. There was a time when KDE4 wasn't stable, but now it is. It's great time to change IMHO.

@deanjo

It works this way - more clicks on distro title = it's more popular on distrowatch :)

[RealNC]

There was a time when KDE4 wasn't stable, but now it is. It's great time to change IMHO.

Imagine the flamage on the lists and forums...

[deanjo]

@deanjo

It works this way - more clicks on distro title = it's more popular on distrowatch :)

I realize that, that's why if you were going use distrowatch as a measuring stick for articles then openSUSE should have more articles then Fedora.

[kensai]

On topic, I just wrote officially that the drivers will be dropped. Thanks for the support on this matter, (Arch) / Linux community.

[energyman]

who cares? Archlinux is broken by design - a good example of how not to do a distribution.

[kraftman]

who cares? Archlinux is broken by design - a good example of how not to do a distribution.

Arch is an ideal distribution for someone who doesn't want to spend all day compiling apps and who wants rolling distro.

[Melcar]

I think in the end this will only hurt Arch, assuming they have a big ATI user base to begin with. Plus, I'm sure someone in the community will eventually step up to the challenge.

[energyman]

Arch is an ideal distribution for someone who doesn't want to spend all day compiling apps and who wants rolling distro.

wrong. With its broken versioning and downgrading it is the worst possible distribution for anybody who wants rolling updates.

[kraftman]

wrong. With its broken versioning and downgrading it is the worst possible distribution for anybody who wants rolling updates.

Which is better if I don't want to compile packages?

[Arch64]

wrong. With its broken versioning and downgrading it is the worst possible distribution for anybody who wants rolling updates.

Broken versioning ? says who ?

Troll detected.

[kensai]

@energyman, you just want to take the time to tell us your achievement frustrations?

@Melcar, 10% of our users use ati open source driver. 9% use catalyst, so no not many users use catalyst, not even ati at all, nvidia is used by 43% of the users.

[grantek]

@energyman, you just want to take the time to tell us your achievement frustrations?

@Melcar, 10% of our users use ati open source driver. 9% use catalyst, so no not many users use catalyst, not even ati at all, nvidia is used by 43% of the users.
Just curious, how many use nv/nouveau?

[Hephasteus]

Just curious, how many use nv/nouveau?

I tried the nouveau the other day. The kernel didn't support it right. I could try it again but I'm trying to keep things to breaking X once a week as it stands.

I know everyone is really frustrated with video driver support on linux. I see some really great work and progress happening and I'm really excited about it. I just downloaded some fedora testing files and broke X but good. After messing with it for about an hour it finally started working. Seeing what the heck happened I find that my computer was being automatically probed for 5 screens. Sure it looked broken and frozen but things are happening.

Microsoft is going to try to generate alot of hardware sales with windows 7 and if it happens people will go back to kissing it's butt again. If it doesn't then linux probably go from 10 plus million desktops to 50 60 million desktops this year. :eek:

[kensai]

Just curious, how many use nv/nouveau?
13% use nv driver. But we have no statistics of the noveau, just yet.

[jeffro-tull]

so, the consensus is "more people use Ubuntu, so that should get the most coverage. If something really big happens with another distro, then it'll get a mention."

Hey guys? That doesn't make sense. By your own logic, this site shouldn't even exist. Windows has orders of magnitude more users than any Linux distribution. So, since more people use it, that should get coverage. Since so few people use Linux, as compared to Windows, it should rarely, if ever, get a mention.

I know, I know, there are many, many, many Linux distributions, and it would be impossible to cover all of them in any kind of great detail. All I'm saying is that news about portage picking up support for solving USE-flag dependencies is more interesting than the code-name for the version of Ubuntu that'll be coming out next year.

[deanjo]

I
Microsoft is going to try to generate alot of hardware sales with windows 7 and if it happens people will go back to kissing it's butt again. If it doesn't then linux probably go from 10 plus million desktops to 50 60 million desktops this year. :eek:

Actually the main focus on Windows 7 is not about added features that require hardware upgrades, it's quite the opposite, It's more about using the present hardware smarter (same can be said with OS X 10.6). MS even has come out with Warp10 in Windows 7 that allows cards that previously were not able to use all the aero effects but now can with the CPU picking up the slack for those effects when needed. It's also been created with systems with limited capabilities in mind such as the netbooks. Then there is also DX 11 with GPGPU support which is again focused on making the most of your hardware. While it may finally prompt some users to upgrade 3+ year old XP systems to run it more effeciently, it is quite at home running on a system such as a AMD 1700+ with 1.5 gig of ram. Minimum system specs are still the same as Vista however it does a better job with lower end hardware (as one journalist found out here http://blogs.zdnet.com/Bott/?p=672).
Many people may upgrade their XP systems now that they feel it's a worthy upgrade but I would imagine that percentage would be much smaller with the people presently running Vista on their systems. Those people will more then likely just get the OS upgrade and stick with the present hardware since it is more then ample to run Win7. Of course OEM's are hoping like hell Win 7 is a success as it may finally get those XP users to switch and that is a welcome chunk of the market in these hard times.

[b15hop]

Energyman:
What distro do you use then?

[kraftman]

@deanjo

Said shorter - they'll optimize aero etc... There's no logic in windowses or macoses. First they're unoptimized crap which demands high end hardware to run properly and then ms and apple release newer versions without meaningful improvements, but just optimize them. Logically it should be different order.

[Stedevil]

On topic:

I would say that the original article seems to leave a much harsher view than what the Arch devs actually seems to have, and more importantly, is quite unclear with the exact gripes they have. For AMD to progress, it's IMO important to give accurate and to the point critique of what and how they do things wrong.

Off topic:
Ubuntu is great for keeping the "z0mg it doesn't work 1 d3mand y00 fixor it NOW or else" crowd out of Debian :D

And the same Debian people saying that wonder why people in general view them as elitist pricks and chose Ubuntu instead... ;)

There are a lot of knowledgeable people tired of such a crappy attitude and instead prefer use eg Ubuntu. And biggest problem with Ubuntu IMO is that it depends on Debian to update packages which means to get your package up to date in *buntu you run the severe risk of having to fight above mentioned elitists and cross your fingers they dont break your perfectly fine working software in their modified source .deb...

And people, dont even get me started on the Gnome/KDE issue... WTF... half the linux community hates KDE the other hates Gnome, and (K)Ubuntu offers both... so get the stick out of your hiny. For sane people, it's a non issue.

[kraftman]

@Stedevil

And people, dont even get me started on the Gnome/KDE issue... WTF... half the linux community hates KDE the other hates Gnome, and (K)Ubuntu offers both... so get the stick out of your hiny. For sane people, it's a non issue.

It is issue for Ubuntu newbies who don't know that KDE exists and they're trolling on Ubuntu forums, because they want features which KDE has, but gnome doesn't. Btw. I want to stop here. /Off topic

[yogi_berra]

Many people may upgrade their XP systems now that they feel it's a worthy upgrade but I would imagine that percentage would be much smaller with the people presently running Vista on their systems. Those people will more then likely just get the OS upgrade and stick with the present hardware since it is more then ample to run Win7. Of course OEM's are hoping like hell Win 7 is a success as it may finally get those XP users to switch and that is a welcome chunk of the market in these hard times.

Off Topic: No one upgrades for the sake of upgrading. Until there is a valid reason for people to upgrade from XP (Like Microsoft ending security support) they won't move to anything new. In some cases even when that support ends XP will stay in place. I know of a place that is still using NT 4 as an interior file server and until the hardware dies they see no reason to upgrade it. That is not an uncommon situation.

On Topic: As long as Arch users are made aware of the situation by the Arch maintainers so no one ends up with a frustrating view of tty1 unexpectedly on upgrade, why is it news worthy? Package maintainers often get burned out (for lack of a better phrase) or disappear from other distributions or decide that upstream is too asshole-ish (ref: Debian's on going argument with Joerg Schilling over wodim) to work with so it really isn't new or even exciting.

[deanjo]

Off Topic: No one upgrades for the sake of upgrading. Until there is a valid reason for people to upgrade from XP (Like Microsoft ending security support) they won't move to anything new. In some cases even when that support ends XP will stay in place. I know of a place that is still using NT 4 as an interior file server and until the hardware dies they see no reason to upgrade it. That is not an uncommon situation.



That is far from the truth, if that was the case there would be no need for the retail versions nor the upgrade versions both of which sell well. What applies in the corporate/business sector does not apply in the home sector. People upgrade their OS's all the time. It happens in linux, OS X, windows, insert favorite OS here.

[yotambien]

That is far from the truth, if that was the case there would be no need for the retail versions nor the upgrade versions both of which sell well. What applies in the corporate/business sector does not apply in the home sector. People upgrade their OS's all the time. It happens in linux, OS X, windows, insert favorite OS here.

Honestly, I don't know anybody who upgraded their OS for the sake of it in Windows land. All the people I know who use a computer runs the operative system that came installed with it. Hardware is today much more affordable than some years ago so I expect this to be the general trend even more than in the past. Is it slow/have a tiny problem? Well, buy a new one (of course I don't share this way of thinking).

I don't know how well the retail versions are doing. However, the point is not whether they are doing 'well' but how do they compare to the number of OS preinstalled in new systems.

[RealNC]

Same here. I don't know a single person who upgraded to Vista. Not a *single* one. Everyone who uses Vista got it pre-installed when buying a desktop or laptop. No one upgraded to it.

[DestroyFX]

What is the problem with ATI drivers?
They don't give you the choice for no multilib drivers??

Did you know that the majority of the Linux user base use a multilib system when on x86_64?

After that, ATI are better and better on each drivers release and I tested both on my system:

Setup : Q6600 4.0Ghz, 8GB DDR2, two screen at different resolution (Xinerama required)

Nvidia 9800 GTX+ :
-I play Eve-Online full screen on my first screen and the rest of the display is laggy (laggy consoles, laggy windows....)
*No solution found, but it seem to be a problem with Xinerama : Thank Nvidia
-Sometime my X start to take 100% of one of my CPU core and the system is "frozen" until I ssh in my box to kill X
*Fix from Nvidia FAQ : start computer with maxcpu=1 or nosmp : Know problem since like five drivers release... WOW! This look so professional!... (this happen more when you use a lot of CPU power(who say BOINC?))

ATI 4870HD :
-I can play with two EVE Apocrypha (the next beta release that use heavily GLSL, and like 246 vec4s in there shaders uniform). Not only I can use two in same time with highest setting with bloom, HDR, but I can use a console and play some movies in the same time (And yes, the consoles don't lag... unlike with Nvidia cards)
-I don't have any X takeover with my ATI.

OK, for ATI, they have a bug in GLSL.. That bug is simple : the drivers did not report the good value for Varying and Shader/Fragments Uniform so OpenGL stuff that use GLSL can choke over that. ATI have this problem since more than two years and people just kept whine about that... WE (EVE-Online user wanting to get EVE Apocrypha working on ATI hardware) isolated and contourned the problem by modifying the way WINE use GLSL. I have posted the facts on the no so secret ATI bugzilla and know what? This problem is now internally fixed in AMD drivers (this took 1 week)...

You want thing working : STOP WHINING AND GO BUG REPORT (Detailed with log and any stuff useful... No whining on bugzilla please)
http://ati.cchtml.com

PS : Sorry for my bad english ;)

[alec]

I know a few, but they didn't pay for it, pirated.

[kensai]

You want thing working : STOP WHINING AND GO BUG REPORT (Detailed with log and any stuff useful... No whining on bugzilla please)
http://ati.cchtml.com

LOL, the problem has been reported upstream long ago. I know you are an ATI fanatic, well, that doesn't help, they have bad support for Linux period. They are improving, but the improvements are so slow that it might take a few years for them to release quality drivers like nvidia, and by then, the open source ati drivers hopefully will be better than the closed source one.

Yeah, nvidia has its issues, but are not as catastrophic as ati. And we are not whining, we are telling our users that by the time xorg-server 1.6 goes to the main repository and ATI hasn't come up with a driver which fixes the issues we are experiencing, which are no non-multilib support, and the compatibility with newer xorg we will remove them. We are a true 64 bit OS, and we follow that philosophy, we have a lot of users content with that, is having that philosophy a crime? I guess not, in fact it supports better the future, which is 64 bit architecture.

[yogi_berra]

That is far from the truth, if that was the case there would be no need for the retail versions nor the upgrade versions both of which sell well. What applies in the corporate/business sector does not apply in the home sector. People upgrade their OS's all the time. It happens in linux, OS X, windows, insert favorite OS here.

It is not far from the truth, I know of people that are still using Windows 98 as their OS. Their needs are met by it, their firewalls still work as do their anti-viral solutions, they see no reason to upgrade, so they don't.

[hunterthomson]

I am an Archlinux Lover. I have tried many other distro's (Ubuntu being the most used and it is vary good for what it is, Bloated) but I can't stand to use them anymore now that I have gotten a tast of what a Linux OS sould be like. Anyone who has not used Archlinux should try it out. Arch dose what you tell it to do, works and "Says Working". Unlike Ubuntu that just randomly brakes 90% of the time do to all the bloated packaged that I never use and the fact that it installs Every driver known to the Ubuntu dev's. Then 6 month latter the whole bug hunt starts agin. But, Ubuntu is the best distro if you know nothing about Linux.

I messed up and thought that ATI had good Linux support. I feel like a dumbass now. ATI suport is a joke. The catalyst dosn't work in SUSE, Debian, Ubuntu, Fedora, which is clames to support, without major tweeking. This is no joke, the ATI catalyst is complet crap and there is no hope for it getting better.

While it sucks that the Catalyst is not supported by Arch now and is left up to the comunity. I have the AUR package installed... That was not hard. The current maintainer has it set up nice and patched to work with 2.6.29 (Though Arch will be on 2.6.30 soon). The hard part was getting it to work. I first did it in Ubuntu becuase Ubuntu is officially supported. Learnd all the tricks to get it to work and then applied what I learnd to Archlinux. Now a month and a half latter I finaly found all the tweeks and pached Xserver and all this crap to get it to work.

So, I was pissed at first but ATI needs a wake up call. Good for you Arch devs. Drop support, I don't expect you to have to deal with this POS catalyst. I am never buying ATI ever agin. I am not even going to buy AMD. I don't even care if I have to pay 10 times more for nVidia or AMD has some super cool CPU. Now, Intel that is a company to support. I love Intel as much as I love Arch :)

[alec]

You are being unfair to ATI/AMD
Video part of linux, xorg, 3d and 2d stuff is a mess. Yes its a progressing mess, but it only means that driver devs need to keep up with changes in the mess.

And in the end we win, because oss driver gets more attention.

[tball]

I am and Archlinux Lover. I have tried many other distro's (Ubuntu being the most used and it is vary good for what it is, Bloated) but I can't stand to use them anymore now that I have gotten a tast of what a Linux OS sould be like. Anyone who has not used Archlinux should try it out. Arch dose what you tell it to do, works and "Says Working". Unlike Ubuntu that just randomly brakes 90% of the time do to all the bloated packaged that I never use and the fact that it installs Every driver known to the Ubuntu dev's. Then 6 month latter the whole bug hunt starts agin. But, Ubuntu is the best distro if you know nothing about Linux.

I messed up and thought that ATI had good Linux support. I feel like a dumbass now. ATI suport is a joke. The catalyst dosn't work in SUSE, Debian, Ubuntu, Fedora, which is clames to support, without major tweeking. This is no joke, the ATI catalyst is complet crap and there is no hope for it getting better.

While it sucks that the Catalyst is not supported by Arch now and is left up to the comunity. I have the AUR package installed... That was not hard. The current maintainer has it set up nice and patched to work with 2.6.29 (Though Arch will be on 2.6.30 soon). The hard part was getting it to work. I first did it in Ubuntu becuase Ubuntu is officially supported. Learnd all the tricks to get it to work and then applied what I learnd to Archlinux. Now a month and a half latter I finaly found all the tweeks and pached Xserver and all this crap to get it to work.

So, I was pissed at first but ATI needs a wake up call. Good for you Arch devs. Drop support, I don't expect you to have to deal with this POS catalyst. I am never buying ATI ever agin. I am not even going to buy AMD. I don't even care if I have to pay 10 times more for nVidia or AMD has some super cool CPU. Now, Intel that is a company to support. I love Intel as much as I love Arch :)

I agree 105 % the first part of your post. I have moved to Arch recently and I love it. Especially AUR :)

But I don't at all have had those bad experiences with ati you got. Which gfx card do you have? My 3650 have always worked with Ubuntu, Suse and Arch without any problems installing it. Though I have had some struggle booting to X with Ubuntu without the option DefaultDepth 24.

[hunterthomson]

You are being unfair to ATI/AMD
Video part of linux, xorg, 3d and 2d stuff is a mess. Yes its a progressing mess, but it only means that driver devs need to keep up with changes in the mess.

And in the end we win, because oss driver gets more attention.

If it is so hard then why dose Intel never have a problem with Linux suport? All there drivers for Wireless, Graphics,... are always rock sold and upto date. ATI need to retool there drivers so they don't rely on stuff that is going to change in the next kernel or Xserver. They need to make it more modular so changes can be made with less effort.

[hunterthomson]

tball: I have a Mobility FireGL v5700 wich is a HD3650...
Here is the suport thread I started on ubuntuforums for my laptop if you would like to see all the stuff I have to do to get it to work.
http://ubuntuforums.org/showthread.php?t=1166667

It looks simple now that I worte that thread but it took me over a month to figure it all out (the catalyst I just found the last few tweeks for) I consider having to use a patched Xserver a major tweek.

But ya was to genoral about the catlayst not working. It dosn't work with my card with any of the suported distors with out major tweeking.

But ya isn't Arch the s/h/i/t :)

Note, the Intel wifi driver problem is not a problem with the driver it is a problem with Ubuntu. It works fine in Arch. Flash is also a problem in Ubuntu but works fine is Arch. In Ubuntu when I play video games like supertuxkart the game thinks I am pressing joydown all the time but in Arch that is not a problem

[mirv]

If it is so hard then why dose Intel never have a problem with Linux suport? All there drivers for Wireless, Graphics,... are always rock sold and upto date. ATI need to retool there drivers so they don't rely on stuff that is going to change in the next kernel or Xserver. They need to make it more modular so changes can be made with less effort.

ATI have probably done that as much as they can - but then we start getting into standardising interfaces, which the kernel devs don't like doing, and is one of the key issues of fglrx / kernel incompatibility. They probably do have support easily for new stuff, but due to their development cycle it'll bake for a month or two before being released (by which time things can change again). That's a con - a pro is that we get updated drivers every month!

[hunterthomson]

ATI have probably done that as much as they can - but then we start getting into standardising interfaces, which the kernel devs don't like doing, and is one of the key issues of fglrx / kernel incompatibility. They probably do have support easily for new stuff, but due to their development cycle it'll bake for a month or two before being released (by which time things can change again). That's a con - a pro is that we get updated drivers every month!

Hum..., standardizing is generally a good thing. I don't know the reasons why the kernel devs don't want to do it though. Maybe they have a good one. Although, Linux is Open Source. They can just look at the code and talk to the devs and make a driver that runs cherry on all distos. The Xserver and Kernel are not disto spisific. The file system layout is though, so ATI should have in place a way for distro's to easly set where stuff should be installed.

The release cycal should not be based on things Like Ubuntu's LTR cycle. They should base it on owe... every other kernel and every Xserver. Also, from what I remember reading, the arch dev in charge of calalyst support and on the mailling list didn't think they were addressing any of the problems that were making it so hard to work with. So, new catalyst driver update every month was probably just adding to the problem not solving it.

[Ex-Cyber]

As I understand it, kernel devs don't like standardizing the internal interfaces because they don't want to be put in a position where out-of-tree concerns end up holding more sway over the internal kernel architecture than in-tree concerns. They want to be free to reorganize, tear down and rebuild anything they think they can improve within the kernel, without being pressured to support every version of every interface going back a decade or more. OTOH they take standardization of userland<->kernel interfaces pretty seriously.

[hunterthomson]

As I understand it, kernel devs don't like standardizing the internal interfaces because they don't want to be put in a position where out-of-tree concerns end up holding more sway over the internal kernel architecture than in-tree concerns. They want to be free to reorganize, tear down and rebuild anything they think they can improve within the kernel, without being pressured to support every version of every interface going back a decade or more. OTOH they take standardization of userland<->kernel interfaces pretty seriously.

Ok, thanks for letting me know. I guess that is a good reason and being open source should make up for changes in the structure. I meen they should let everyone know ahead of time that they will be overhalling something.