ATI dropping support for <R600 - wtf!?

[DirtyHairy]

Before going to sensible work again, I have to vent myself here with a little rant. For the starting point: I am the proud owner of a Thinkpad T60 equipped with a X1300 mobility which I got 2 1/2 years ago and which has served me well ever since. I use it mainly for doing calculations, latex, coding at work and for video and some light gaming in my free time. I am not a zealot when it comes to open source; I like the concept of open source software and try to use it whenever possible, but I don't regard the attempt to earn money with software as a crime and have no reservations when it comes to well-written closed source software.

When I got the device I installed the fglrx driver and was rather pleasantly surprised by it working stable and reliably - I had read other stories. There were some glitches, but I though to myself: "wait for the next revisions, they'll iron it out" and - nothing improved, some bugs went away, others came. This game has been going on till now, and at NO POINT IN TIME, a revision of the driver did all that it advertised, but the thing was always broken in some way or the other. Specifically, the most annoying areas were:

- Xv: working well for some months, then completely disappearing, and never working again with vsync or on all players (mplayer + xine + vlc) till this day.
- Basic OpenGL: I don't need it badly, but I have a graphics accelerator, so it'd better do it. The first revisions worked, but there was some lag between commands submitted to the processor and the display casuing an inacceptable lag between input and reaction in ALL OpenGL programs. I had to hack around by preloading a library that did some waiting I don't remember exactly before swapping buffers. This got better over time but never really went away until the new generation of drivers hyped on phoronix appeared.
- Antialiasing: this has suffered from the lag described above until this day. It seems better in the last revisions.
-Vsync in OpenGL: this never ever worked really reliably, causing the lag described above.
- Visual artifacts: numerous revisions of the driver would display heavy corruption or artifacts in different situations, ranging from random lines over blackened window borders to the thing called the "checkerboard of doom" in these forums
- ... I could continue this list for some time ....

Now, AMD is dropping support completely for this generation of hardware from the driver, and, bottom line: for the whole life cycle of this support, the driver NEVER worked correctly. It was usable, but lacking many features that were advertised and that I expect from this type of hardware.

I am very concious of the opensource efforts, and appreciate the release of programming documentation but, honestly, in my oppinion this shouldn't be a big favour, but just the rule: if I buy something, I should be given the documentation on how to use it. Using hardware for me means being given the possiblity of writing code that uses it, not just running a set of pre-selected OSsed with varying support on top of it. The fact that this isn't the rule doesn't make it more of a favour.

Apart from releasing the documentation, the "open source effort" consists of paying one developer working on the open source code. Effectively, this means for me that ATI/AMD is offloading the linux driver support to the community resp. other companies like RedHat and Novell, therefore dropping support for this generation of hardware on linux nearly completely, which is not acceptable to me.

I highly appreciate the efforts of the developers that work on delivering support for this hardware in the opensource drivers which has shaped up nicely till now, but that doesn't take the responsibility from the company building the actual hardware. There are examples out there how to do it better: intel, HP with their opensource printing driver, NVidia (allthough it is closed source, they maintain support for all generations of hardware, the legacy drivers also get forward-ported to newer revisions of X and linux if I am informed correctly) etc.

So, consequence for me: I will try to avoid ATI hardware in the future whenever possible. I either want well-working closed source support (which is what people on Windows or OSX also want) or, even better, opensource support _actively maintained_ by the hardware manufacturer. Intel does the latter, and the performance of their graphics hardware is totally capable of satisfying my needs. ATI/AMD have proven to deliver none of both. Their strategy is better than the situation with a company that doesn't produce any programming info or a reliable binary driver, but I usually try to avoid those, too.

[bridgman]

We haven't talked much about specific funding and staffing levels (and don't plan to do that in the future either), but "paying one developer" to work on open source drivers is pretty far from the truth. We funded a lot of the radeonhd work, and we currently have three in-house developers working on open source drivers (Alex, Richard, Cooper) in addition to the outside development community.

We started ramping up open source efforts a year or so ago, which helped to get solid open source driver support in place for these generations of hardware.

[RealNC]

It seems like AMD is in for tough times. The other day I saw news that AMD makes dramatic cuts to the prices of Radeon cards.

I hope we won't be left with NVidia having a 95% market share in the end... AMD needs to come up with something. Right now, AMD vs NVidia is almost analogous to Linux vs Microsoft. Your lower market share kills you.

[Saist]

It seems like AMD is in for tough times. The other day I saw news that AMD makes dramatic cuts to the prices of Radeon cards.

I hope we won't be left with NVidia having a 95% market share in the end... AMD needs to come up with something. Right now, AMD vs NVidia is almost analogous to Linux vs Microsoft. Your lower market share kills you.

um. A: Nvidia's market share and stocks have been PLUMMETING : http://enthusiast.hardocp.com/news.html?news=Mzc5NTUsLCxoZW50aHVzaWFzdCwsLDE=

B: As far as I'm aware, the price cuts on the HD series are not hurting profitability (much I think), and if anything were cut to kick Nvidia while Nvidia was down.

The big problem is that the big green, or Nvidia, is in trouble. They are still largely riding on the G80 architecture introduced with the 8800 back in 9/11/2006 (http://www.techarp.com/showarticle.aspx?artno=88&pgno=3), and they had the disaster with the GPU's that caught on fire, and now they have the lawsuits with Intel over making an x86 cpu (http://www.bit-tech.net/news/hardware/2009/03/04/nvidia-reveals-plans-for-x86-cpu/1). With AMD having brought ATi in house, AMD has a competent chipset platform, and can compete with Intel on a platform basis.

Nvidia used to be dependent on the AMD market for it's chipsets, it's only been in recent years that Nvidia has made Intel chipsets, and Intel says that license agreement doesn't cover new processors (http://www.bit-tech.net/news/hardware/2009/02/18/intel-files-lawsuit-against-nvidia/1).

The chipset market gets even worse for Nvidia with Intel following AMD's lead and putting the Memory controller on chip (http://www.betanews.com/article/Intel-Officially-Confirms-Integrated-Memory-Controller-for-45nm-Nehalem/1175123788), removing one of the primary selling points of Nvidia hardware. Remember back on the AthlonXp series when and Nforce chipset easily blew Via and Sis out of the water?

Now Via and Sis are largely gone from the chipset market, and Nvidia's in position to follow.

***

So no, the chance of Nvidia taking 95% of a market they are losing, rapidly, is pretty low.

Rather, what needs to be remembered is that Intel chips, by far, make up the largest percentage of GPU's in use (http://www.xbitlabs.com/news/video/display/20080808090756_Market_of_Graphics_Adapters_Flat_as _Intel_Grabs_Highest_Market_Share_in_Years.html). A complete collapse of Nvidia would pit Intel versus AMD not just in processors, but in GPU's as well. It wouldn't be surprising given Intel's market position to see them leverage AMD into a niche corner.

[RealNC]

What's the current market of AMD/ATI vs NVidia? Isn't it something like 30% vs 70%? Not sure where I read that though.

[Saist]

What's the current market of AMD/ATI vs NVidia? Isn't it something like 30% vs 70%? Not sure where I read that though.

Um... nooo... I linked to Xbit's August report in the previous post

It's Intel that has the largest market share, with just under 50%.

AMD and Nvidia generally swap back around forth around 25-35% each.

Currently, AMD is on a role with the HD 4x00 series that trumps the Nvidia offerings at pretty much each price point.

[roothorick]

Good post, but...

Rather, what needs to be remembered is that Intel chips, by far, make up the largest percentage of GPU's in use (http://www.xbitlabs.com/news/video/display/20080808090756_Market_of_Graphics_Adapters_Flat_as _Intel_Grabs_Highest_Market_Share_in_Years.html). A complete collapse of Intel would pit Intel versus AMD not just in processors, but in GPU's as well. It wouldn't be surprising given Intel's market position to see them leverage AMD into a niche corner.

Patently false, and you'd have to be severely misinformed to think that. Intel's GMA "GPUs" have all the graphics capabilities of a Voodoo 5 packaged in a convienient integrated-motherboard solution -- they won't be running any serious games with any amount of capacity. Besides, you can't even buy them in discrete card form! They're not even in the same market. And even then, their performance even lags behind the GeForce 61xx and Radeon Xpress series, which are nV and ATI's integrated motherboard offerings respectively. Larrabee might change that, but by the looks of things, it's still at least a year away.

Back on topic, AMD is likely doing what they can to keep their dying Catalyst codebase going one more year. Remind me again, why didn't they just toss it out on its head and start over? Is there some spec release out and they're waiting for a FOSS driver to take the reins?

[bridgman]

Remind me again, why didn't they just toss it out on its head and start over? Is there some spec release out and they're waiting for a FOSS driver to take the reins?

The biggest focus of the Catalyst Linux drivers has always been professional workstation graphics (FireGL, FirePRO etc..) although over the last year we have been putting more effort into some consumer features as well.

From a technical point of view Catalyst Linux is our vehicle for letting Linux users see the performance and features from our proprietary shared-across-OSes code base. Practically speaking, for Linux users that means spiffy 3D performance and features, including things like Crossfire and Multiview. These are super-important for the professional workstation market but seem to be less interesting for consumer Linux users with the exception of running Wine.

In most other respects, the open source drivers have the advantage of staying current with the evolving framework (and it is evolving REALLY quickly right now) since new framework features (like EXA improvements) are usually worked out by making simultaneous changes to the open source drivers and the X/DRI framework. You can't do that kind of thing with closed-source drivers.

Anyways, bottom line is that the Catalyst Linux driver will continue to be extremely important for some market segments, but will probably never be a perfect fit for all of them -- but the open source drivers will be a good solution for the rest.

[Saist]

Patently false, and you'd have to be severely misinformed to think that. Intel's GMA "GPUs" have all the graphics capabilities of a Voodoo 5 packaged in a convienient integrated-motherboard solution -- they won't be running any serious games with any amount of capacity. Besides, you can't even buy them in discrete card form! They're not even in the same market. And even then, their performance even lags behind the GeForce 61xx and Radeon Xpress series, which are nV and ATI's integrated motherboard offerings respectively. Larrabee might change that, but by the looks of things, it's still at least a year away.

Roothick : you missed the point. Performance isn't the question here. Intel has the largest percentage of GPU's in use. Intel has almost as many graphics cards in use as AMD and Nvidia... combined.

Okay, I'm not going to argue that the performance sucks. If you've taken any time to read http://zerias.blogspot.com you'd have read me bashing on Intel's performance, and if you've read any of the posts I've made about Intel hardware on Mepislovers, I'm constantly going after the lack of hardware shaders.

High End cards, as much as we, as gamers, love them, only make up a small percentage of overall profits.

ATi, 3DFX, SiS, Trident, the SiS/Trident merger of Xabre, then XGI from that, Nvidia, Matrox, insert your favorite graphics card vendor here, all of them depend on low-end sales to drive their business.

I think (and if Mr. Bridgman reads this and access to the numbers to correct me if I'm wrong) that AMD's figures for sale of high-end cards, such as the Radeon 9800 XT, then the x1800 XT, then the HD 2900, then the insert high end card here, only makes up 5% or less of the actual series product sales.

Valve backs this up with their hardware reports : http://store.steampowered.com/hwsurvey/

I think this is where RealNC got his figure from: Steam only polls people who submit to the hardware report, doesn't report correctly from within Cedega or WINE, and only covers people who have Steam installed. In other words, NOT a reliable source of actual market share.

According to steam, the most popular Shader Model 2.0 card is the Radeon 9600... not exactly a high end card.

The most popular Shader Model 3 card is the 7600, which again, is NOT a high end card.

The most popular DX10 card is the 8800, which has been on the market for... well... more than 2 years, and it holds around 22%.

At 13% is the 8600... again. NOT a High end card.

At 9% though is the growing ATi market share, the 4800, which is referred to as series.

***

So, while Steam is a limited survey, it does show that mid-range and lower-end cards tend to be more popular.

[RealNC]

It's Intel that has the largest market share, with just under 50%.

I wrote "AMD vs NVidia" though, not "Intel vs AMD vs NVidia".

AMD and Nvidia generally swap back around forth around 25-35% each.

Looking at those stats, it's AMD vs NVidia: 36% vs 64%. So I wasn't too far off. So I stand by what I said. And you've just confirmed it.

Currently, AMD is on a role with the HD 4x00 series that trumps the Nvidia offerings at pretty much each price point.

Problem is, NVidia outperforms them currently with the more expensive cards and AMD doesn't offer anything equivalent.

[bridgman]

Yeah... in terms of units the low-end parts have by far the highest volumes. Most people still do pretty basic stuff with their PCs, and the most basic IGP parts satisfy a lot of users. You sure see the difference when you start gaming, however...

In terms of dollars I suspect the midrange parts (say HD4670 today) are the most significant, but I'm not sure about that.

[Saist]

Problem is, NVidia outperforms them currently with the more expensive cards and AMD doesn't offer anything equivalent.

um... ... no. Nvidia doesn't have anything to answer AMD's high-end offerings.

I'll quote HardOCP on their recent Phenom II DD3 review : http://www.hardocp.com/article.html?art=MTYyMSw1LCxoZW50aHVzaWFzdA==


I am sure you are wondering, “WTF happened to all our resolutions to compare?” Well interestingly enough, none of our current test setups had enough CPU to push the setups to being GPU limited even at 2560x1600. Those 4870 X2 in CrossFireX kick some major ass! big grin I could show you a big graph, but it would just be flat. Instead, what you see above represents every resolution. We had no set of scores that differed by more than 3 frames. Anyway.

Also, if you check Newegg here and look at the $400+ bracket (http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=2000380048+4020&Configurator=&Subcategory=48&description=&Ntk=&SpeTabStoreType=&srchInDesc=), you'll find the HD 4870 punching all of the Nvidia cards in the nose with a lower price point.

***


Looking at those stats, it's AMD vs NVidia: 36% vs 64%. So I wasn't too far off. So I stand by what I said. And you've just confirmed it.

Um. No. The Xbitlabs report from August 2008 shows Nvidia GPU's at 31% of shipment, and AMD GPU's around 18% shipment. So now. That doesn't comfirm what you said.

The only significant design win Nvidia has to their name are the parts in Apple's. Okay, I'll grant you that Apple's sell pretty decently, but that also ignores updated designs for AMD products over the past 6 months.

[bridgman]

Um. No. The Xbitlabs report from August 2008 shows Nvidia GPU's at 31% of shipment, and AMD GPU's around 18% shipment. So now. That doesn't comfirm what you said.

Actually in Aug 08 that might have been right; if you take out Intel then you need to double the "total market share numbers" to get just AMD vs NVidia and that gives you 62% vs 36%, which is pretty close.

That said, the Aug 08 numbers were shortly after we launched 7xx so probably don't reflect what is happening today.

[Saist]

Actually in Aug 08 that might have been right; if you take out Intel then you need to double the "total market share numbers" to get just AMD vs NVidia and that gives you 62% vs 36%, which is pretty close.

That said, the Aug 08 numbers were shortly after we launched 7xx so probably don't reflect what is happening today.

okay. I honestly didn't ready the numbers that way.

[RealNC]

um... ... no. Nvidia doesn't have anything to answer AMD's high-end offerings.

I'll quote HardOCP on their recent Phenom II DD3 review : http://www.hardocp.com/article.html?art=MTYyMSw1LCxoZW50aHVzaWFzdA==

A Phenom II is not a GPU :P


Also, if you check Newegg here and look at the $400+ bracket (http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=2000380048+4020&Configurator=&Subcategory=48&description=&Ntk=&SpeTabStoreType=&srchInDesc=), you'll find the HD 4870 punching all of the Nvidia cards in the nose with a lower price point.

Go through the pages here:

http://www.driverheaven.net/reviews.php?reviewid=711&pageid=8

NVidia is pretty much stomping on the 4870. I don't see how the 4870 outperforms NVidia. Not by any stretch of the imagination.

[bridgman]

okay. I honestly didn't ready the numbers that way.

That's the beauty of numbers; there are so many ways to read them :D

[roothorick]

The biggest focus of the Catalyst Linux drivers has always been professional workstation graphics (FireGL, FirePRO etc..) although over the last year we have been putting more effort into some consumer features as well.

From a technical point of view Catalyst Linux is our vehicle for letting Linux users see the performance and features from our proprietary shared-across-OSes code base. Practically speaking, for Linux users that means spiffy 3D performance and features, including things like Crossfire and Multiview. These are super-important for the professional workstation market but seem to be less interesting for consumer Linux users with the exception of running Wine.

In most other respects, the open source drivers have the advantage of staying current with the evolving framework (and it is evolving REALLY quickly right now) since new framework features (like EXA improvements) are usually worked out by making simultaneous changes to the open source drivers and the X/DRI framework. You can't do that kind of thing with closed-source drivers.

Anyways, bottom line is that the Catalyst Linux driver will continue to be extremely important for some market segments, but will probably never be a perfect fit for all of them -- but the open source drivers will be a good solution for the rest.

For 4xxx owners like myself, that "choice" is a choice between no 3D acceleration whatsoever, and many glaring non-OpenGL problems. Using radeonhd, I can't even play Quake 3. Using fglrx, I have black lines on my windows that shouldn't be there, panning in GQView is embarassingly slow, there's big tearing lines in Xv video playback, and I have to calculate custom modelines if I want a refresh rate higher than 75Hz. If I had an nV card, or even an Intel GMA or something ancient, I wouldn't have to deal with this. Based on my experience with this card, I probably won't be buying ATI in the future.

[bridgman]

A Phenom II is not a GPU :P

I think the point was that the high end GPUs are so fast that you end up CPU-limited on benchmarks, not GPU-limited. You can see that when the framerate stops increasing as you lower the resolution.

NVidia is pretty much stomping on the 4870. I don't see how the 4870 outperforms NVidia. Not by any stretch of the imagination.

You can spend your whole life arguing this; every time prices change the relative price-performance of the competitors changes completely. The very high end performance is primarily for bragging rights; all of the other purchase decisions are made on bang-for-the-buck.

[bridgman]

If I had an nV card, or even an Intel GMA or something ancient, I wouldn't have to deal with this.

Right now if you had an older GPU from anyone including ATI you would have an easier time -- the newest GPUs from all vendors have the most problems with Linux. That is changing pretty quickly though.

Using radeonhd, I can't even play Quake 3.

Agreed, but that's why we're focusing *more* resources on fglrx support for newer GPUs in parallel with finishing the open source support for 6xx/7xx. This is about shifting resources, not cutting them.

Using fglrx, I have black lines on my windows that shouldn't be there, panning in GQView is embarassingly slow, there's big tearing lines in Xv video playback, and I have to calculate custom modelines if I want a refresh rate higher than 75Hz.

I don't remember seeing these issues reported before; tearing in Xv is a known issue (for most vendors, not just us) but you should be able to get good tear-free playback through OpenGL. If you haven't already posted details about the black lines on your windows can you start a thread or point to a bug ticket somewhere ?

Thanks,
JB

[Hephasteus]

I think the point was that the high end GPUs are so fast that you end up CPU-limited on benchmarks, not GPU-limited. You can see that when the framerate stops increasing as you lower the resolution.



You can spend your whole life arguing this; every time prices change the relative price-performance of the competitors changes completely. The very high end performance is primarily for bragging rights; all of the other purchase decisions are made on bang-for-the-buck.


You're not seeing cpu limiting. Things haven't been cpu limited for years. You're seeing parallel limiting. The framerate stops increasing as you lower the resolution because the cards are designed to handle x number of calculations as well as they can handle y number of calculations. The biggest cards are able to do max resolution or less than max resolution with alot of post processing (Anitropic filtering, Anti aliasing, etc.) You'll only hit bump ups in framerate when you hit sweet spots on resolution, post processing etc being able to be divided down into streams or rop's that don't stall very often.

[Melcar]

...



Go through the pages here:

http://www.driverheaven.net/reviews.php?reviewid=711&pageid=8

NVidia is pretty much stomping on the 4870. I don't see how the 4870 outperforms NVidia. Not by any stretch of the imagination.


Think he meant the HD4870X2. The HD4870 is at a much lower price bracket and is not meant to compete with neither of nvidia's high end.

[roothorick]

Right now if you had an older GPU from anyone including ATI you would have an easier time -- the newest GPUs from all vendors have the most problems with Linux. That is changing pretty quickly though.

There's major documented problems with older ATI cards going back to the X1300 and beyond. After all, how was this thread started? Right. I'm fairly confident that if I had bought a GTX280 instead, I'd have good video playback and 2D performance.

I don't remember seeing these issues reported before; tearing in Xv is a known issue (for most vendors, not just us) but you should be able to get good tear-free playback through OpenGL. If you haven't already posted details about the black lines on your windows can you start a thread or point to a bug ticket somewhere ?

Wait, you have a bug tracker?

In any case, the Xv tearing is an fglrx-specific issue unlike what you think. It doesn't happen with radeonhd, it doesn't happen on Intel GMA with the DRI drivers, and it doesn't happen on nV cards with either their binary blobs or the nv driver. Yes, OpenGL is what I'm using now, but it requires that the player have a properly optimized OGL backend -- you can forget about MythTV.

Oh, and I forgot to mention that fullscreen FLV playback via Adobe's Flash plugin for Firefox is unusably slow. The video is a slideshow and the controls are so unresponsive that every time I try it I have to resist the urge to tab out to a tty and kill X outright. This doesn't happen on nV cards with their binary blobs.

-E- Another (minor) issue I just remembered: If I tab out to a tty and kill an OGL app, the screen is corrupted. It does correct itself when tabbing back to X, at least.

[bridgman]

There's major documented problems with older ATI cards going back to the X1300 and beyond. After all, how was this thread started? Right. I'm fairly confident that if I had bought a GTX280 instead, I'd have good video playback and 2D performance.

Understood; I was talking about the specific problems you mentioned.

Wait, you have a bug tracker?

It should be in the Catalyst Linux release notes, I'll check. Anyways, it's at http://ati.cchtml.com .

In any case, the Xv tearing is an fglrx-specific issue unlike what you think. It doesn't happen with radeonhd, it doesn't happen on Intel GMA with the DRI drivers, and it doesn't happen on nV cards with either their binary blobs or the nv driver.

Yes and no. Tear-free support only went into the radeonhd driver very recently and is still a bit problematic there (check IRC or mailing list). Tearing with Xv isn't an issue with any of the older GPUs (including both Intel and AMD/ATI) which use hardware overlay for video playback, but on the more recent GPUs (or older GPUs using textured video instead of overlay) I think even the latest Intel Xv code has tearing issues.

Oh, and I forgot to mention that fullscreen FLV playback via Adobe's Flash plugin for Firefox is unusably slow. The video is a slideshow and the controls are so unresponsive that every time I try it I have to resist the urge to tab out to a tty and kill X outright. This doesn't happen on nV cards with their binary blobs.

My guess is that it doesn't happen with the ATI open source drivers either, will check.

Anyways, I guess the main point here is that for the GPU generations which were dropped the open source drivers are in pretty good shape already and development work on those drivers is continuing to grow, including both AMD and community resources.

[DoDoENT]

-E- Another (minor) issue I just remembered: If I tab out to a tty and kill an OGL app, the screen is corrupted. It does correct itself when tabbing back to X, at least.

This happens to me with fglrx too. And sometimes tabbing back to X doesn't fix the corruption.

That's why I'm looking forward to use open source drivers for my Mobility Radeon X1600. But now I expect FOSS drivers to have PowerPlay and OpenGL 2.1+ support for R500 cards by the June.

ATi, since now you are dropping support for R500 and older cards, feel free to open parts of fglrx that handle R500 and older cards. Now you don't have to hide that parts of fglrx code, as you don't support those chips anymore. And I'm sure that those bits of code could help FOSS driver developers a lot in making PowerPlay and OpenGL 2.1+ support for R500-based (and older) cards.

I think that'd be a good move for you, ATi.

[Saist]

ATi, since now you are dropping support for R500 and older cards, feel free to open parts of fglrx that handle R500 and older cards. Now you don't have to hide that parts of fglrx code, as you don't support those chips anymore. And I'm sure that those bits of code could help FOSS driver developers a lot in making PowerPlay and OpenGL 2.1+ support for R500-based (and older) cards.

I think that'd be a good move for you, ATi.

They can't.

This was one of the reason's behind funding Novell's RadeonHD.

FGLRX contains several licensed technologies and patents that AMD doesn't own, with sources ranging from (old)SGI to IBM to (old Artx) to Intel. AMD can ONLY open up the Source Code to technology that AMD owns outright, or has created.

Sure, it's a nice idea to just open FGLRX up, but it's just not legally possibly.

[Melcar]

This happens to me with fglrx too. And sometimes tabbing back to X doesn't fix the corruption.

That's why I'm looking forward to use open source drivers for my Mobility Radeon X1600. But now I expect FOSS drivers to have PowerPlay and OpenGL 2.1+ support for R500 cards by the June.

ATi, since now you are dropping support for R500 and older cards, feel free to open parts of fglrx that handle R500 and older cards. Now you don't have to hide that parts of fglrx code, as you don't support those chips anymore. And I'm sure that those bits of code could help FOSS driver developers a lot in making PowerPlay and OpenGL 2.1+ support for R500-based (and older) cards.

I think that'd be a good move for you, ATi.


They already did. The licensed bits still have to remain closed, but that's not really ATI's fault. It's not at ATI now that you should be putting pressure at, but rather the radeonhd/radeon developers. The ball is in their court now.
I called previously how people would complain if fglrx ever dropped support for cards again, and it seems I called it right. It's really not a surprise. The drivers meant for the Linux desktop user are radeon/radeonhd (for ATI users), so if any pressure should be applied it should be towards these pair of drivers. Fglrx is a worstation driver that just so happens to support desktop GPUs. That's ATI's main market in Linux and it will always be.

[DirtyHairy]

We haven't talked much about specific funding and staffing levels (and don't plan to do that in the future either), but "paying one developer" to work on open source drivers is pretty far from the truth. We funded a lot of the radeonhd work, and we currently have three in-house developers working on open source drivers (Alex, Richard, Cooper) in addition to the outside development community.OK, glad you corrected me, bridgman, I only knew of Alex Deucher. Still, I don't feel good about this move. The opensource support is far from being perfect in my experience (not only on R500, but also R300); powerplay is not there (although dynamic clocks does an acceptable job for me), and on both chip generations, a lot of 3D stuff doesn't work as intended (including even some xscreensaver modules, try the ant ones like antinspect) yet. While I can live with this and certainly don't want to lament about non-functional xscreensaver modules, I don't like getting only half-baked support over more than two years and then official support being dropped completely, being left with the hope that the opensource drivers will be able to do with help of the "outside development community" in due time what the proprietary driver never delivered.

[DoDoENT]

FGLRX contains several licensed technologies and patents that AMD doesn't own, with sources ranging from (old)SGI to IBM to (old Artx) to Intel. AMD can ONLY open up the Source Code to technology that AMD owns outright, or has created.

So let me get this straight: Some parts of the chip I have in my laptop are not made by ATi, but by someone else??? Then how could have ATi sold me a chip that was only partly their? Is it so complicated to call those respective owners and ask them nicely for the permission of opening up the code (like e.g. Intel or IBM)? AFAIK, Intel and IBM support open source...

[Melcar]

So let me get this straight: Some parts of the chip I have in my laptop are not made by ATi, but by someone else??? Then how could have ATi sold me a chip that was only partly their? Is it so complicated to call those respective owners and ask them nicely for the permission of opening up the code (like e.g. Intel or IBM)? AFAIK, Intel and IBM support open source...


Licensing. It's not an alien concept. Everyone does it to an extent. AMD has little power over things that they are licensing. They obviously can ask whoever is licensing to them, but it's usually not that straight forward, and the answer is usually "no".

[Saist]

Originally Posted by DoDoENT View Post
So let me get this straight: Some parts of the chip I have in my laptop are not made by ATi, but by someone else??? Then how could have ATi sold me a chip that was only partly their? Is it so complicated to call those respective owners and ask them nicely for the permission of opening up the code (like e.g. Intel or IBM)? AFAIK, Intel and IBM support open source...Licensing. It's not an alien concept. Everyone does it to an extent. AMD has little power over things that they are licensing. They obviously can ask whoever is licensing to them, but it's usually not that straight forward, and the answer is usually "no".

Melcar pretty much beat me to the answer. I touched on this back in my original post : http://www.bit-tech.net/news/hardware/2009/03/04/nvidia-reveals-plans-for-x86-cpu/1 :: http://www.bit-tech.net/news/hardware/2009/02/18/intel-files-lawsuit-against-nvidia/1

Intel basically owns x86 technology. Only companies with a license from Intel can make x86 chips. If you just want a taste of the fights that have occured over this, do a web search for the terms Intel x86 license

Many of the large tech companies hold various patents for hardware design that they allow other companies to use, sometimes in exchange for cash, and sometimes in exchange for equal technologies, like the AMD / Intel cross-license agreement on x86.

***

The situation can quickly deteriorate when companies that some technology for go out of business, or are merged. Tracking down who owned what technology, at what time, can be a legal heart-ache. Even when a vendor is fairly certain they HAVE licensed the technology properly, somebody else may show up with a broad patent claim, or having bought up interest in a patent portfolio : Look up patent troll


***

Also, there's a difference here between hardware and software. Yes, AMD / ATi designs the chip hardware, but that doesn't mean they use internal software for it. A good case in point is the old TV-output, which has been covered several times on Phoronix during the development of TV-output into the X.org ATi driver. Check out this post by Mr. Bridgman from 2007 for example : http://www.phoronix.com/forums/showpost.php?p=62565&postcount=2

[bridgman]

The fglrx driver also contains code shared across multiple OSes, and all the OSes except Linux have strong DRM (Digital Rights Management, not Direct Rendering Manager) requirements. It's certainly possible to pick the driver apart and separate out the bits which can be safely exposed, but the result would look a lot more like ground beef than like a steak ;)

Note that Intel doesn't release docs on all the hardware in their chips either. That is not meant as criticism; we are all dealing with the same constraints here. I didn't think IBM ever supported open source driver development for their graphics chips; Our FireGL guys might remember -- they used IBM GPUs before switching to ATI).

In all seriousness, I don't think documentation or sample code is a factor for power management or for higher levels of OpenGL support.

Power management is primarily waiting for things to settle down in the command submission portions of the driver stack so that power management code can dynamically adjust to drawing workload and display modes etc...

OpenGL 2.1 theoretically comes for free once Gallium3D drivers are running on 3xx-5xx. Gallium3D in turn is currently built over DRI2 and GEM/TTM, which are also in progress.

For anyone wondering why we expect the open source drivers to make you happy even if fglrx did not, the reason is simple. The open source drivers are aimed directly at the mix of functionality and performance that most of you are asking for and *only* contain code for that functionality. Fglrx is still aimed primarily at professional workstation users on a small set of enterprise Linux distibutions, and includes well over 10x as much hardware-specific code as the open source drivers.

You really need all that code to get every last bit of 3D performance out of the GPU, but a *much* smaller driver can provide all the functionality most of you expect along with perhaps 70% of the performance -- and can be tweaked to work well on a wide variety of distros and systems much more readily than our workstation driver.

[DoDoENT]

In all seriousness, I don't think documentation or sample code is a factor for power management or for higher levels of OpenGL support.
Great then! So we can expect open source 3D rendering as fast as fglrx?

Power management is primarily waiting for things to settle down in the command submission portions of the driver stack so that power management code can dynamically adjust to drawing workload and display modes etc...
So, I guess this is WIP?

OpenGL 2.1 theoretically comes for free once Gallium3D drivers are running on 3xx-5xx. Gallium3D in turn is currently built over DRI2 and GEM/TTM, which are also in progress.

What about OpenGL 3.0?


You really need all that code to get every last bit of 3D performance out of the GPU, but a *much* smaller driver can provide all the functionality most of you expect along with perhaps 70% of the performance -- and can be tweaked to work well on a wide variety of distros and systems much more readily than our workstation driver.

WTF??? Only 70%? So I guess I won't be playing Nexuiz on high resoultion with all effects turned on on an open source driver? :(

I'd understand 90+%, but 70%?? That's just not enough, or is it?

[susikala]

I don't really get why people complain about this move. You can't maintain those things forever, especially when you have a good alternative. It is just a proof that AMD's open source strategy pays off. I mean, effectively, dropping Catalyst support for <R600 means that many Windows users will have to either switch to Linux or buy a new card in order to use the latest technologies. Linux users, on the other hand, will continue to profit from the development of the FOSS driver(s), backed by AMD or community-driven. Also, as someone here pointed out, more people moving to the open source drivers will help squashing bugs etc. So the ones who'll really profit from this move are the Linux users -- I'd be more pissed off if I were using Windows and knew Catalyst had some bug that's never going to get fixed.

Honestly, I'd have them completely drop Catalyst support for Linux and have a few more people working on FOSS drivers.

[roothorick]

Hold on a minute here bridgman, I was kind of under the impression that fglrx shares most of its code with the Windows drivers and you had some "Unified Driver Architecture" type middleware in the mix. Is that simply not true? Is the fglrx driver a completely different codebase specifically written for professional applications, in comparison to the gaming-oriented Windows drivers? And what's this 70% number? Are you telling me that Linux open source solutions will only have just over two thirds the performance of the current fglrx driver, which is already a fair distance behind the Windows blobs? What are we looking at, half the framerate? Excuse me if I find that hard to swallow.

[Melcar]

The performance of the driver will depend entirely on how much optimization the open source devs. are willing and/or able to put into the drivers.

[bridgman]

Hold on a minute here bridgman, I was kind of under the impression that fglrx shares most of its code with the Windows drivers and you had some "Unified Driver Architecture" type middleware in the mix. Is that simply not true? Is the fglrx driver a completely different codebase specifically written for professional applications, in comparison to the gaming-oriented Windows drivers?

The fglrx driver shares big chunks of code with other OSes, and includes code paths for both workstation and consumer products. I call it a "workstation driver" because both the development and test focus of the Linux-specific bits are biased towards professional workstation use, particularly in terms of distro support (RHEL and SLED are not the most common consumer distros).

And what's this 70% number? Are you telling me that Linux open source solutions will only have just over two thirds the performance of the current fglrx driver, which is already a fair distance behind the Windows blobs? What are we looking at, half the framerate? Excuse me if I find that hard to swallow.

AFAIK the Windows and Linux binary driver performance should be pretty much identical today; I think Linux stopped being "a fair distance behind" around the end of 2007.

I asked our architects what performance levels they felt could be obtained with a "clean, simple, well written driver" but without any application-specific performance optimization and their estimate was an average of between 60 and 70% of proprietary driver performance.

We have released enough programming info to get to 100% but the driver code size and effort grows exponentially as you go far that last 30%. Getting the last bit of performance optimization out of a driver/GPU combination is extremely time-consuming and just plain hard work -- and none of the devs I have spoken with feel it will be necessary.

This was based on the assumption of identical performance between Windows and Linux; if you are seeing something different (other than running through Wine) please let me know.

[DoDoENT]

I asked our architects what performance levels they felt could be obtained with a "clean, simple, well written driver" but without any application-specific performance optimization and their estimate was an average of between 60 and 70% of proprietary driver performance.

We have released enough programming info to get to 100% but the driver code size and effort grows exponentially as you go far that last 30%. Getting the last bit of performance optimization out of a driver/GPU combination is extremely time-consuming and just plain hard work -- and none of the devs I have spoken with feel it will be necessary.


So, this means only 2100 out of 3000 FPS with glxgears on my Mobility Radeon X1600. (in 3rd powerstate) I'm not impressed. Sorry. FGLRX gave me 2500+ FPS in 2nd powerstate and over 3000 FPS in 3rd powerstate.

But, one day I'll learn how to make drivers and I'll optimize my copy of radeon to get 100% of my card. That's the beauty of open source. :cool:

[Melcar]

So, this means only 2100 out of 3000 FPS with glxgears on my Mobility Radeon X1600. (in 3rd powerstate) I'm not impressed. Sorry. FGLRX gave me 2500+ FPS in 2nd powerstate and over 3000 FPS in 3rd powerstate.

But, one day I'll learn how to make drivers and I'll optimize my copy of radeon to get 100% of my card. That's the beauty of open source. :cool:


Glxgears in not a benchmark. And you can't really blame ATI for the state of the open source drivers... not anymore at least. The info. is available. Developers have the task of making the drivers and optimizing 3D performance.

[bridgman]

Yeah, I wasn't really thinking about glxgears in my posts; I didn't want to have to explain to the architects why everyone uses a benchmark which hardly uses any of the GPU functionality ;)

The dev community is already working on the top priorities for improving 3D performance in the open drivers :

#1 - memory manager (needed for GL 1.5 and higher) - GEM/TTM

#2 - redo the command submission and buffer management code (current driver stack doesn't pipeline CPU and GPU operation as much as it could) - bufmgr, radeon-rewrite

#3 - shift to a driver model designed around shader-based GPUs rather than fixed-function GPUs (ie Gallium3D)

Once those are done (and all are making great progress) I think you'll see driver development move back to the incremental improvements you are used to seeing. Right now there is perhaps 18 months of work accumulated in branches and alternate code paths, and all of that should start to show up in releases over the next few months.

[DoDoENT]

Yeah, I wasn't really thinking about glxgears in my posts; I didn't want to have to explain to the architects why everyone uses a benchmark which hardly uses any of the GPU functionality ;)

So, glxgears actually doesn't fully utilize GPU?


The dev community is already working on the top priorities for improving 3D performance in the open drivers :

#1 - memory manager (needed for GL 1.5 and higher) - GEM/TTM

#2 - redo the command submission and buffer management code (current driver stack doesn't pipeline CPU and GPU operation as much as it could) - bufmgr, radeon-rewrite

#3 - shift to a driver model designed around shader-based GPUs rather than fixed-function GPUs (ie Gallium3D)

Once those are done (and all are making great progress) I think you'll see driver development move back to the incremental improvements you are used to seeing. Right now there is perhaps 18 months of work accumulated in branches and alternate code paths, and all of that should start to show up in releases over the next few months.

OK, now I understand: latest open source drivers are actually in good shape regarding performance, but those latest bits of code still aren't included in most of distributions (what users actually see). So, could we see good open source 3D performance and power management by the end of this year (I mean in Ubuntu 9.10/Fedora 12)?

[bridgman]

So, glxgears actually doesn't fully utilize GPU?

Right. The glxgears program uses only fixed-function graphics (no shaders), and draws shaded triangles with no textures. Between shaders and textures that's maybe between 80% of a modern GPU sitting idle. Vertex shaders get used to emulate the fixed-function transform and lighting, but that's about it. The only block that works hard is the ROP/RBE, ie the part that handles depth compare (Z-buffer) and writes pixels into video memory.

Even worse, since older chips used relatively more of their silicon area for ROP/RBE than new chips, it's not unusual for an old GPU to outperform a new GPU on glxgears.

OK, now I understand: latest open source drivers are actually in good shape regarding performance, but those latest bits of code still aren't included in most of distributions (what users actually see). So, could we see good open source 3D performance and power management by the end of this year (I mean in Ubuntu 9.10/Fedora 12)?

I think so. The end-of-year distros are probably the first ones which will be able to ship with all this new stuff, but it should be available for download & build sooner.

Power management is the only part that hasn't really been implemented yet (everything else has been implemented but not integrated), but for 5xx and below I *think* it should come together fairly quickly once the invasive drm changes start to settle down.

[RealNC]

DX10 cards even have those "unified shaders" anyway, where the distinction between geometry, vertex and pixel processors becomes fuzzy. I think glxgears doesn't have any access to this feature, so it uses the small (fixed) percentage of shaders that are assigned to it by the driver. I think ATI cards give about 30% of their processors to vertex/geomerty. The other 70% is dedicated to pixel shading and is not accessible to glxgears at all. So in other words, glxgears only uses 30% of a modern GPU.

I think this 30%/70% ratio started with the X1900.

(I hope I got the above right :P)

[Mr_Ed]

Right now the current fglrx driver gives me much better performance (2d and 3d) on my 9600 XT/TV (rv350) card then the open source driver does.

But from reading this thread I understand that fglrx will not support my 9600 card anymore and I am forced to switch to the open source driver?

Is that correct?

[RealNC]

Yes. Though I preferred the open driver on my X1950XT anyway. Catalyst was faster, but bugged like hell. With the open driver, everything was working correctly, even though 3D was slower.

[BlackStar]

(I hope I got the above right :P)
You are not far off the mark. The unified architecture was introduced with R600 (the 2000 series). R500 is still old-style vertex & fragment shaders, which is why it can be accelerated using the same code as R300/R400.

Other than that, unified shaders means that the drivers and/or hardware can dynamically balance resources according to the utilization. In other words, if your program is 90% vertex shader and 10% fragment shader computation, a unified architecture will be able to allocate the hardware resources to match that. An old-style architecture, like R500, would leave the fragment processing hardware underutilized, in this 90/10 scenario.

But from reading this thread I understand that fglrx will not support my 9600 card anymore and I am forced to switch to the open source driver?

Is that correct?
Yes. You'll be able to use fglrx with the 2.6.28 kernel and XServer 1.5 (and maybe 1.6), but you'll have to change to radeon/radeonhd once new versions arrive.

[bridgman]

Right now the current fglrx driver gives me much better performance (2d and 3d) on my 9600 XT/TV (rv350) card then the open source driver does.

I agree that 3d performance is better on fglrx than on the open source drivers, but most users seem to be reporting better 2d performance with the open source drivers.

Are you using XAA or EXA acceleration with the open source drivers ? I think the default is still XAA, but EXA often gives better performance these days.

[Mr_Ed]

Conclusion: I will not ever going to be able to use the fancy stuff like desktop effects (compiz, AWN, screenlets, blurred window borders, etc.) since this is slow like hell when using the open source driver. Google earth doesn't even work with it.

While I have been waiting for a few years for the fglrx driver to solve all this for me, I feel pretty much screwed now by ATI.

Bummer!

[Mr_Ed]

Yeah, I tried the open source driver with EXA, it doesnt work well.

[Saist]

Conclusion: I will not ever going to be able to use the fancy stuff like desktop effects (compiz, AWN, screenlets, blurred window borders, etc.) since this is slow like hell when using the open source driver. Google earth doesn't even work with it.

interesting... I was able to use Beryl on a Radeon 8500 using the X.org ATi driver off of Mepis 6.5 with no problem, and I seriously doubt you can get much slower than that when it comes to 3d rendering on the open source X.org ATi driver.

[bridgman]

Yeah, something is wrong here, maybe drm not initializing properly (it won't start up unless you fully uninstall fglrx and reboot before starting with the open driver. That would explain both 2d and 3d running slow (but it's only a guess).

[Mr_Ed]

Here's the google earth bug thing:

https://bugzilla.redhat.com/show_bug.cgi?id=474063

[Mr_Ed]

Ok, I guess I have to buy me a new(er) graphic card.

Which card should I buy? Suggestions?

[Mr_Ed]

I remember my firts ATI card was a diamond stealth (VESA local bus), I have been using ATI cards ever since like the All-in-wonder and later the RAdeon-all-in wonder. But my laptop (hp nw9440) has an Nvidia FX1500M which I also like very much in linux. What is the best choice?

[susikala]

What is the best choice?

If you use Linux, AMD/ATI. It's pretty much the _only_ serious choice.

No wait, if you use computers, AMD/ATI. XD

[Saist]

I remember my firts ATI card was a diamond stealth (VESA local bus), I have been using ATI cards ever since like the All-in-wonder and later the RAdeon-all-in wonder. But my laptop (hp nw9440) has an Nvidia FX1500M which I also like very much in linux. What is the best choice?If you use Linux, AMD/ATI. It's pretty much the _only_ serious choice.

No wait, if you use computers, AMD/ATI. XD

seconding that.

Until Nvidia gets serious about an Open Source strategy, if you really want to support Open Source, AMD is your only graphics choice.

However, if you really don't care about the theology, or voting with your wallet, buy whatever you want.

***

Ed : right now I'd say look around for the HD 3870. Okay, Crossfire support isn't enabled yet, so I personally am a bit of a litter box on that, but they still pack a punch as a single card. They also have a wide driver support range, so if something's screwy on the bleeding edge 9.x set, you can still drop back some drivers.

[DirtyHairy]

So, in this whole discussion, always popping up seems to be the train of thought: "the info is out there, now it is up to the community to pick up the ball and make a good driver". Heck, someone even said "bugger them about it". While I can't possibly know if this is really ATIs strategy or whether they are going to actively put money and workpower into the opensource drivers, this is just plain wrong for me: in my opinion, it is the job of the _hardware vendor_ to provide software support for his products, be it closed- or open source. Providing the necessary specs for programming the thing is the _bare minimum_ of support imaginable for me, and not going further than that means offloading the support to other companies and individuals to save money.

Apart from that, not all is well in the open source ATI world for me; the drivers do work fairly well given the fact that they are based on reverse engineering and/or documentation that has been available for only some 14 months, nothing more. Regarding that EXA hype: I tried XAA versus EXA several times on my X1300 as well as on my 96something R300 based card in my powermac: while EXA _is_ signigicantly faster when I do compositing, the whole desktop feels snappier with XAA when I disable compositing. For example, scrolling down huge (web) pages is noticably smoother on XAA than on EXA.

[susikala]

So, in this whole discussion, always popping up seems to be the train of thought: "the info is out there, now it is up to the community to pick up the ball and make a good driver". Heck, someone even said "bugger them about it". While I can't possibly know if this is really ATIs strategy or whether they are going to actively put money and workpower into the opensource drivers, this is just plain wrong for me: in my opinion, it is the job of the _hardware vendor_ to provide software support for his products, be it closed- or open source. Providing the necessary specs for programming the thing is the _bare minimum_ of support imaginable for me, and not going further than that means offloading the support to other companies and individuals to save money.

The deal is more like: they sell you hardware. It works well, I hear, for the absolute bulk of users (Windows). If you move to alternative/niche OSes, you can reasonably expect a drop in support. By estimation, I'd say the level graphics is supported on Linux is currently much better than games in general are -- which is good, solid ground for more people using Linux. This is all thanks to AMD, though.

Fortunately, it's up to you to improve support if you want. But I don't see how AMD is in any way commited to provide you 100% of their Windows support on Linux -- and they _are_ dropping the drivers on Windows too.

Edit: I also wanted to add that whining about this move also shows many users expect AMD to have a "one OS mentality" (which is blatantly ridiculous; if anything, logic says they should only concentrate on Windows). I mean, users of <R600 on Windows will be able to keep using their cards with no problems -- no new xservers and other new technologies, etc.

[RealNC]

Conclusion: I will not ever going to be able to use the fancy stuff like desktop effects (compiz, AWN, screenlets, blurred window borders, etc.) since this is slow like hell when using the open source driver. Google earth doesn't even work with it.

Uhm, it's the other way around. The open source drivers are much faster with desktop (both 2D and 3D compositing). Catalyst in only faster with pure 3D OpenGL apps (mainly games), but very slow with 2D and desktop effects (Compiz, KDE4, or whatever).

[DirtyHairy]

The deal is more like: they sell you hardware. It works well, I hear, for the absolute bulk of users (Windows). If you move to alternative/niche OSes, you can reasonably expect a drop in support. By estimation, I'd say the level graphics is supported on Linux is currently much better than games in general are -- which is good, solid ground for more people using Linux. This is all thanks to AMD, though.

Fortunately, it's up to you to improve support if you want. But I don't see how AMD is in any way commited to provide you 100% of their Windows support on Linux -- and they _are_ dropping the drivers on Windows too.I don't care about games, and I don't care whether the feature set is similar to the windows driver; all I am complaining about is the fact that I bought something that hasn't worked as advertised most of the time, and now the company that produced it is dropping support.

Concerning me improving the open source driver: yeah, I can just dig into it in my free time, take the register guide and improve it. Don't get me wrong, I do hack into open source software if I urgently have to get rid of a bug, I do file bug reports, and I have produced my (small) share of patches to open source code over the years. However, just startíng to hack on a graphics drivers is a little bit above my head. If have choosen this hardware because ATI has advertised linux support (meaning _they_ support it), not because I was longing for the adventures of writing a graphics driver.

[bridgman]

So, in this whole discussion, always popping up seems to be the train of thought: "the info is out there, now it is up to the community to pick up the ball and make a good driver".

That's not what we are saying. See below.

Providing the necessary specs for programming the thing is the _bare minimum_ of support imaginable for me, and not going further than that means offloading the support to other companies and individuals to save money.

That is also what most people were asking for until recently. We provided documentation AND co-funded the initial driver development with Novell AND hired in-house developers and testers to work on the open source drivers. Other than giving everyone a pony, what are we missing ?

[Saist]

That's not what we are saying. See below.

That is also what most people were asking for until recently. We provided documentation AND co-funded the initial driver development with Novell AND hired in-house developers and testers to work on the open source drivers. Other than giving everyone a pony, what are we missing ?

What you are missing is human nature. There are people within the Open Source communities that have looked for any reason to be anti... something. They don't care what. AMD makes a good target because it's a decent size company that's seen as profitable. Some people just want to attack that.

The concept is on full display here with people who have never filed bug reports or gotten involved on any level of development stating they have unfixable bugs, and things are broken, etc, etc, etc.

Then there are those that declare that their perfectly functional card on a perfectly functional system is now broken and unusable. It doesn't matter that they are using the system, right now, to do everything they presumably want to do, since they are using it. All that matters is that support for their product won't be the same in the future.

Then there are those that proclaim their card is / was broken since a period of time ago. When confronted with people saying "it worked me on such and such a system" and when confronted with offers to try and, you know, solve the problem, those people simply don't want to listen.

Then there are those that are die-hard distro-backer's or bashers. I'll be the first in line to say I'm biased against Ubuntu, or any Gnome based distro. One of the main points of Linux though is it's flexibility. If one distribution does not work for you, try another one. Distributions are not created equally. Some have different x-servers, different kernels, different module pre-loads, and differt etc, etc, etc. Some people simply don't want to change or adjust, and the idea of going to another distribution that might have a work around or solution is just unimaginable.

***

Now, I personally tend to agree with the state of X.org Radeon right now being able to take over for Fglrx. Lets be honest, if I am a serious gamer, I would already be budgeting to upgrade my x1800 XT and x1900 systems. Well. Okay. I am a serious gamer and I am budgeting to replace those cards. I know that the Shader Model 2 graphics cards are going to have a hard time with the Unreal Engine 3 and ID Tech 5 graphics. If I want to play games on those platforms, I AM going to need something newer, driver support not-withstanding.

So, lets junk the performance 3D argument. It simply doesn't hold any weight.

Outside of performance 3D, the X.org ATi driver is actually better in 2D display. That's important when 1920*1200 LCD monitors are getting to be affordable.

Then there's the tear-free video support...

and basic support for 3D operations. I CAN use stuff like Compiz Fusion on X.org ATi on R100-R500 hardware. Granted, I really don't think I'd enjoy the experience on a Radeon 7500, but the option is there.

So, as an everyday user, X.org ATi is simply good enough for the average consumer, and lets say it, the type that buys from Dell.

I've also seen an expressed interest from the actual Developers, you know, the guys WRITING the code, in having bug reports filed and proper procedures followed through for those who are having problems with their display.


***

I'm also going to leave one parting shot here that might just open some people's eyes as to how committed AMD is to Linux and Open-Source.

R300-R500 is ALSO getting dropped from Windows Catalyst. Currently Windows users have no alternative drivers. At all.

This also means that upcoming Windows 7 users... won't be able to use the R300-R500 GPU's ... at all. There might be a basic driver provided by Microsoft, but they won't be getting updates. There will be no driver path.

Chew on that.

[DarkFoss]

Will owners of hd 2600 through hd 3850 AGP be cut off as well in both Linux and Windows ?

[Svartalf]

That is also what most people were asking for until recently. We provided documentation AND co-funded the initial driver development with Novell AND hired in-house developers and testers to work on the open source drivers. Other than giving everyone a pony, what are we missing ?

The in-house part wasn't always made clear John. Not everybody was on that same page with you and your employer (even though I was... ;) ).

I'm not WHOLLY sure that AMD got it's money's worth with the Novell portion of that deal, but all in all, I'm tickled with what you've done to the point of putting Radeons back into my machines and buying new ones (Which, unfortunately seem to have issues with fglrx and things like user switching...).

I suspect a somewhat larger investment into the overall Linux support picture would make people happier- but it's a tough sell to AMD upper management. I know, I didn't get very far myself with that one when I had the chance. :D

[Saist]

Will owners of hd 2600 through hd 3850 AGP be cut off as well in both Linux and Windows ?

um. no support for R600+ chips are being dropped. So no.

[Stormking]

I don't really get why people complain about this move. You can't maintain those things forever
It's not about maintaining things forever. It's about ATI promising the driver to get better but never delivering. It's about their final admittance that there will never be a working driver for my hardware.

If there had been a fully functional driver for the last years and now support gets dropped, I would completely understand that. But there hasn't. Something was always broken and now all hopes that *someday* things might just work has vanished into thin air.

[Saist]

It's not about maintaining things forever. It's about ATI promising the driver to get better but never delivering. It's about their final admittance that there will never be a working driver for my hardware.

If there had been a fully functional driver for the last years and now support gets dropped, I would completely understand that. But there hasn't. Something was always broken and now all hopes that *someday* things might just work has vanished into thin air.


deary. I've already called you out for trolling. Knock it off.

[cutterjohn]

All I can say is: WOW! WTF?!

I just bought an MSI GT725 notebook specifically because I wanted to try an ATI card again, but seeing the proprietary X11 AND WIndows(!) driver dumping parts so fast makes me feel even more queasy than I already do looking at the overall lack of adequate ATI X11 support.

On the good side for me, I'm at the latest and greatest mobile radeon part (4850) so I'm good for a bit. (But I get the feeling that I'd better start looking for a good deal on an 9800M GT MXM part before long at this rate of dropping support...)

It's kind of funny how nVidia supports GPUs way way way back to what the 5XXX series now? Not to mention their proprietary X11 drivers work peachily. I just can't see myself with another ATI card again given their rapid dropping of support v. their only real competitor, and you'd think that ATI would be bending over backwards to at least appear to be giving better support given their lower customer base, but I guess this just goes to show how much ATI/AMD are hurting.

Can't really comment directly on the ATI X11 proprietary driver ATM as I'm making a last ditch effort to get graphical install going for 64b Ubuntu 8.10 before falling back to expert text mode... (I'm not expecting much given various threads here though...)


IIRC last Steam survey that I saw was like 75% nVidia 25% Other including ATI, but rpesumably mostly ATI...

But this rapid dropping of support even if their arch MAY be slightly better than nVidia's is enough to send me right back to nVidia for my desktop GPU upgrade now, where I HAD been considering a 4850 or 4870. Actually with this announcement I'm REALLY wishing that I had gone for the ATI GPU in the desktop and gotten the MSI Gt627 as my notebook(9800M GS) as it's ALOT cheaper and easier to swap a desktop GPU than a notebook GPU even if it is an MXM part...)

Apparently AMD/ATI don't understand seemingly little things like support are in reality MUCH more important than a slightly better/"cooler" piece of sand...

Fool me once, shame on you. Fool me twice, shame on you. Fool me thrice, shame on me. Tell your PR/marketing department that they're doing an OUTSTANDING job for nVidia...

[Stormking]

deary. I've already called you out for trolling. Knock it off.
Well, if you think that dismissing serious criticism as trolling is the right choice - go on, live in your fantasy world where a perfect working fglrx has actually been coded by unicorns ...

[DirtyHairy]

That's not what we are saying. See below.Bridgman: neither did I say I was aiming at you with that comment, which I wasn't :)
That is also what most people were asking for until recently. We provided documentation AND co-funded the initial driver development with Novell AND hired in-house developers and testers to work on the open source drivers. Other than giving everyone a pony, what are we missing ?I'll try to express my criticism once more: from my perspective, I was sold card/driver combi that worked between 50%-80% of the advertised feature set with the continuous promise of things getting better. Now support is being dropped completely, with the alternate open source driver being aroung 50%-80% finished and it being unclear (at least to me) what ATI is going to ensure that I will be able to use that thing in the future. I don't actually doubt that the support in the open source driver will improve in the near future, but I consider this to be a task for the actual hardware manufacturer and not for the community or other companies.

I would not complain if I was passing from a 100% working closed soure driver to a 100% working open source one, or a 50% working one with ATI actively working on delivering the remaining 50%, or if linux support never had been advertised.

So, that's once again my two cents on this, others may of course disagree, and compared to other companies ATI has gotten quite linux friendly, but at least for me, this is still not what I would call "good support". People may agree or disagree with this, but I wouldn't consider it "trolling", "whining" or anything of that sort. Oh, and by the way, I don't file a bug report for every issue I encounter in open source software (as preparing a proper bug report is a time consuming task), but I do it often enough, and also try to fix bugs when I find time, so I wouldn't consider myself "people who have never filed bug reports or gotten involved on any level of development stating they have unfixable bugs, and things are broken, etc, etc, etc.". End of statement from me :)

[BlackStar]

On the good side for me, I'm at the latest and greatest mobile radeon part (4850) so I'm good for a bit. (But I get the feeling that I'd better start looking for a good deal on an 9800M GT MXM part before long at this rate of dropping support...)

Just to put things into perspective, Ati has fully supported R300 chips for 7 years. That's a quite a bit better than nvidia's support for 5xx0 chips.

R500 owners are the ones that are hurt most by this move (total support for only 3-4 years), but at least there are working open-source drivers, which - get this - work *better* than fglrx for anything but 3d games. The real shaft goes to Windows users that wanted to upgrade to Windows 7.

In any case, your R770 will be supported for at least 3-4 years, possibly more. By the time this support ends, the open drivers will be mature and fast.

You may find this move alarming, but the truth is that you won't be affected by it at all. :D

[val-gaav]

You may find this move alarming, but the truth is that you won't be affected by it at all. :D

He will be .. I mean with less things to support in fglrx his card can only work better then worse with it.

So it is a good situation for him ...

[bridgman]

I just bought an MSI GT725 notebook specifically because I wanted to try an ATI card again, but seeing the proprietary X11 AND WIndows(!) driver dumping parts so fast makes me feel even more queasy than I already do looking at the overall lack of adequate ATI X11 support.

On the good side for me, I'm at the latest and greatest mobile radeon part (4850) so I'm good for a bit. (But I get the feeling that I'd better start looking for a good deal on an 9800M GT MXM part before long at this rate of dropping support...)

It's kind of funny how nVidia supports GPUs way way way back to what the 5XXX series now? Not to mention their proprietary X11 drivers work peachily. I just can't see myself with another ATI card again given their rapid dropping of support v. their only real competitor, and you'd think that ATI would be bending over backwards to at least appear to be giving better support given their lower customer base, but I guess this just goes to show how much ATI/AMD are hurting.


Just to be clear here; for Windows we are doing exactly the same as NVidia; shifting to a legacy branch and then updating the drivers on a slower schedule. For Linux, we have already decent open source drivers (an option not available to NVidia today) and most people switching to the open source drivers have ended up preferring them, so we're focusing on continuing to improve the open source drivers rather than maintaining a legacy branch of fglrx.

[lucky_]

Just to be clear here; for Windows we are doing exactly the same as NVidia; shifting to a legacy branch and then updating the drivers on a slower schedule. For Linux, we have already decent open source drivers (an option not available to NVidia today) and most people switching to the open source drivers have ended up preferring them, so we're focusing on continuing to improve the open source drivers rather than maintaining a legacy branch of fglrx.

Moreover, it might even be a good thing for people working on the radeonhd driver, as their user base will substantially grow (judging by the number of angry people here).
It will be profitable for the people who were already using radeonhd for political or practical reason.

I wish I would benefit from an open source driver for my 2400 and my 3470.
(I know it will eventually happen, but still)

[bridgman]

You probably already know the status of open source support for your GPUs, but just in case...

- EXA and Xv are working in both radeonhd and radeon

- code has been merged to master in X drivers, but you still need drm code from the 6xx-7xx branch of mesa/drm or from airlied's drm-next tree

- mesa driver support is "limping" in house; gears is running with some artifacts, working on textures now, will push to public repo soon

[Mr_Ed]

seconding that.
right now I'd say look around for the HD 3870. Okay, Crossfire support isn't enabled yet, so I personally am a bit of a litter box on that, but they still pack a punch as a single card. They also have a wide driver support range, so if something's screwy on the bleeding edge 9.x set, you can still drop back some drivers.

What about the Sapphire HD3850?

On my P4 I am stuck with AGP so the Sapphire HD3850 is as close as I can get.

From what I read of this card it's performance is impressive (the best for what is possible on AGP).

Is it suitable for Linux?

Regards,

E.

[Saist]

What about the Sapphire HD3850?

On my P4 I am stuck with AGP so the Sapphire HD3850 is as close as I can get.

From what I read of this card it's performance is impressive (the best for what is possible on AGP).

Is it suitable for Linux?

Regards,

E.

I honestly wouldn't know. I can't get over the price difference AGP cards carry over their PCI-E cousins... I would "presume" it would work fine, but somebody else who actually has a card would be better suited to answer.

[bridgman]

I guess I'll be the first to suggest you at least think about upgrading to PCIE if your power supply has enough 12v output (older systems needed lots of +5V, newer systems need lots of +12V). You could start with something like a 780-based mobo and still have higher performance than your 9600, although not a *lot* faster.

[Saist]

I guess I'll be the first to suggest you at least think about upgrading to PCIE if your power supply has enough 12v output (older systems needed lots of +5V, newer systems need lots of +12V). You could start with something like a 780-based mobo and still have higher performance than your 9600, although not a *lot* faster.


also gonna make a point here... I just went to Newegg and the cheapest PCI-E HD 3850 is around $65 : http://www.newegg.com/Product/Product.aspx?Item=N82E16814140103

The cheapest AGP 3850 is around $115: http://www.newegg.com/Product/Product.aspx?Item=N82E16814131141

That's a $50 price-tag difference.

That's also just $6 off the difference of getting a well rated Asus budget brand AsRock Socket AM2+ motherboard : http://www.newegg.com/Product/Product.aspx?Item=N82E16813157149 : with a price of $56

So, for the price of buying a single high-end AGP card (which is using a now medium low gpu), you could get the same exact chip, AND a new motherboard, for literally the same price.

It's your money, but if you have $120 to shell out for upgrades, I wouldn't spend it on buyin an AGP card.

***

I also just realized you said P4. I'm presuming here that you are using DDR2 memory on that, so you'll probably be set for memory.

For just $40 you could pick up a Brisbane 4400+ on Socket AM2 : http://www.newegg.com/Product/Product.aspx?Item=N82E16819103221

So, for LESS than the price of a new AGP card for your system you could pick up a motherboard with integrated graphics that are just as fast AND a processor with much better performance. ($56 + $40 = $96).

[BlackStar]

P4 == DDR1. Add $20 for 2GB of memory and you get a whole new system ($116, or the price of the AGP 3850).

AGP parts are not a good buy at this day and time.

[Svartalf]

So, glxgears actually doesn't fully utilize GPU?

In a word, NO. glxgears is so far removed from a benchmark it's tragic. ;) I believe the people associated with 3D drivers on Linux have been saying that it wasn't a very good benchmark for performance since I kind of quit maintaining the Utah-GLX source tree years ago. :D

[Svartalf]

P4 == DDR1. Add $20 for 2GB of memory and you get a whole new system ($116, or the price of the AGP 3850).

AGP parts are not a good buy at this day and time.

In this day and age, unless you're just simply trying to recycle a machine, it's just better if you go get a cheap Core-Duo or Phenom system upgrade for a couple of hundred and get you a new PCI-E GPU card. AGP parts will not perform nicely for you. I'm using AGP because I'm nursing along an Athlon64 Solo2 motherboard for 64-bit and AGP verification support... ;)

[deanjo]

In a word, NO. glxgears is so far removed from a benchmark it's tragic. ;) I believe the people associated with 3D drivers on Linux have been saying that it wasn't a very good benchmark for performance since I kind of quit maintaining the Utah-GLX source tree years ago. :D

In fact IIRC it even originally had to have a switch that was something like


glxgears --I-acknowledge-that-this-is-not-a-benchmark


before it even ran.

[Mr_Ed]

I guess I'll be the first to suggest you at least think about upgrading to PCIE if your power supply has enough 12v output (older systems needed lots of +5V, newer systems need lots of +12V). You could start with something like a 780-based mobo and still have higher performance than your 9600, although not a *lot* faster.

Bridgeman, this confuses me. I thought the issue here was my slow and old videocard and the lack of ATI's driver support for it.

But now you say that it also leis on the fact that I am not using PCIE?

Must I conclude now that what you actually are saying that the AGP bus speed is insufficient?

Elaborate please.

Regards,

E.

[Zhick]

Bridgeman, this confuses me. I thought the issue here was my slow and old videocard and the lack of ATI's driver support for it.

But now you say that it also leis on the fact that I am not using PCIE?

Must I conclude now that what you actually are saying that the AGP bus speed is insufficient?.
No, he just suggests you that if you want to do an upgrade, you realy shouldn't upgrade to an agp-card because they're often problematic (afaik with both fglrx and radeon(hd)). And as others have pointed out, the price difference is so big you could get the same PCI-E card with a new MoBo at almost the same price-tag as the AGP-card,

[Saist]

P4 == DDR1. Add $20 for 2GB of memory and you get a whole new system ($116, or the price of the AGP 3850).


Um. Okay, your memory was a bit better. I forgot that Socket 478 was DDR. DDR2 came with 775.

[Saist]

Bridgeman, this confuses me. I thought the issue here was my slow and old videocard and the lack of ATI's driver support for it.

But now you say that it also leis on the fact that I am not using PCIE?

Must I conclude now that what you actually are saying that the AGP bus speed is insufficient?

Elaborate please.

Regards,

E.

You kind of missed the point. AGP is actually "fast" enough for most graphics operations.

But as I already outlined, you are going to pay as much for an AGP card that IS supported, if not more, than it would cost to upgrade major components of your system to something that has greater performance with lower heat output.

[bridgman]

Yeah, AGP is just a general problem area, more in terms of reliability than speed (although they are related). Some systems are rock solid, others not so much. I think the main problems is that the signal margins in the spec are too small to handle the noise levels found on a typical motherboard, so depending on which components you mix & match you can get hangs which the driver can't always recover from. Slowing down the bus speed often helps but not always.

The PCIE bus is a lot more reliable, and the specs are tight enough that you can mix & match components without worrying how well they will work together.

That said, there is so much good work happening with the open source drivers right now that I expect you could continue using the current card for quite a while longer if you wanted. It sounds like AGP is not causing you any trouble right now; we're really just saying that if you *are* going to upgrade there are better ways to spend the money than a new AGP card.

[Mr_Ed]

Yeah, AGP is just a general problem area, more in terms of reliability than speed (although they are related). Some systems are rock solid, others not so much. I think the main problems is that the signal margins in the spec are too small to handle the noise levels found on a typical motherboard, so depending on which components you mix & match you can get hangs which the driver can't always recover from. Slowing down the bus speed often helps but not always.

The PCIE bus is a lot more reliable, and the specs are tight enough that you can mix & match components without worrying how well they will work together.

That said, there is so much good work happening with the open source drivers right now that I expect you could continue using the current card for quite a while longer if you wanted. It sounds like AGP is not causing you any trouble right now; we're really just saying that if you *are* going to upgrade there are better ways to spend the money than a new AGP card.

Well, I just went back to the radeon driver:

Google-Earth doesnt work with the open source driver.

When DPMS mode is enabled and the screen goes to sleep it will not wake up from it anymore.

Pretty much everything else I do with this driver is at least 50% slower then with the fglrx driver.

i.e. Watching a video on discovery it hicks like hell.

And it is by far not as stable as the fglrx driver.

Sorry for my whining about it.

Oh, and you all say that it is much cheaper buying a pci mobo but you all forget the new cpu and power-supply that I will be forced to buy too. From what I read about it, PCI cards need a lot more power. I am defenitely not going to stick my old P4 CPU on a brand new mobo, that's for sure. If I would have to buy a new mobo I'd probably would go for at least a dual core or maybe a quad core cpu.

I.o.w. I think I'd be better off buying a whole new pc.

[bridgman]

Oh, and you all say that it is much cheaper buying a pci mobo but you all forget the new cpu and power-supply that I will be forced to buy too. From what I read about it, PCI cards need a lot more power. I am defenitely not going to stick my old P4 CPU on a brand new mobo, that's for sure. If I would have to buy a new mobo I'd probably would go for at least a dual core or maybe a quad core cpu.

I.o.w. I think I'd be better off buying a whole new pc.

We didn't forget about CPUs; it's just that a basic dual-core CPU is real cheap these days and is still faster than your P4.

I don't think PCIE cards draw more power (a bit less if anything), it's that over the years systems have shifted from using lots of 5V power to using lots of 12V power, so depending on the age of your PC you *might* need to upgrade the P/S (that was the qualifier in my initial post).

[bridgman]

I feel like I need to repeat something here for everyone.

We are not going to be driving to your house and taking away the fglrx driver you already have, and we are not saying that the open source driver is equal to fglrx in the 3D area for the 3xx-5xx chips (although most other parts of the driver area equal or better for typical consumer usage).

What we *are* saying is that IN THE FUTURE (ie starting a couple of months from now) we think that the open source drivers will be a good solution. If we spent the efforts on supporting legacy fglrx updates rather than open source development the first "quarterly update" would be five or six months from now, and by that time if you could choose between fglrx and the open drivers I think nearly all of you would pick the open drivers.

[Mr_Ed]

At least you keep motivating us :-)

[grazzt]

I feel like I need to repeat something here for everyone.

We are not going to be driving to your house and taking away the fglrx driver you already have, and we are not saying that the open source driver is equal to fglrx in the 3D area for the 3xx-5xx chips (although most other parts of the driver area equal or better for typical consumer usage).

What we *are* saying is that IN THE FUTURE (ie starting a couple of months from now) we think that the open source drivers will be a good solution. If we spent the efforts on supporting legacy fglrx updates rather than open source development the first "quarterly update" would be five or six months from now, and by that time if you could choose between fglrx and the open drivers I think nearly all of you would pick the open drivers.

Non existant TVout in the R4xx open source driver is not "equal or better for consumer usage" IMO.

[susikala]

Non existant TVout in the R4xx open source driver is not "equal or better for consumer usage" IMO.

His point is that most things other than 3D work (well). It boils down to: there's always something to complain about if you want, but AMD's decision is justified enough. It is a _positive_ decision, and reflects the company's healthy attitude to open source in the last years.

[deanjo]

His point is that most things other than 3D work (well). It boils down to: there's always something to complain about if you want, but AMD's decision is justified enough. It is a _positive_ decision, and reflects the company's healthy attitude to open source in the last years.

I think your missing a major selling feature of the IGP solutions. One of the biggest selling points of the RS690 was that it was a viable HTPC platform and without TV-out support those people that bought the board for that use are SOL. The 690g chipset also is still heavily used in many current motherboards.

[DoDoENT]

In a word, NO. glxgears is so far removed from a benchmark it's tragic. ;) I believe the people associated with 3D drivers on Linux have been saying that it wasn't a very good benchmark for performance since I kind of quit maintaining the Utah-GLX source tree years ago. :D

So, what software would you recommend for graphics benchmarking (instead of phoronix test suite)?

And btw. what is the 'fgl_glxgears' and why is it only available when fglrx is installed? How does it differ from glxgears (except rendering a different animation)?

[deanjo]

So, what software would you recommend for graphics benchmarking (instead of phoronix test suite)?

And btw. what is the 'fgl_glxgears' and why is it only available when fglrx is installed? How does it differ from glxgears (except rendering a different animation)?

Using a suite such as PTS is really the only true way of judging real world performance as it gives you a multitude of tests based on real apps and not some synthetic bullshit benchmark. Without a broad span of tests you really cannot isolate what is possibly bottlenecking. Is it driver or just shitty code in the app?

[Saist]

Well, I just went back to the radeon driver:

Google-Earth doesnt work with the open source driver.

When DPMS mode is enabled and the screen goes to sleep it will not wake up from it anymore.

Pretty much everything else I do with this driver is at least 50% slower then with the fglrx driver.

i.e. Watching a video on discovery it hicks like hell.

And it is by far not as stable as the fglrx driver.

Sorry for my whining about it.

Oh, and you all say that it is much cheaper buying a pci mobo but you all forget the new cpu and power-supply that I will be forced to buy too. From what I read about it, PCI cards need a lot more power. I am defenitely not going to stick my old P4 CPU on a brand new mobo, that's for sure. If I would have to buy a new mobo I'd probably would go for at least a dual core or maybe a quad core cpu.

I.o.w. I think I'd be better off buying a whole new pc.

Did you read ANY of my posts?

[Mr_Ed]

Did you read ANY of my posts?

Yes I did. Thank you for your advice.

Regards,

Eddie.

[bridgman]

I think your missing a major selling feature of the IGP solutions. One of the biggest selling points of the RS690 was that it was a viable HTPC platform and without TV-out support those people that bought the board for that use are SOL. The 690g chipset also is still heavily used in many current motherboards.

I thought tvout on 690 was working with radeon. You need to add :

Option "ATOMTvOut" "true"

to xorg.conf. There were some reports of colours not being displayed with PAL systems (ie everything was monochrome), not sure if that is still the case.

[deanjo]

I thought tvout on 690 was working with radeon. You need to add :

Option "ATOMTvOut" "true"

to xorg.conf. There were some reports of colours not being displayed with PAL systems (ie everything was monochrome), not sure if that is still the case.

Composite and svideo are still horribly broken from overscan issues, color anomalies, jittery picture, hardlocks, a freeking round halo around the cursor the size of texas that disrupts pixels on it's edge as well as there is still tearing present even with the newest drivers. All and all, sorry to be blunt, but it is the worst TV out that I have personally seen (including comparison to a bloody SiS and Trident tv out). And forget about a smooth XBMC experience.