Future of my support for ATI

[cruiseoveride]

When a thread starts off by trying to drain blood from a rock, you know as well as I do, it can never be _productive_.

Nothing short of dropping large sums of money at the door of ATi would be considered "productive".

Don't get upset I dissed your friend.

I wont be unproductive again.

[Redeeman]

as much as you might hate to admit it, he is right.

fglrx's failure to run wine is really quite embarrasing, especially since it goes ever unfixed.. the same with the other issues which simply makes fglrx not an option.

I dont care about fglrx, I do however care about documentation, and what the free drivers are achieving, but i believe cruiseoveride is correct, fglrx is more damaging than good - if people run fglrx and get a bad experience, they arent gonna blame the ati driver, but linux, and they are gonna keep that opinion for the next 20 years. and people WILL have a bad experience with fglrx..

i can tell you this much, i want SO much to go out and buy AMD 4870/4850 cards, but every time im about to press order in the webshops, it all ends up coming back to the fact that i dont need a highend paper weight, which is what ill be getting with fglrx..


Yes, AMD is doing progress on fglrx, but really, i do believe the tales of glory and "feature parity with nvidia" and "end of catch up with nvidia" in your articles are really misleading to people..

AMD should be applauded for giving documentation, but for fglrx? hell no.

[Extreme Coder]

fglrx's failure to run wine is really quite embarrasing, especially since it goes ever unfixed.. the same with the other issues which simply makes fglrx not an option.

Yeah.. Even though cruiseoveride's post is a bit too aggressive and harsh, I'm sure it shows how one would feel after a lot of time of waiting for fglrx to get fixed..
TBH, FGLRX works pretty fine for me on any 3d native game, Xv video is working fine for me, the only problem is that I can't run Wine with most 3D games, and 3d and xv flicker in compiz.
I mean, why is it a stretch to expect fglrx having a compatibility as that of nVIDIA in WINE, and non-flicker Xv and 3D sometime this year?

The problem with Wine, if I get it correctly, has something to do with Wine using memory emulation tricks or something, which for some reason don't work with Fglrx. and I doubt that fglrx devs will be doing anything about that any soon, and I don't know why the wine devs fix this.
I sure hope that this is the only problem that needs to be fixed for ATI to have the same level of compatibility as nVIDIA in terms of 3D games.

[siggma]

As the thread starter oh so long ago, and having installed several incarnations of various open and proprietary ATI drivers since, the underlying issue is that there is still no way I can use Linux with an ATI graphics card on a daily basis without watching giggly water looking video tearing in nearly every video that plays.

If you look at the history of ATI-linux drivers, taking r128 for example. How obsolete was the hardware before a decent driver was finally written?

By the time the driver actually worked the Radeon cards were out and relatively inexpensive so most of us upgraded. I clearly recall fighting to get my All in wonder TV tuner to work under linux because ATI offered NO support whatsoever for it at the time (2003). I recall playing the waiting game with the "next" version of xfree86 to offer support.

So, now I upgrade to a new x1650 (r560) card and vuala, no support for most of it's features under linux. Claiming that LInux has a lower market share and therefore deserves less effort or support is nonsense, especially considering ATI has already been paid. The amount of effort aimed at linux support compared to Microsoft cannot be an accident. There has to be some pressure being applied somewhere and I believe it's indirect financial pressure.

The question is:
How does ATI profit more from supporting Microsoft than from supporting Linux?
If the Windows video system makes it easier to implement cards without an overlay, why does it require MORE support efforts?

Again, I sense obfuscation and steering from the ATI representative reading here and have to wonder if there will ever be a usable driver for the card that I already purchased from ATI. If so, how obsolete will the hardware be when it does finally work?

I can think of many solutions to the video issue, all of them very easy to implement. Triple buffering, retrace hold-off, etc. From the number of people complaining about this issue (every person who owns an r3xx+ ATI card running Linux), I have to continue wondering why no solution has been implemented.

So, here's a direct question to ATI:
Why is/was xfreex86 or freedesktop.org supporting ATI hardware?

I've heard lots of "reasons" but none of them make sense except from a political, marketing-ONLY viewpoint. Market share is important to keep a company making products but exclusive support-limited marketing for the sole purpose of steering the public towards other people's products, or a more profitable product is tantamount to a conglomerate-monopoly and clearly socially irresponsible, IMHO.

When a society decays to the point where the term "Business" means "getting as much from you as I can with as little effort on my part as possible", society as a whole suffers. What kind of society will my grandchildren be living in if we all think only of the moment and how profitable this or that is for the select few?

-Tom

[bridgman]

As the thread starter oh so long ago, and having installed several incarnations of various open and proprietary ATI drivers since, the underlying issue is that there is still no way I can use Linux with an ATI graphics card on a daily basis without watching giggly water looking video tearing in nearly every video that plays.

If you play video through OpenGL with "sync to vblank" turned on what do you get ?

Right now video playback through Xv is probably more satisfying with the open source drivers than with fglrx. If you are sensitive to tearing you might want to try Alex's tear-free patch for radeon.

If you look at the history of ATI-linux drivers, taking r128 for example. How obsolete was the hardware before a decent driver was finally written?

By the time the driver actually worked the Radeon cards were out and relatively inexpensive so most of us upgraded. I clearly recall fighting to get my All in wonder TV tuner to work under linux because ATI offered NO support whatsoever for it at the time (2003). I recall playing the waiting game with the "next" version of xfree86 to offer support.

I wasn't involved with Linux back then, but I don't think we even offered support during the R128 timeframe other than supplying programming info under NDA to the open source community.

Also I don't think we have ever claimed Linux support for AIW functionality. Again, I wasn't involved back then but that's my recollection.

So, now I upgrade to a new x1650 (r560) card and vuala, no support for most of it's features under linux.

This statement surprises me. What kind of features are you talking about ?

Claiming that LInux has a lower market share and therefore deserves less effort or support is nonsense, especially considering ATI has already been paid. The amount of effort aimed at linux support compared to Microsoft cannot be an accident. There has to be some pressure being applied somewhere and I believe it's indirect financial pressure.

The question is:
How does ATI profit more from supporting Microsoft than from supporting Linux?
If the Windows video system makes it easier to implement cards without an overlay, why does it require MORE support efforts?

There are some conflicting views of market share, but very roughly Windows has 90% of the desktop/laptop market while Linux has something under 1%, ie Windows market share is roughly 100x that of Linux. I don't understand why you feel that investing resources in proportion to market share is "nonsense", perhaps you could explain ?

We do allocate resources based on market share, although we tend to spend relatively more on Linux than the market share would indicate. The problem is that from a user perspective what you would like is identical investment (or at least identical results) even though the market share and user base is perhaps 1/100th of Windows. Everyone would like to spend more on Linux, but you understand that in a business you do have finite resources to spend (driven by sales) and so every extra dollar we spend on one environment takes money away from other environments where we make the bulk of our sales, and hurts our competitive position there.

It's probably obvious, but that is where the "allocate resources based on market share" rule-of-thumb comes from. Even if we were to obtain 100% of the Linux client market share (ie if our products were so good that nobody bought Intel, NVidia etc.. for Linux) that would still only be perhaps 1/50th of the Windows market and 1/5th of the MacOS market, so the "if you build it they will come" argument doesn't really hold water.

The Windows video system includes support for all the things you are talking about below -- triple buffering, flow control for video through the compositor etc... -- all of those are being added to the Linxu/X stack over time but they are not there yet. Even syncing Xv playback to vblank is problematic and keeps getting yanked and/or rewritten in the open source driver stack.

Again, I sense obfuscation and steering from the ATI representative reading here and have to wonder if there will ever be a usable driver for the card that I already purchased from ATI. If so, how obsolete will the hardware be when it does finally work?

I can think of many solutions to the video issue, all of them very easy to implement. Triple buffering, retrace hold-off, etc. From the number of people complaining about this issue (every person who owns an r3xx+ ATI card running Linux), I have to continue wondering why no solution has been implemented.

See above. Note that the old-style overlay was only removed starting with 5xx/RS6xx so if you use Xv with overlay on a 3xx or 4xx card with open source driver you should get largely tear-free playback. Doesn't help you with a 5xx card but Alex's tear-free patch for radeon should help there.

So, here's a direct question to ATI:
Why is/was xfreex86 or freedesktop.org supporting ATI hardware?

I've heard lots of "reasons" but none of them make sense except from a political, marketing-ONLY viewpoint. Market share is important to keep a company making products but exclusive support-limited marketing for the sole purpose of steering the public towards other people's products, or a more profitable product is tantamount to a conglomerate-monopoly and clearly socially irresponsible, IMHO.

When a society decays to the point where the term "Business" means "getting as much from you as I can with as little effort on my part as possible", society as a whole suffers. What kind of society will my grandchildren be living in if we all think only of the moment and how profitable this or that is for the select few?

-Tom

Not sure I understand your original question. Can you check the wording ?

Over the last couple of years we have been making a bigger investment in Linux support -- first with the new OpenGL driver in Sep 07, and with other improvements and open source support since then. If anyone is expecting all issues to be fixed in the next monthly driver release I have to say "sorry, not likely to happen" but you have to admit there has been a lot of progress over the last 12 months and if you extrapolate a bit I think you will end up quite happy with the results.

If, however, you are expecting an identical investment of resources into Windows (90% share) and Linux (0.8% share) client support, then I hate to say it but you are not likely to be happy with the results from any vendor.We are able to "beat the system" a bit by using common code across multiple OSes, but even that can only go so far.

[energyman]

wow, scale back the paranoia my friend...

[siggma]

If you play video through OpenGL with "sync to vblank" turned on what do you get ?

I'm using the radeon driver because the November fglrx install didn't work for me under Ubuntu Ibex and I don't feel it's worth the effort to try to get it to work unless the issue has been directly addressed. I've installed the repository version of fglrx and it's less functional than the radeon driver, plus last time I installed it the repo had a later version.

But that's not your question, so far forcing vblank (using driconfig) has no effect at all. The driver is allowing writes to the same RAM as the video crtc device. There is no other way this kind of tearing can occur. It's not that complicated to correct which is why I think it's political.

Right now video playback through Xv is probably more satisfying with the open source drivers than with fglrx. If you are sensitive to tearing you might want to try Alex's tear-free patch for radeon.

And after reading further I note this quote from one of the developers:
"The tear-free code is more of a proof of concept than anything else. I doubt it'll actually get merged to master."
(See: http://www.phoronix.com/forums/showt...t=11458&page=3)

So, is your "tear-free" reference a solution or an attempt at political avoidance. If I spent two hours installing, downloading and compiling only to find out the patch is not a fix, how much do i trust what you write next time. Worse, how much do I trust AMD in the future?

I wasn't involved with Linux back then, but I don't think we even offered support during the R128 timeframe other than supplying programming info under NDA to the open source community.

Also I don't think we have ever claimed Linux support for AIW functionality. Again, I wasn't involved back then but that's my recollection.

This statement surprises me. What kind of features are you talking about ?

In reference to what? I didn't mention any features.
The bottom line is the driver didn't work until the card was obsolete and that's what seems to be developing with this driver as well. The r300+ cards have been in production for several years now. How long is too long?

See next post

[siggma]

There are some conflicting views of market share, but very roughly Windows has 90% of the desktop/laptop market while Linux has something under 1%, ie Windows market share is roughly 100x that of Linux. I don't understand why you feel that investing resources in proportion to market share is "nonsense", perhaps you could explain ?

First, if the windows system is easier to get working, and it's now working very well, why haven't resources been allocated to getting AMD products working in those areas there is trouble?

We do allocate resources based on market share, although we tend to spend relatively more on Linux than the market share would indicate. The problem is that from a user perspective what you would like is identical investment (or at least identical results) even though the market share and user base is perhaps 1/100th of Windows. Everyone would like to spend more on Linux, but you understand that in a business you do have finite resources to spend (driven by sales) and so every extra dollar we spend on one environment takes money away from other environments where we make the bulk of our sales, and hurts our competitive position there.

Wait a sec. If it only takes 200 man hours to get a working Windows driver out the door and you're already invested those man hours, what you're saying makes no sense at all. If market share were the only reason to support a product, no support would even be necessary... think about it. Again, I sense there are other reasons and I think it's sneaky and dishonest to claim windows needs more resources because you sell to more windows customers. Especially when the development needs are far less...

If AMD doesn't see Linux as an opportunity, it's really dishonest to string people along by providing unusable drivers.

I think the "competitive position" is hurt worse by being "corporate unresponsive" or worse, "corporate dishonest" to important issues like expecting people to use a driver that won't work. I hear you minimizing the impact of this lack of support, but how much time have you spent trying to watch DVD's or mpeg videos on an r5xx+ system running either radeon or fglrx?

I'd like an answer to the question, not another batch of statistics.

What it comes down to is this:
The current driver(s) are UNUSABLE for the card I paid for. Just because it makes images on the screen does not make it usable. Load up a copy of Ubuntu using an r5xx or r6xx card and drag a window around or fiddle with the autohide menus for a while. You've been responding to these posts for at least a year and I keep reading the same complements about Alex did this and Alex did that. But, it still does not work and that's what really matters.

Load up a DVD and try to watch it on an HDTV before you respond so you are experiencing the same thing as i am. Then we can have a discussion.

-There is no substitute for experience.

It's probably obvious, but that is where the "allocate resources based on market share" rule-of-thumb comes from. Even if we were to obtain 100% of the Linux client market share (ie if our products were so good that nobody bought Intel, NVidia etc.. for Linux) that would still only be perhaps 1/50th of the Windows market and 1/5th of the MacOS market, so the "if you build it they will come" argument doesn't really hold water.

That's not my argument so where did it come from?

I don't care about corporate politics or rules of thumb. I just want to be able to watch videos under Ubuntu using my X1650 video card so I'm not stuck with Microsoft. If rules of thumb are the reason it won't work then AMD is in big trouble.

The Windows video system includes support for all the things you are talking about below -- triple buffering, flow control for video through the compositor etc... -- all of those are being added to the Linxu/X stack over time but they are not there yet. Even syncing Xv playback to vblank is problematic and keeps getting yanked and/or rewritten in the open source driver stack.


See above. Note that the old-style overlay was only removed starting with 5xx/RS6xx so if you use Xv with overlay on a 3xx or 4xx card with open source driver you should get largely tear-free playback. Doesn't help you with a 5xx card but Alex's tear-free patch for radeon should help there.

Another vague reference to a correction. If it were that good why hasn't it been included in the current radeon version? I can't believe it takes four months to include a tear-free patch. If this were a Microsoft issue how long would you expect people to put up with it? It effects fewer people so it doesn't really matter?

I'm not happy because the driver that AMD supplies is unusable and I'm not the only person. Have you read the thousands posts around the internet asking specifically about video tearing in r300+ cards? I'd think it would be a primary issue to deal with, even before hardware acceleration or EXA.

The really stupid thing is that it's not a difficult issue to correct. Don't write to the video page that's currently being output... I think your driver folks call it page flipping. It's even worse when the option is included in the documentation but does not work.

Incidentally, page flipping is included in the current radeon driver xorg.conf syntax and seems to really speed up the display. But, it has no effect on video playback.

How can that be?
I think it's intentional. Sorry to be so blunt but I do. I think it's intended to steer people away from Linux so they'll "give up" and buy Microsoft. What I don't get is what AMD get by steering people towards Microsoft.

So, do I keep purchasing AMD/ATI products or do I assume AMD is in bed with Microsoft and won't really support anything else?

I don't expect my grumblings to change the way AMD does business but they may at least effect the people who read them. I also think there are many, many thousands of people out there with AMD video cards that don't use Linux because it's not well supported. So again I ask, what does AMD get by not fully supporting Linux?

That's the real issue and that's the underlying reason for posting this.

Not sure I understand your original question. Can you check the wording ?

Over the last couple of years we have been making a bigger investment in Linux support -- first with the new OpenGL driver in Sep 07, and with other improvements and open source support since then. If anyone is expecting all issues to be fixed in the next monthly driver release I have to say "sorry, not likely to happen" but you have to admit there has been a lot of progress over the last 12 months and if you extrapolate a bit I think you will end up quite happy with the results.

If, however, you are expecting an identical investment of resources into Windows (90% share) and Linux (0.8% share) client support, then I hate to say it but you are not likely to be happy with the results from any vendor.We are able to "beat the system" a bit by using common code across multiple OSes, but even that can only go so far.

Political rhetoric and I'm not in the market for rhetoric.
If there is a way to watch video under linux without all the tearing I'd like to hear about it. I suspect many people just toss them in the trash or post them on ebay. It may not effect the business now but it will in the very near future. Especially as the general public that grew up on computers becomes the prodominant purchaser of computer hardware.

We have a new President...every four years. Millions of children are born every year and every year the previous year's generation graduate High School and enter the marketplace as purchasers and producers. How many of them are willing to accept the deviant definition of "Business" that's become the standard in this century. How long can a society or our country afford dishonest business practices.

THIS:http://www.phoronix.com/forums/showp...&postcount=115 is exactly what I'm talking about. Third party interests are more important than customer support or a working product. It's really immature and very socially damaging. This kind of "gimme gimme" egomania is one of the reasons our economy is on the edge of a precipice...
-Tom

[bridgman]

In reference to what? I didn't mention any features.

Your comment was "So, now I upgrade to a new x1650 (r560) card and vuala, no support for most of it's features under linux." I was asking which features you were talking about.

I think it's intentional. Sorry to be so blunt but I do. I think it's intended to steer people away from Linux so they'll "give up" and buy Microsoft. What I don't get is what AMD get by steering people towards Microsoft.

If we wanted to do that, why would we bother investing in Linux support at all ? Wouldn't it be easier to only support Windows and MacOS ?

The current driver(s) are UNUSABLE for the card I paid for. Just because it makes images on the screen does not make it usable. Load up a copy of Ubuntu using an r5xx or r6xx card and drag a window around or fiddle with the autohide menus for a while.

My primary machine is running Ubuntu 8.10 with the in-box open drivers on an X1950PRO (ie rv570), running Compiz with wobbly windows, rotating cube etc... No artifacts or problems that I can see. Haven't played any DVDs though, just video files including some high-def H.264 files (eg Big Buck Bunny) which all play very cleanly. I have a fairly recent quad-core CPU though, maybe that is helping.

I'm currently running radeon rather than radeonhd because the radeon package on 8.10 works properly with 3D while the radeonhd package does not, but I'm hoping to get the latest radeonhd built and running later this week.

Another vague reference to a correction. If it were that good why hasn't it been included in the current radeon version? I can't believe it takes four months to include a tear-free patch.

The patch blocks all rendering activity until the next vblank. If your primary use of the system is for video playback this isn't a problem, but it does slow down 3d and some 2d drawing. Alex was pretty clear that this would probably not make it in as a general purpose solution, just as something which would help users whose primary application was video playback.

If you follow the mailing lists and IRC chat you will see that the underlying mechanisms for synchronizing the ddx driver with vblank is being redesigned -- current thinking is some kind of helper thread hanging off drm -- but that there is debate about whether even this approach is worth doing or whether the whole mechanism should be re-implemented around compiz (since the path from initial render to user visible screen is quite different once you add a compositor).

The reality here is that there are relatively limited resources available for re-architecting the X/DRI stack, and that right now video playback is not on the top of anyone's priority list. Almost all of the improvements users would like to see in Linux graphics first require a standard memory manager in the kernel (ie drm) -- that, plus the directly related initiatives (kernel modesetting, DRI2/RDR) are the top priorities right now. Once those are done, I think you will see much faster progress in both OpenGL support and video playback.

[Melcar]

@ siggma:

Just so you know, most of the complaints you pose already have been discussed to death on previous occasions, and bridgman & Co. have already thoroughly explained the situation. Repeating the same points over and over again comes out a bit like whining.