Announcement

Collapse
No announcement yet.

7.12 - speculation time...

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • The majority of what ATI owners are moaning about is pretty basic stuff, it's not much to ask for working drivers. Rialto based cards work but badly, poor scrolling in Firefox, major memory leak in any 3D app etc. Frankly ATI/AMD should be ashamed of their drivers atm.

    If you release a driver that isn't marked as unstable/alpha/beta it should work! Major show stoppers simply should not get through the QA department. Phoronix have reported how drivers are released with a rolling system, a driver spends at least a month in QA.

    We are also not moaning about something provided for free. We all spent a good amount of money on the hardware, we expect decent drivers under supported OS's.

    Having said all that AMD have released the new code base, CCC is finally getting ported to Linux and Michael hinted that Rialto based cards are being properly supported next release. I can't wait, I'm hoping it will fix most of the major problems and they can go from there adding new features and better performance.

    Comment


    • Originally posted by Svartalf View Post
      Heh... It probably should have been moderated "insightful" instead of "funny" back when he made it- because he wasn't kidding when he posted it. In many ways, the RE'd driver worked more stably and better than fglrx in many 2D situations.

      It's why we're dead keen on seeing what they release in tech data to help us help them. They're just taking longer than they probably ought to considering the sad state of things in the proprietary driver. Now, if they can fix tens if not hundreds (yes...) of serious flaws in the new codebase driver (things like that bad memory leak on certain chipsets, things like glaring omissions of OpenGL 2.X support that everybody else that can support 2.X HAS- including the OLD codebase driver...), that would also be a good thing.
      My suggestion to ATI would be to drop fglrx completely
      and instead integrate the GL blob with the RHD driver, and assign the people working on the 2D part
      to it. So at least 2D would be well supported, and
      the adventurous folks can have a bleeding edge 3D blob that doesn't depend on a particular kernel and
      xorg version (I bet the freebsd people would like that too).

      Comment


      • As long as they continue with the improvements in each release and keep up to the promise of more gpu documetation for the open source world i see great future for Amd/Ati.

        The situation is not as bad as people want it to be. After Amd take over Ati things seems inded to be improving. Thats atleast my opinion.

        I now got 2 different ati architecture(2600pro x600xt) working under linux, and have no need to change them even thought i got a 6600GT ready if i need it.
        The driver have inded improved , and each new release brings some fixes with it. Im looking forward to catalyst 7.12


        and btw:Nvidia cards are not totally bug free either

        Comment


        • Originally posted by mlau View Post
          My suggestion to ATI would be to drop fglrx completely
          and instead integrate the GL blob with the RHD driver, and assign the people working on the 2D part
          to it. So at least 2D would be well supported, and
          the adventurous folks can have a bleeding edge 3D blob that doesn't depend on a particular kernel and
          xorg version (I bet the freebsd people would like that too).
          I agree with you, but as far as I know, I don't think it is possible until they don't release the specs (read most of the specs).
          And what about older graphic card ? Maybe they simply cannot release anything for older models they inheritate from ATI.
          But for sure the best solution would be a common open source base with some bits of closed sources.

          I won't buy new hardware for a while but by the end of next year, AMD gpu are well supported by an open source driver then I will buy AMD.
          As an owner of a firegl v3200 (notbook t43p) I don't use fglrx for about 6 months.
          It's insane that we always have to choose between an old kernel or an old driver. When you compare to ath ati driver. Speaking of QA, when you see that the ati driver enters in ditributions after quite immediately, without much trouble... For instance gutsy release by the end of october had to provide 8.37.6.
          Last edited by lucky_; 12-18-2007, 12:49 PM.

          Comment


          • Originally posted by teroedni View Post
            As long as they continue with the improvements in each release and keep up to the promise of more gpu documetation for the open source world i see great future for Amd/Ati.

            The situation is not as bad as people want it to be. After Amd take over Ati things seems inded to be improving. Thats atleast my opinion.

            I now got 2 different ati architecture(2600pro x600xt) working under linux, and have no need to change them even thought i got a 6600GT ready if i need it.
            The driver have inded improved , and each new release brings some fixes with it. Im looking forward to catalyst 7.12


            and btw:Nvidia cards are not totally bug free either
            Well, nothing is bug free really...but Nvidia cards have far less (fatal) problems then the ATI binary equivalent.

            But yes, the situation is at least seemingly getting better. Only time will tell now though.

            Regarding the suggestion to drop fglrx, it won't happen. Not all documentation can be released...specifically anything containing content protection mechanisms as well as other items that has not crossed my mind yet. For the users who want the 'full functionality' they will still need to use fglrx.

            Comment


            • Originally posted by Svartalf View Post
              One should be very, very careful about making comments like this... You might find yourself embarrassed.

              Among my open source credits...

              Versalogic VSBC6 Industrial I/O drivers for Linux

              Maintenance of the Intelogis Parallel Port networking drivers.

              Developer (one of many...) and one-time maintainer of the Utah-GLX drivers.

              Several different USB check scanners...

              I've done numerous closed-source drivers for companies as well, under contract, for Linux.

              I know device drivers. It's my bread and butter, if you will. Just because I do consulting work for LGP, it's not the sustaining income, not because it can't be- it's because I'm making a lot more money doing the driver and embedded systems consulting than I could doing the LGP work alone.

              I even KNOW how hard it is to do 3D drivers. It's harder than one would think, but not as hard as you're making it out to be.

              To each and every one of the people commenting on that we shouldn't be complaining like we are, I will YET AGAIN observe that if this was Windows they were doing this stuff on they would be out of business- period. It doesn't matter that you think that we should be nice. Sorry, we've been that for some time now- if you're just coming in and think that this is going to get you anywhere, you're sadly mistaken. In the 12+ years I've been doing professional Linux development work, it doesn't matter if you're nice or not. Either they do it or they don't- you're viewed, rightly or wrongly, as a very, very small niche market at best.

              I know, I see it every day now. Have been for 12 years.

              It is flatly not acceptable to ship drivers with the kinds of memory leaks that 8.11 shipped with, and the three previous versions shipped with. I don't care how "nice" they're being- it's NOT acceptable however you slice it. This is coming from a software engineer with over 20 years of industry experience.

              Don't expect me to be nice (though I really am being that way, though you might not see it...) when they do things like that.
              As replied, Linux Community is growing, growing fast. In my experience with Gentoo, I have not been able to experience such memory leak, I say this to illustrate my point, ATi has merged with AMD, AMD then said that support for Linux users would be increased. Did happen, they are releasing the specs on their chips and started working on a driver for Linux, unfortunately, when one says Linux, one may refer to ANY of the current distros, each customized with a number of configurations for these systems as large as the community itself. Too many variables to check, Before the merge, ATi only existed for Windows, so why would they need to even botter to support a part of the market that is non-profitable?. That was, now that the head has changed, the devs are working in something (I hope so) relatively new to them, I think that time will tell... It seems that ATi is beggining its experience with Linux from 0.

              Comment


              • Do you know if we will have the Catalyst driver of December before the Christmas week?

                Comment


                • Originally posted by pierluc View Post
                  Do you know if we will have the Catalyst driver of December before the Christmas week?
                  It will probably before Christmas yes, each driver is released around the 20ish of the month.

                  Comment


                  • Originally posted by Crunchy View Post
                    The majority of what ATI owners are moaning about is pretty basic stuff, it's not much to ask for working drivers. Rialto based cards work but badly, poor scrolling in Firefox, major memory leak in any 3D app etc. Frankly ATI/AMD should be ashamed of their drivers atm.

                    If you release a driver that isn't marked as unstable/alpha/beta it should work! Major show stoppers simply should not get through the QA department. Phoronix have reported how drivers are released with a rolling system, a driver spends at least a month in QA.

                    We are also not moaning about something provided for free. We all spent a good amount of money on the hardware, we expect decent drivers under supported OS's.

                    Having said all that AMD have released the new code base, CCC is finally getting ported to Linux and Michael hinted that Rialto based cards are being properly supported next release. I can't wait, I'm hoping it will fix most of the major problems and they can go from there adding new features and better performance.
                    Can't agree more. We pay for our cards with hard earned money. The very least would be reasonable drivers/specs? Luckly AMD/ATi is working on it, and it takes time, I know. I'm glad enough that it is happening.

                    BTW I don't think CCC is being ported, as the win32 CCC is .net and we don't ahve to luckly endure that horror.

                    P.S. Micheal, i know NDA n stuff, but something 'I have or haven't tested it' would have been _some_ answer It would be nice if I finally got to use my card is all

                    Comment


                    • Originally posted by mlau View Post
                      My suggestion to ATI would be to drop fglrx completely
                      and instead integrate the GL blob with the RHD driver, and assign the people working on the 2D part
                      to it. So at least 2D would be well supported, and
                      the adventurous folks can have a bleeding edge 3D blob that doesn't depend on a particular kernel and
                      xorg version (I bet the freebsd people would like that too).
                      There's a serious impedance mismatch between what they're doing in the fglrx driver and what we're going to be able to do with the 2D driver in x.org. It'll be lessened once the AtomBIOS stuff gets fully implemented and gelled, but it's still got a lot of nasty things going on that won't mesh up nicely. Ideally, they need to clean up their stuff or go all-open on the Linux side of things. They're going open enough that if we get what they've been promising us, they'll be doing better overall, even if they don't get the fglrx mess cleaned up quick enough.

                      Comment


                      • Originally posted by lucky_ View Post
                        And what about older graphic card ? Maybe they simply cannot release anything for older models they inheritate from ATI.
                        Well, I believe that most of the R200 stuff is being discussed at this point as well as the R300/R400 and the R500 and R600 info. As it stands, I do need to get back in touch with John Bridgman to see if they're okay with releasing the sanitized RagePRO/Rage128 spec data that I had been given for the UtahGLX project work I'd done- but the both of us seem to be busy. (I'm hoping it's due to trying to get more people in there to fix the problems his employer has on the proprietary driver and possibly to get the specs out the door. I would honestly like to see the fglrx driver do much better and to see us catch mostly up with them and make this a win-win situation for ALL.)

                        We're just going to have to see what comes of 7.12/8.1 and go from there, I guess.

                        Comment


                        • Originally posted by Svartalf View Post
                          Considering that NVidia got SUED for their "Vista Ready Drivers" that had similar issues on the Vista side, I think you might want to stop and think about what is all going on here.

                          If you think I'm being negative, heh, what the heck was that from the Windows world?
                          See I'm talking about the development environment , if anybody thinks they can win by suing AMD/ATI about the linux driver then feel free that'll do something , bitching about it towards the developers isn't going to do jack. See if anybody wants to sue AMD/ATI then I have absolutely no problem with that , that may actually improve the situation , it's the actions that are at best neutral and at worst harmful that I don't see the point of.
                          Originally posted by Svartalf View Post
                          I'm just stating facts here- pointing out serious problems which should have been show stoppers but weren't. To not do so is sticking one's head in the sand.
                          There is absolutely no problem with pointing out problems and being helpful , however there is a vast difference between that and what some people are doing here.
                          Originally posted by Svartalf View Post
                          And you mention moving on, etc. What do you think this is, hm? You're taking me to task for pointing out unacceptable things coming from them and telling them that I'm moving on.
                          I haven't taken you to task for moving on , in fact I've suggested it and telling AMD/ATI Management, let them know they are losing customers because of their linux driver quality.
                          Originally posted by Svartalf View Post
                          Professionally, I can't really do that (I have to use their stuff to test for function on the stuff I do so that I know it works under Liunx.
                          Well then you can either sue them , or contact them (and probably pay them) for support. If it's such a big problem. If it's not then suck it up and wait. That's part of what the professional world is about.
                          Originally posted by Svartalf View Post
                          They COULD have better stuff coming out than they fielded at this point. They still can. Whether their upper management opts to give out adequate resources for the people working on things to do so, remains to be seen. The certainly have had the entire year to get it to this point and they haven't managed it so far- not because the people working on it are unable because of skills or competence, but because they're too few to be actually working on the stuff, especially in the stated development process they're using to produce the proprietary driver.
                          So you admit the developers can't do anything? So why condone people making the development environment hostile for them? How is that going to attract more developers? which AMD/ATI is having trouble finding?

                          Once they find the developers they've got to get them up to speed on the hardware, current driver, and AMD/ATI corporate culture. This takes time. I'm not surprised it's taking them this long. Yes it's unfortunate and if I was a business customer of them or someone else who distributes their products where it was costing me money and the risk/reward ratio for suing them was in my favour too bloody right , I'd sue them but I'm not so I have too look at what's the best I can do in my situation and that's hope some big boys sue them , speak out against people who are making the environment for the developers hostile and persuade them to focus their efforts where it'll actually do some good.
                          Originally posted by Svartalf View Post
                          What? I'm just supposed to "suck it up" and hope they get their act together? Not likely. You state that this is a problem with Linux getting ahead. Nope. Not even close. They end up encountering as bad or worse (Way worse...being rebuked online like this is nothing compared to getting sued over something...) stuff from the Windows world- and people still keep making stuff for it. Eagerly.
                          There is a vast difference between what has been going on here , that has been targeting the developers and instead doing something that will make AMD/ATI's management take notice.

                          Without talking to intellectual property lawyers you might find the case against ATI on the linux driver side much harder to win than on the windows side.

                          Windows is also a much bigger market and that changes the game an incredible amount. We only have a couple of developers for linux and AMD/ATI needs more , if I was a potential developer for the linux driver I'd have a long around , see what's going on here and think fuck that for a joke.

                          If someone asks whether an AMD/ATI graphics card is suited for linux then by all means tell them no , but there are people here calling the developers monkeys , morons and idiots and suggesting when people don't ask that they change to NVidia when they post a problem.

                          If we could somehow have a positive development environment for the developers and maturely focus the negative aspects about the quality of the linux drivers and the consequences of that on AMD management I'd be perfectly happy.

                          Instead we've got a bunch of folks who are annoyed they are having so much hassle with their gfx card on linux and taking it out at anybody they can.

                          I'm simply pointing out in the long run that unless it's focused in a mature manner (which may very well involve the courts) on AMD management then they aren't helping the situation and are only going to prolong their problems.
                          Originally posted by mlau View Post
                          My suggestion to ATI would be to drop fglrx completely
                          and instead integrate the GL blob with the RHD driver, and assign the people working on the 2D part
                          to it. So at least 2D would be well supported, and
                          the adventurous folks can have a bleeding edge 3D blob that doesn't depend on a particular kernel and
                          xorg version (I bet the freebsd people would like that too).
                          Do you realise how difficult that would be? It would be more difficult than getting the whole thing to work properly.

                          Comment


                          • Ze. The community has waited a long time for them to get their act together, probably too long... Been there, done that. The only ones making it hard for the developers to actually code the driver were ATi themselves by simply having an understaffed team, and a HUGE problem on their hands, not to mention that every ~6 months a new product would come out and every ~18 months a whole new generation of products and hence underlying architecture. It is disproportionate to compare the Windows Driver team and the Linux Driver team, at one point IIRC one of us in the Rage3D forums joked about how many Windows Driver developers were there for one Linux developer.

                            Svartalf is not criticizing the actual merits of the driver team, particularly Matthew Tippet (leader of the Linux driver team), but rather that the gargantuan task they are assigned to is simply too big and they are just too few to accomplish it in a reasonable amount of time... And by that I mean graphics market reasonable amount of time, which is ~18 months... That was what the community was expecting (and still is). In the current state of things AMD/ATI have indeed made a HUGE leap forward (from where I left them, anyway), but they still have an equally or even bigger gap to close before things will be smooth enough, and I don't mean speed necessarily (see Svartalf points about OpenGL spec implementation, basic X features, etc, etc, etc).

                            If anyone made the development environment hostile to the developers were ATi themselves, as I said before. With this new open way of doing things, I'm sure ATi will see a much better reaction from the community, not because of the open driver, but because most likely things will start to "just work" (at a LONG last!). AMD has the great opportunity by maintaining a dual model driver (open and proprietary) to actually benefit from the open development on the open source driver and implement that into their proprietary driver, but keep the open driver so "things just work" in new systems and installations.

                            Most of us former ATi users grew skeptical of ATi/AMD on Linux. Not for the people behind the code, but for the people higher than these in management. I agree with most what Svartalf said about the current situation and what does the opening of the specs and an open driver might mean for ATi/AMD, I don't think ATI/AMD will regain the lost users any time soon, but for new users, and people buying systems integrated with ATI/AMD parts, this means that ATI/AMD will at least be able to start having a growing user base in Linux, and these users won't
                            1. Go running in tears back to Windows.
                            2. Go out and buy an nVidia card.

                            Like you, I don't think anyone will sue ATI/AMD for their Linux drivers any time soon, and the market segments that could, have moved on to that other IHV, cheaper than any legal action, and you get actual work done; However, I don't think Svartalf implied that anyone would do so either.

                            Comment


                            • Originally posted by Svartalf View Post
                              Well, I believe that most of the R200 stuff is being discussed at this point as well as the R300/R400 and the R500 and R600 info. As it stands, I do need to get back in touch with John Bridgman to see if they're okay with releasing the sanitized RagePRO/Rage128 spec data that I had been given for the UtahGLX project work I'd done- but the both of us seem to be busy. (I'm hoping it's due to trying to get more people in there to fix the problems his employer has on the proprietary driver and possibly to get the specs out the door. I would honestly like to see the fglrx driver do much better and to see us catch mostly up with them and make this a win-win situation for ALL.)

                              We're just going to have to see what comes of 7.12/8.1 and go from there, I guess.
                              It would be awesome to see spec for this cards, in this case may we see a merge between ati and radeonhd ? (one driver to rule them all). The killing question is : how long will it take to be that smooth ? i.e How long people should delay their buying to have an amd experience as smooth as with intel ?
                              Considering that there is very little people buying amd/ati powered lappy due to the current situation since at least one year (at least for linux enthusiast). It sounds pretty pessimistic but I still hope for the good time to come, AMD cpu's always satisfied me.

                              Comment


                              • Originally posted by Ze.. View Post
                                So you admit the developers can't do anything? So why condone people making the development environment hostile for them? How is that going to attract more developers? which AMD/ATI is having trouble finding?
                                I did no such thing. What I stated was that they have too few people to manage to pull together a version of the codebase from the Vista/XP R600 driver in a timely manner. They're shipping code that intrinsically had problems to begin with (missing OpenGL 2.X functionality, amongst other things...) and that the bulk of the few people they do have don't "get" Linux, never mind that it's easier to develop drivers for it than under Windows. There's even as good or superior tools for free that they could have used to find many of the issues they have outstanding- but they didn't bother to even go looking. The big problem is that there's not a lot of OpenGL driver developers out there, available for AMD to hire- most of them work for Intel or NVidia now; ATI kind of hurt themselves by their upper management choices and now AMD has to fix that mess.

                                Once they find the developers they've got to get them up to speed on the hardware, current driver, and AMD/ATI corporate culture.
                                Heh... That takes a lot less time than one would think. Biggest complaint that I have is that everyone over at AMD knows what the real market is (Larger than you'd think- and getting larger by the day, believe it or not...) and they didn't do what needed to be done to resolve the problem. That's one of the places that I've my biggest beef.

                                There is a vast difference between what has been going on here , that has been targeting the developers and instead doing something that will make AMD/ATI's management take notice.
                                You'll note you're verbally jousting with me, and not the ones taking the swipes at the developers- and I've NEVER done so. In fact, I've pointed out that it's not the developers that're the problem and tried to get people to tone that crap down as well.

                                Without talking to intellectual property lawyers you might find the case against ATI on the linux driver side much harder to win than on the windows side.
                                Open sourcing the driver is probably out of the question. Releasing the technical data, sans anything to do with things like Macrovision control, isn't- and should never have been an issue. AMD's rectifying that one right now, they just got to make sure there's no leakage of protection stuff, etc. so it takes time for them to validate that they're doing their due dilligence there.

                                Windows is also a much bigger market and that changes the game an incredible amount. We only have a couple of developers for linux and AMD/ATI needs more , if I was a potential developer for the linux driver I'd have a long around , see what's going on here and think fuck that for a joke.
                                Actually, it's not as big as you'd think. Microsoft's numbers are heavily skewed- and that's what everyone and his dog is going off of. But, even THAT impression is changing. I watched at one of my client companies where not one, but TWO of the main OEM players (you figure which two out of Dell, Lenovo, IBM, HP, etc. they were... ) tell the client company that they had to have compelling performance on their network part ON Linux, preferably Open Sourced driver support- or they weren't going to carry their product. PERIOD. It was part of why they'd hired me- to salvage the situation.

                                If someone asks whether an AMD/ATI graphics card is suited for linux then by all means tell them no , but there are people here calling the developers monkeys , morons and idiots and suggesting when people don't ask that they change to NVidia when they post a problem.
                                Most of the people in the groups weren't doing that until some of the AMD fanboy types (Yes, there were some in here...) started making snide remarks. Someone made an ill-chosen choice of comments and then it really was on. Most of the people on this forum actually DID what you stated. Myself, being one of the more prevalent commentators on the subject in question.

                                Until AMD gets it straightened out, one way or another, their GPU products, under Linux, is solely for the brave at heart or the lucky. That's what most people have actually been saying.

                                If we could somehow have a positive development environment for the developers and maturely focus the negative aspects about the quality of the linux drivers and the consequences of that on AMD management I'd be perfectly happy.
                                You'll find that this was a situation that got out of hand because someone took a swipe at the people indicating that if you could, you'd be better off with an NVidia card instead of an AMD part because the AMD stuff didn't work right for most people- or that they were tired of waiting for AMD and getting an NVidia part. 'GTFO' (To Crunchy: Sorry for bringing it up, I know you weren't being a fanboy just being as irritated as the rest, but it IS what started the crap...) was what was chosen to be used in response to just what I indicated- and then it was off to the races with the fanboy crap from both sides. And, I've been trying ever since to get people to cool their jets. It's not working. I wonder why?

                                Instead we've got a bunch of folks who are annoyed they are having so much hassle with their gfx card on linux and taking it out at anybody they can.
                                YOU try dealing with a part that the supplier claims is supported by the driver, only to find out that it only works in a degenerate, suboptimal mode of operation- many of the Xpress200M owners have had that problem for YEARS now. They know it was broken. They didn't fix it. Until recently, the very people responsible for not giving the resources to the Linux driver team to FIX that problem were still in charge over at the former ATI division of AMD. It's only probably changed, either in management view or in management personnel, with regards to what needed to be done at AMD for this problem within the last 4-6 months (About the timeframe for the sudden about face on things Linux)- but they had nearly 12 to actually have done this.

                                The average person doesn't know whom to blame, so they pick the "likely" target. I don't and I've been trying to get them to quit taking potshots at the devs.

                                I'm simply pointing out in the long run that unless it's focused in a mature manner (which may very well involve the courts) on AMD management then they aren't helping the situation and are only going to prolong their problems.
                                Actually, they've got a clue. It's just taken them nearly a year longer than it ought to have. And it'll take them about 6-12 months, unless they get lucky with staffing, to have them see a real impact on this turn around because it's 6-12 months later than it should have been.

                                It's still inexcusable to ship with the problems that the current new codebase drivers have. I know that the devs are doing what they've been told. But the middle management should have planted their feet in the ground like a mule or donkey would have and said that this stuff wasn't at all ready. It's very apparent that their QA staff doesn't know what to test under Linux or things are getting swept under the carpet.
                                Last edited by Svartalf; 12-19-2007, 10:17 AM.

                                Comment

                                Working...
                                X