Announcement

Collapse
No announcement yet.

AMD Catalyst 7.12 Linux Driver -- The Baby's In Surgery

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

  • Presumably it's related to the PreInitDAL failed error I get.

    Comment


    • they should just rewrite their driver from scratch instead of doing everything at once. which will get them nowhere fast.

      too bad it's not possible with a proprietary solution, though :/
      yep, every month they would rewrite the full driver from scratch!!! what a brilliant idea...

      WTF is ATI doing?
      Now I can't even use my 1400x1050 resolution anymore....
      It seems they have some major quality control issues! So if you're reading this ATI, just drop me a message...

      I got so fed up with the ATI s&*t that I switched my desktop machine to nVidia and I switched back my laptop to vesa. No compiz but I got suspend back! And it just works. Take a hint from that ATI.... I really liked you ATI but you keep delivering drivers that are only getting worse...
      i have 3 different ati processor types and one nvidia and the results are:
      - the ati opensource driver is worse than fglrx on all the boards (r300, r500, r600)
      - the r500 outperforms the nvidia 7xxx serie that i have and the r600 does it too
      - i can use compiz on all of them, but on r300 is slow
      - i don't have all the problems that continue to come out with outher users, or i'm able to fix them when they come out

      now, speaking directly with you:
      1- nobody has forced you to install the new driver. this driver hasn't the quality for being a "distribution" release as stated by ati (michael should be able to confirm it). the last distribution release should be the 8.40 if i remember right.
      2- you have errors?! return to the catalyst 7.11 or to the 8.42. all distributions have their patches to install and make work the last "stable" fglrx driver.
      3- the driver is far from perfect and lacks features, but if you take a look on the progress that amd has made with fglrx in the last year you may be able to say: hey the guys aren't doing so bad. amd has to catch up with nvidia after 5-6 years of lack of linux support. they've done in one year what nvidia has done in more time (they've restarted to write the driver from scratch about one year ago, while nvidia has reached over the years a quite good base).
      4- nobody forces you with amd/ati. buy nvidia if you're not satisfied or wait about 6 months to have a quality driver.
      5- if any amd/ati employee would read your comments i'm sure he'd ignore you thanks to your "kind and gentle" words. if you address to someone like this in real world i'm sure that the most you'd get is a "yeah, whatever..."

      if it's not broken - don't fix it. golden rule when it comes to ati drivers.
      i don't understand your comment in reply to what you've quoted...
      as a software dev i can assure you that a lot of time fixing one issue would come with other worse issues. from i can understand of ati developing i can tell you that the devs are experimenting with portions of lot of undocumented code. i don't know if you have any idea of how hard is to read undocumented code that is written in a bad way and that should do something but it does something else. that's why they've decided to restart the developing from scratch. the problem is that portions of code is not ok for all the boards around and the testing is not enough. you can only test cases that are likely to happen and miss issues because of the code working only for that particular case. i've been through undocumented code that had the problem of running well only in certain circumstances (false positives). this applies a lot to the portions of code ported from the previous fglrx (i'm sure that there are there these portions) which over time would be demystified. don't take developing the driver as a job that is a simple task, cause it's not simple writing code, figuring out what the undocumented code does, writing documentation, testing. it's clear that the number of devs working full time on the code for the moment is not enough and amd has admitted it various times. this lays to the following sentence:
      "bugs that have been around from a lot of time either cannot be fixed without breaking worse the driver or are in parts of code that the devs aren't figuring out what do."

      I guess the question is this:

      I am running a ATI x1650 Pro with driver 8.40.4 (or whatever, the 8.40 one) and everything works pretty smoothly. Should I upgrade to 7.12 just to get AIGLX? Is it worth it?
      in my opinion compiz is useless... it doesn't have any production increase and is only for fun. i can assure you that on a production workstation is totally useless since it drives you mad after less than a hour of work. so, while aiglx is useful only for running compiz, then the reply is clear. if you want to toy around with compiz upgrade else don't do it.
      the reason why i think that you should try at least the 7.11 (i find this the best of the latest releases) is that from 8.41 the driver has inserted a performance increase in them and since the latest fglrx should be quite painless for these boards. if i were to chose from the drivers released, i'd tell you to "test" the 7.12 and if that works remain with it, else downgrade to 7.11. if you experience problems just post a new thread and ask for help. i'm sure that people will try helping you out if they can. the upgrade/downgrade is not difficult and is quite rapid so in the worst case you'll be turning back to 8.40 in a rapid way.

      Comment


      • givemesugarr:
        excellent post and thanks for putting things in perspective. I think, we end users have started demanding a tad too much out of everything that comes out into the market. We should be giving them credit for atleast trying to come upto speed. As a programmer myself, I realise the difficulties faced when putting out a product. And also, someone will always be unhappy (though in this case, the disgruntled lot seems to be a sizable percentage). Having said that, I would expect the features to either work or not work. It is a disappointing to read, FireGL supported, but how well, cant say. I had been waiting to get a release that had all the improvements ATI has been putting into fglrx with support for my card. I am not too keen on using previous releases with PCI-ID patches.

        While I disagree with you on that that compiz is useless, I was able to get by just fine before it came along, it is the smoothness in opening windows (experience), desktop scaling, expo plugins that i find useful, that is a different discussion altogether.
        BTW, I dont know if ATI/AMD deem 7.12 release distribution ready, but they sure have placed it for general download on their website, that has not been the case for certain releases like 8.41, 8.42, 8.43. I might be wrong, but in the release notes for those releases, I remember reading words like not ready for distribution but this one did not have those words. But, on the whole, I agree with you. I am with you ATI
        Last edited by shishir; 27 December 2007, 03:05 PM.

        Comment


        • Originally posted by givemesugarr View Post
          - i don't have all the problems that continue to come out with outher users, or i'm able to fix them when they come out
          [...]
          now, speaking directly with you:
          4- nobody forces you with amd/ati. buy nvidia if you're not satisfied or wait about 6 months to have a quality driver.
          Hi all,

          I read about problems with ati cards and I'm going to buy MB asus p5k-e + ati/asus 3850 512 MB: plz can someone confirm this graphic card works with Debian/Ubuntu/Red Hat/Mandriva/Suse distributions at least for generic and 2D jobs without problems? I use this system for work...

          Thanks in advance...

          Comment


          • Originally posted by gnuser View Post
            Hi all,

            I read about problems with ati cards and I'm going to buy MB asus p5k-e + ati/asus 3850 512 MB: plz can someone confirm this graphic card works with Debian/Ubuntu/Red Hat/Mandriva/Suse distributions at least for generic and 2D jobs without problems? I use this system for work...

            Thanks in advance...
            well, michael reported it to work both with catalyst 7.11 and 7.12 (you'd have 2d and 3d opengl acceleration, hdmi). on the other hand the radeonhd 1.1.0 should support your board too, but it hasn't any 2d or 3d acceleration and has hdmi (i'm not really sure about it though).

            BTW, I dont know if ATI/AMD deem 7.12 release distribution ready, but they sure have placed it for general download on their website, that has not been the case for certain releases like 8.41, 8.42, 8.43.
            if i recall right 8.42 and catalyst 7.11 (former 8.43) was downloadable but not "distribution".

            I am with you ATI
            i'm not with ati since in the last years has only played with our expectations, but with ati devs. it's a little different. i continue to read around that the devs are monkeys and so and that pisses me off a little cause it's obvious that whoever talks like that, in my opinion, doesn't even know what a programming language is... saying this, i'm still waiting to see how the things evolve for the new pc components.

            Comment


            • Originally posted by givemesugarr View Post
              if i recall right 8.42 and catalyst 7.11 (former 8.43) was downloadable but not "distribution".
              I Am not so sure about that. They were downloadable alright, but not on the official driver page that now sports the new 8.443 driver. And, after going through the release-notes, it is clear that they deem it distribution ready.

              Originally posted by givemesugarr View Post
              i'm not with ati since in the last years has only played with our expectations, but with ati devs. it's a little different. i continue to read around that the devs are monkeys and so and that pisses me off a little cause it's obvious that whoever talks like that, in my opinion, doesn't even know what a programming language is... saying this, i'm still waiting to see how the things evolve for the new pc components.
              I am willing to give ATI some time for a big reason, they are releasing their specs to Open Source community with no NDA, which I think is a big deal. ATI devs, it was never a question. When all blame ATI, i assume its the co. and not the devs who are being blamed. A programmer can do only so much, it is the management that decides on the release methodology. If they are so focused on following the 20th of month relesae train model, but willing to sacrifice quality/stability, it is their lookout. They could have kept this release as distribution unfriendly and marked it as an incremental build. That might have helped keeping some disappointment in check.
              Last edited by shishir; 27 December 2007, 10:05 PM.

              Comment


              • Heh... You've no idea how much all the calling the devs "monkeys" bothers the HELL out of me.

                I know what it takes to make OpenGL drivers. Believe me, I know all too well how hard it is. It's not as hard as some would have you believe, but it's NOT easy either. Not by a long shot.

                The people at AMD's ATI division are some of the best and brightest minds in graphics algorithms and programming- really, they ARE. But the problem is that they're TOO FEW due to a lot of serious mis-management by ATI's former upper management (who is no longer there...). Especially when you consider what they have on their plate. Worse, I have a niggling suspicion that many of them know graphics but don't know any driver level innards other than Windows or MacOS. While MacOS might be similar in nature to what we do, it's not similar enough. It's got it's own interrupt rules, memory management rules, etc. Windows... Heh... Anyone that thinks they can safely apply a Windows centric edge to a Linux problem is ignorant or crazy. And if you don't understand the gotcha's of all three interfaces you will have a HELL of a time making the truly cross-platform bulk of what makes a driver work right and efficiently on all three platforms. One is going to suffer badly. Guess which one is the one they don't have a lot of expertise in...

                My frustration is with AMD. They HAD to know that there was some SERIOUS problems with the driver with respect to Linux support. They've had the better part of a whole year to realize they've got the problems they have right now. They failed to hold off and get the documentation out the door like they're trying to do now. They failed to hold the driver back, trying to make a marketing deadline when they KNEW it was not going to pass muster.

                The devs had nothing to do with what is going on. Upper management at ATI and then AMD are the people to be the ones taking the heat here. So, folks, please, please, PLEASE quit ragging on the devs here on this. There's not enough people there to FIX the problems they know are there. They're working on fixing that problem, but that takes some time.

                Comment


                • Originally posted by givemesugarr View Post
                  yep, every month they would rewrite the full driver from scratch!!! what a brilliant idea...
                  I don't think that was what the poster was implying there... I think they were talking about what they had and trying to band-aid a busted design doing major reworks every month, which is what it appears to the uninitiated to be what they're doing.


                  i have 3 different ati processor types and one nvidia and the results are:
                  - the ati opensource driver is worse than fglrx on all the boards (r300, r500, r600)
                  - the r500 outperforms the nvidia 7xxx serie that i have and the r600 does it too
                  - i can use compiz on all of them, but on r300 is slow
                  - i don't have all the problems that continue to come out with outher users, or i'm able to fix them when they come out
                  Lucky you. I've not been QUITE as lucky. Which R500 do you have? An X1950XTX? It's going to bash just about any G70 part out there because it's the baddest, highest performing part in it's class- even with suboptimal drivers it was a suitable choice for Linux (Though for bang for buck, it was a losing proposition- even now it's not the best buy...). I've R500's that can't compete with the G70's I've got. It's all in what part you've got. Furthermore, some of the R500's seem to generate the issues people are seeing. Not all R500's are magic good. Worse, the IGPs all are evil devices, one and all, that should not be compared with discrete chips. Performance and stability are not the same with those parts. I've an R300 part that just isn't good all the way around. I resorted to giving the laptop, reconfigured for XP, to my mother because it wasn't usable by myself- and bought a laptop with an NVidia G70 series laptop part that actually performed WELL under Linux and favorably with everything else out there.

                  You have been lucky. Others have NOT. Please don't be going about claiming everything is just fine (Which is what comes across here, whether that is what you intended or not...) Spare me the line. I've got all kinds of parts and I have to rely on the 3D performance being there along with stability for doing the cross-platform visualization software and the Linux game porting work I do for people. AMD's stuff just isn't there yet for anything other than test for proper function (which much of it fails right now...).

                  as a software dev i can assure you that a lot of time fixing one issue would come with other worse issues. from i can understand of ati developing i can tell you that the devs are experimenting with portions of lot of undocumented code.
                  The driver happens to be derived from their "Orca" codebase. There's VERY little undocumented code and it's all largely new compared to what they've done in the past. That's not the reason. Not at all.

                  in my opinion compiz is useless... it doesn't have any production increase and is only for fun. i can assure you that on a production workstation is totally useless since it drives you mad after less than a hour of work. so, while aiglx is useful only for running compiz, then the reply is clear. if you want to toy around with compiz upgrade else don't do it.
                  Compiz may be "useless" but AIGLX is far from it. AIGLX allows one to have accelerated remote GLX sessions as if they were local. While that may seem like a silly thing to do, it's very, very useful in some visualization applications.

                  Just because YOU see something as being useless or a waste of resources, doesn't make it so.

                  Comment


                  • I should probably mention that "Orca" is just the OpenGL component, so only the OpenGL portion is all-new.

                    When we asked users which new features were most important, AIGLX was apparently a clear #1, which surprised me a bit at the time.
                    Last edited by bridgman; 28 December 2007, 02:16 AM.
                    Test signature

                    Comment


                    • Originally posted by bridgman View Post
                      I should probably mention that "Orca" is just the OpenGL component, so only the OpenGL portion is all-new.

                      When we asked users which new features were most important, AIGLX was apparently a clear #1, which surprised me a bit at the time.
                      I would have thought that that part would have been part of the interface edge for the Orca code. Interesting, John. And AIGLX opens up a bunch of things, Compiz being just one piece of that whole picture. Is it that we got AIGLX earlier than planned that's part of the issues we're all seeing with things?

                      Comment

                      Working...
                      X