Announcement

Collapse
No announcement yet.

NVIDIA Is Still Killing AMD Over Linux OpenGL Performance

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

  • #41
    Originally posted by smitty3268 View Post
    Question for anyone who knows. What's up with this "privative" word? Is that some kind of common slang in europe?

    I've been seeing it used all over the place recently, from a bunch of different people. It can't be people just copy/pasting, i don't think.
    I was wondering the same thing so duckduckgo'd it and got "Causing deprivation, lack, or loss." From that I assumed it was some kind of pun, but then again maybe it wasn't intentional.

    Anyway, I've been considering an upgrade sometime this year from my 6550D (on r600g) to probably the R7 260X. Based on recent benchmarks it doesn't look as though that upgrade will gain me much just yet. Maybe in 6 months either radeonsi or nouveau will be on par with the quality of r600g and I'll make my choice then.

    Comment


    • #42
      Originally posted by ChrisXY View Post
      You can be both. Three possibilities: Either you are fine with supporting the status quo and keep doing whatever you do, or you are not and try to change it or you don't care.

      So I take it you will not buy a fairphone until it actually is 100% fair, right?


      Try using a real? window manager like kwin and set vsync to repaint the full screen every time. There's some explanation why this is actually intel's preferred method of repainting the screen that you can google if you like.
      1. The plague of tearing on intel is common across linux desktops. You claim kwin to be superior while every article I've read on the subject state that full-screen repaints is just a workaround making any other affected compositor as real? as kwin is. For Gnome Shell's Clutter it's CLUTTER_PAINT=disable-clipped-redraws:disable-culling and CLUTTER_VBLANK=True options.

      While we're at it it's worth highlighting that some genius behind the intel tearing bollocks is responsible for tearing on linux desktops unfixed for over a year! Until Xorg 1.15 and Mesa 10 land in distros of those affected by the issue.

      2. I failed to get any interesting results on intel's preferred method of repainting.

      Originally posted by zanny View Post
      Then why even use Linux? It is so easy to just be in the Windows world, where the NSA can spy on you and you have to pay the same company every couple years to get the latest versions of the same software. Where if you don't like the network traffic algorithm, you kind of have to deal with it because the whole OS is a black box.

      Same can be said of gpus - oh, your hardware is old. So we're not supporting it anymore. Have fun! Or, a new API comes out - like opencl - oh hey, yeah, we're a competitor, we're not going to support it. Or maybe we are going to do software crippling of certain hardware parts to make larger margins with the same silicon. Such an easy way to raise profits, and if the drivers are blobs and undocumented, hey, not like you can do anything about it right?

      Or your parts don't work with non-Android kernels. Because our driver is a big fat blob of who knows what running on your cpu. But hey, its convenient, so why care? Or more appropriately, its convenient until it isn't, at which point you are absolutely fucked.
      Because contemporary Linux works! And that's all there is to it. It's [works] against [doesn't work*, but oooh, I'm all about freedom].

      *or has performance crippled by ~20%, which is same, as far as I'm concerned.

      Originally posted by zanny View Post
      (...)oh, your hardware is old. So we're not supporting it anymore. Have fun!
      It's not wise to mention such things waving your pro-AMD banner, you know... Does AMD declaring anything older than radeon 5000 'legacy' ring a bell?
      Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

      Comment


      • #43
        Originally posted by squirrl View Post
        Nvidia is the only graphics company actively adressing performance of it's cards underneath Linux.

        AMD has good drivers, PS4/Sony is somehow getting a great driver for their FreeBSD kernel.

        Somebody is lying. I've got an APU -- e450 -- and it just midway through last year got a decent driver where XBMC would play 1080p without stuttering.
        It's a beta driver.
        I had to hack out the watermark. Fantastic!

        Quality work.
        Huh? I used the 12.10 catalyst on my e350 for quite some time with xvba-xbmc build. Worked very nice! Only the newer catalysts had a xvba bug - do not use those.
        At the end of 2011 it even worked with vaapi-xbmc and older catalysts, also fluent. Maybe because I compiled xbmc myself...

        Two months ago I had to upgrade to mesa-driver cause catalyst 12.10 supported only x-server 1.13. But with xbmc13alpha the oss driver uses vdpau for xbmc and thus is as great (even better in stuff like HD audio support) as the xvba support.

        Cheers tomme

        Comment


        • #44
          Originally posted by Bucic View Post
          It's not wise to mention such things waving your pro-AMD banner, you know... Does AMD declaring anything older than radeon 5000 'legacy' ring a bell?
          http://phoronix.com/scan.php?page=ar..._legacy2&num=1
          You're proving zanny's point here. The fact that AMD declared things as legacy has all the bearing on closed drivers and no bearing on OSS drivers. In fact, the HD 4000 series on OSS drivers is pretty much up to its ripening stage at this point. It has almost complete feature parity with Catalyst.

          Comment


          • #45
            Originally posted by GreatEmerald View Post
            You're proving zanny's point here. The fact that AMD declared things as legacy has all the bearing on closed drivers and no bearing on OSS drivers. In fact, the HD 4000 series on OSS drivers is pretty much up to its ripening stage at this point. It has almost complete feature parity with Catalyst.
            My point is the OSS driver is useless for gaming even on 'well supported' chipsets (r600?) and I said nothing that contradicts that.
            zanny's point is the OSS driver is worth buying and AMD graphics for whatever use, I don't know, and that OSS driver guarantees support. So his point is valid only for non-gaming user cases. And I may be too generous here:
            The Catalyst driver remains superior when it comes to OpenGL compliance, OpenGL performance, power management, video acceleration, and numerous other features like AA/AF modes, CrossFire, PowerXpress, and much more.
            What I meant in my previous post was one shouldn't even mention AMD and decency in supporting graphics in one post.

            Comment


            • #46
              Originally posted by Bucic View Post
              The Catalyst driver remains superior when it comes to OpenGL compliance, OpenGL performance, power management, video acceleration, and numerous other features like AA/AF modes, CrossFire, PowerXpress, and much more.

              What I meant in my previous post was one shouldn't even mention AMD and decency in supporting graphics in one post.
              Again, that's false when talking about the HD 4000 series. Mesa has better OpenGL compliance, the performance is on par with Catalyst, the power management is on par with Catalyst, video acceleration is much better than on Catalyst, most of the other features are supported on OSS drivers as well (including PowerXpress).

              Comment


              • #47
                Originally posted by GreatEmerald View Post
                Again, that's false when talking about the HD 4000 series. Mesa has better OpenGL compliance, the performance is on par with Catalyst, the power management is on par with Catalyst, video acceleration is much better than on Catalyst, most of the other features are supported on OSS drivers as well (including PowerXpress).
                'on par' normally means 'eoughly equal', but for OSS fans it usually means 'it's almost 80% as fast in some cases, less than 80% as fast in other cases', so could you back what you're saying with some article on the subject?

                Whenever I'm tempted to compare my hd 3650 with catalyst legacy 13.1 vs OSS driver I recall all those times I tried just to get some mere 30% performance on OSS driver.

                Comment


                • #48
                  Originally posted by Bucic View Post
                  1. The plague of tearing on intel is common across linux desktops. You claim kwin to be superior while every article I've read on the subject state that full-screen repaints is just a workaround making any other affected compositor as real™ as kwin is. For Gnome Shell's Clutter it's CLUTTER_PAINT=disable-clipped-redraws:disable-culling and CLUTTER_VBLANK=True options.

                  While we're at it it's worth highlighting that some genius behind the intel tearing bollocks is responsible for tearing on linux desktops unfixed for over a year! Until Xorg 1.15 and Mesa 10 land in distros of those affected by the issue.

                  2. I failed to get any interesting results on intel's preferred method of repainting.



                  Maybe that guy on reddit is wrong with the explanation(?), but I can only say, I don't see any tearing with kwin and vsync set to full screen repainting (on ivy bridge + x.org 1.14), but with kwin + xrender compositing I see horrible tearing.

                  Comment


                  • #49
                    It's not wise to mention such things waving your pro-AMD banner, you know... Does AMD declaring anything older than radeon 5000 'legacy' ring a bell?
                    That is exactly why I mention it, and is exactly why I only even look at AMD for hardware I can run on open drivers. They demonstrate one of the many issues with proprietary drivers while also being the only discrete gpu option with open source ones (can't really say free, because you need the firmware blobs).

                    Maybe that guy on reddit is wrong with the explanation(?), but I can only say, I don't see any tearing with kwin and vsync set to full screen repainting (on ivy bridge + x.org 1.14), but with kwin + xrender compositing I see horrible tearing.
                    My main rig uses hd4600 under Arch, I have no tearing on 2x screens with automatic vsync, on opengl 3.1 + raster graphics. Hell, I never knew it was a problem, and I built this new pc around 4 months ago, so whatever fixed it has been in mainline Arch for a while.

                    Comment


                    • #50
                      Originally posted by felipe View Post
                      a real gamer play those games when they come out, 10 years ago... metro last ligth and serious sam 3 sucks i don't know who call AAA to metro
                      die...

                      Comment

                      Working...
                      X