Announcement

Collapse
No announcement yet.

AMD Catalyst vs. NVIDIA OpenCL Performance

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

  • #31
    Originally posted by mmstick View Post
    I'm not painting anything other than the truth. I use Linux as a desktop and server OS. HL2 and Portal isn't very taxing; try playing TF2 and watch Radeon HD 7000 GPUs commit suicide. Serious Sam 3 doesn't work 'well' at all; it's many magnitudes slower than the drivers on Windows. Indie games don't need the full power of a Radeon HD 7950 nor do they use a lot of the complicated features that break in the drivers. Steam itself crashes quite a lot due to the fglrx drivers.
    I'm also only reporting truth.
    Then our definition of 'well' is different. SS3 runs fast enough for me, and looks pretty good enough too. that is what I mean as 'well'. I never ran it in Windows, so have nothing to compare to.
    I have also have never had Steam crash on me, then again I don't clock in hundreds of hours gaming a week.

    Comment


    • #32
      Originally posted by johnc View Post
      I got a hearty laugh.
      I facepalmed myself when I read today what I wrote, but I was drunk :/

      Comment


      • #33
        Article still not updated to mention whether this is double precision or single precision. Good job, Phoronix. This really makes you look like you have no clue what you're writing about.

        Comment


        • #34
          Curiously, is AMD still committed to the same open-source mandate from 2006 after acquiring ATI? Has the development of truly open sourced drivers slowed down, as would seem? Why? Legal issues are what I'm hearing, but it would appear ailing bugdet concern's with an all round cost-cutting across the board would seem more likely after they failed to capitalise after AMD64 are what I'm seeing.

          Either, either, or. Maybe some insight in to how things are faring at AMD 'generally' would be nice (links, quotes, personal work flows unbound by NDA's, whatever). I like the gear enough to keep the faith and Intel honest (doing a pretty damn good job of late, obviously). I keep buying AMD knowing quite often it's not the wisest idea at the time, and I've stop this practice just to accomplish even the most simplest of tasks (such as home computing and recycling older hardware with a newer adaptor, but with nVidia's, not AMD's. A mostly useful yet unsupported and underperformant VGA with a screaming fan is not 'cool', so to speak. And then there's the onboard stuff =( )

          Again, just curious how things are looking at AMD, and what can be expected.
          Hi

          Comment


          • #35
            Originally posted by stiiixy View Post
            Curiously, is AMD still committed to the same open-source mandate from 2006 after acquiring ATI? Has the development of truly open sourced drivers slowed down, as would seem? Why? Legal issues are what I'm hearing, but it would appear ailing bugdet concern's with an all round cost-cutting across the board would seem more likely after they failed to capitalise after AMD64 are what I'm seeing.
            What on earth are you talking about. They've just released UVD and dynamic power management, which are by far the most difficult things to get out, technically and legally. There was a MASSIVE code drop just days ago.

            OpenCL is taking a while, but this is not only AMD's business. Intel doesn't have complete OpenCL either, as the whole open stack (Gallium, LLVM, etc.) is not ready yet. But they're making progress nonetheless.

            Comment


            • #36
              Originally posted by pingufunkybeat View Post
              What on earth are you talking about. They've just released UVD and dynamic power management, which are by far the most difficult things to get out, technically and legally. There was a MASSIVE code drop just days ago.

              OpenCL is taking a while, but this is not only AMD's business. Intel doesn't have complete OpenCL either, as the whole open stack (Gallium, LLVM, etc.) is not ready yet. But they're making progress nonetheless.
              Umm, well, I was talking about exactly what you just replied to in part and in a vague sense, so yeah, thank you, in part ina vague manner. I guess.

              All I see regarding AMD these days are news snippets and the bemoaning of AMD's current driver issues, open or otherwise. And I was merely after how AMD had progressed after their promose made 7 years ago regarding opening their graphics up. If you don't know, or understand, try asking instead of gettin' on your high-horse. I wasn't for or against AMD. I'm after the track record =)
              Hi

              Comment


              • #37
                Originally posted by stiiixy View Post
                Umm, well, I was talking about exactly what you just replied to in part and in a vague sense, so yeah, thank you, in part ina vague manner. I guess.

                All I see regarding AMD these days are news snippets and the bemoaning of AMD's current driver issues, open or otherwise. And I was merely after how AMD had progressed after their promose made 7 years ago regarding opening their graphics up. If you don't know, or understand, try asking instead of gettin' on your high-horse. I wasn't for or against AMD. I'm after the track record =)
                I've been on AMD's OSS drivers ever since r700 first started rendering a triangle. I'd say that I know and understand enough

                It just seems that some people are hell-bent on not acknowledging any work. Now we have GL 3.1, 80% or more of Catalyst performance, Dynamic PM, UVD (!!!), works out of the box on all modern distros, and people are still bitching about something.

                The other day, I upgraded Debian from Squeeze to Wheezy at work. With Nvidia binary driver. Fucking hell, I keep forgetting how awful that is.

                Comment


                • #38
                  Originally posted by stiiixy View Post
                  Umm, well, I was talking about exactly what you just replied to in part and in a vague sense, so yeah, thank you, in part ina vague manner. I guess.

                  All I see regarding AMD these days are news snippets and the bemoaning of AMD's current driver issues, open or otherwise. And I was merely after how AMD had progressed after their promose made 7 years ago regarding opening their graphics up. If you don't know, or understand, try asking instead of gettin' on your high-horse. I wasn't for or against AMD. I'm after the track record =)
                  The fact is that the track record shows AMD OSS drivers have improved tremendously in the last half year or so. The radeon OSS driver provides better support, stability, desktop performance, etc than the catalyst driver across the board.

                  Comment


                  • #39
                    Originally posted by Marc Driftmeyer View Post
                    Either you're laughing at him or you don't grasp that CUDA is dying.
                    Google Trends does not agree with you:



                    It seems that you have not used CUDA nor OpenCL. OpenCL is ridiculously verbose and the interface is crappy. cl_mem? are you serious? managing contexts, compiling/loading code in runtime? have you looked at the syntax to launch kernels? CUDA is 3-4 years ahead in terms of features and it targets application developers. OpenCL is OK for some libraries (e.g. video decoders) but it is not meant to be directly used in applications. Even AMD is now pushing for HSA, that supports other languages like C++AMP.

                    Comment


                    • #40
                      Keep dreaming

                      Originally posted by Marc Driftmeyer View Post
                      Either you're laughing at him or you don't grasp that CUDA is dying.
                      Google Trends does not agree with you:



                      By the way, have you ever written a single line of CUDA or OpenCL? OpenCL is not targeted to application developers. In High-Performance Computing everybody uses CUDA. It is 3-4 years ahead in terms of features (C++, dynamic parallelism, pointers!).

                      Comment

                      Working...
                      X