Announcement

Collapse
No announcement yet.

Radeon Linux Driver Picks Up Support For Another Vega M GPU

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

  • #11
    Originally posted by Marc Driftmeyer View Post
    Who cares? 99% of these card varieties never arrive or they are 18 months behind schedule. The only aspect AMD is hitting grand slams on is the CPU division. The GPU divisions is a dumpster fire of mismanagement.
    wow a bit harsh! AMD from my perspective is doing an incredible job considering their size. While I’d love to see them do even more in the GPU and frankly the APU arenas I have to balance that against the reality of what they have to work with.

    Comment


    • #12
      Everything AMD Vega is of course interesting, of course a much better match for open source OS and open spec and drivers. Just started to test the XPS 15 w/ Intel Core + AMD Vega graphics. https://www.youtube.com/watch?v=nVxsmhAxBmY guess this mean PRIME hybrid graphic support needs to be improved for more fine grained and automatic witching, ..!?

      Comment


      • #13
        Originally posted by rene View Post
        Everything AMD Vega is of course interesting, of course a much better match for open source OS and open spec and drivers. Just started to test the XPS 15 w/ Intel Core + AMD Vega graphics. https://www.youtube.com/watch?v=nVxsmhAxBmY guess this mean PRIME hybrid graphic support needs to be improved for more fine grained and automatic witching, ..!?
        Hi Rene, I watch your reviews constantly. Thanks for your work!

        You no longer need xrandr setprovideroffloadsink. A simple DRI_PRIME=1 env variable, would suffice.

        Something like:
        Code:
        [mike@mikes-nest ~]$ glxinfo | grep -i 'opengl renderer'
        OpenGL renderer string: Mesa DRI Intel(R) HD Graphics 5500 (Broadwell GT2)
        
        [mike@mikes-nest ~]$ DRI_PRIME=1 glxinfo | grep -i 'opengl renderer'
        OpenGL renderer string: AMD Radeon (TM) R9 M375 (VERDE, DRM 3.27.0, 4.19.9-300.fc29.x86_64, LLVM 8.0.0)

        Comment


        • #14
          Originally posted by xxmitsu View Post

          Hi Rene, I watch your reviews constantly. Thanks for your work!

          You no longer need xrandr setprovideroffloadsink. A simple DRI_PRIME=1 env variable, would suffice.

          Something like:
          Code:
          [mike@mikes-nest ~]$ glxinfo | grep -i 'opengl renderer'
          OpenGL renderer string: Mesa DRI Intel(R) HD Graphics 5500 (Broadwell GT2)
          
          [mike@mikes-nest ~]$ DRI_PRIME=1 glxinfo | grep -i 'opengl renderer'
          OpenGL renderer string: AMD Radeon (TM) R9 M375 (VERDE, DRM 3.27.0, 4.19.9-300.fc29.x86_64, LLVM 8.0.0)
          cool, thanks, will try. Guess the biggest imperfection right now with PRIME is that the user needs to more manually switch it than it probably is quite dynamic and transparent on Windows (not really using Windows now, only guessing), with the MUX based setup is certainly is fully transparent on macOS, ... But as I said in any case working AMD solution is >>> NVidia binary stuff ;-)

          Comment


          • #15
            Originally posted by rene View Post

            cool, thanks, will try. Guess the biggest imperfection right now with PRIME is that the user needs to more manually switch it than it probably is quite dynamic and transparent on Windows (not really using Windows now, only guessing), with the MUX based setup is certainly is fully transparent on macOS, ... But as I said in any case working AMD solution is >>> NVidia binary stuff ;-)
            I understand what you mean...

            By using this env var, the user can use per application settings.
            For example, on fedora there's 'Launch using Dedicated Graphics Card' right click option. If I inspect process env vars, I can see that what does is actually exporting
            Code:
            DISPLAY=:0 DRI_PRIME=1
            among others...


            The aforementioned is for OpenGL.

            For Vulkan, one would have to export a different icd file:
            Code:
            [mike@mikes-nest ~]$ VK_ICD_FILENAMES=/usr/share/vulkan/icd.d/intel_icd.x86_64.json vulkaninfo | grep -i 'gpu id'
            GPU id       : 0 (Intel(R) HD Graphics 5500 (Broadwell GT2))
            [mike@mikes-nest ~]$ VK_ICD_FILENAMES=/usr/share/vulkan/icd.d/radeon_icd.x86_64.json vulkaninfo | grep -i 'gpu id'
            GPU id       : 0 (AMD RADV CAPE VERDE (LLVM 8.0.0))

            Last edited by xxmitsu; 21 December 2018, 03:57 PM.

            Comment


            • #16
              IIRC, you can also specify the GPU in your .drirc file.

              Comment

              Working...
              X