Announcement

Collapse
No announcement yet.

Radeon ROCm 3.5 Released With New Features But Still No Navi Support

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

  • Radeon ROCm 3.5 Released With New Features But Still No Navi Support

    Phoronix: Radeon ROCm 3.5 Released With New Features But Still No Navi Support

    Radeon Open Compute 3.5 (ROCm 3.5) is now available with a number of improvements but surprisingly still no GFX10/Navi support...

    http://www.phoronix.com/scan.php?pag...m-3.5-Released

  • #2
    Would be nice to see opencl upstreamed into Mesa gallium for radeon cards. It just would be a more useful out of the box experience in the medium to long run.

    Comment


    • #3
      So Navi is abandoned for compute - I hope that changes with Big Navi/Sienna Cichlid hardware.

      Comment


      • #4
        Originally posted by dbont View Post
        So Navi is abandoned for compute - I hope that changes with Big Navi/Sienna Cichlid hardware.
        If RDNA1 is an indicator how things will be handled for RDNA and CDNA going forward, things don't look good for Linux consumers who want to employ ROCm.

        Comment


        • #5
          This is getting ridiculous...
          ## VGA ##
          AMD: X1950XTX, HD3870, HD5870
          Intel: GMA45, HD3000 (Core i5 2500K)

          Comment


          • #6
            Originally posted by darkbasic View Post
            This is getting ridiculous...
            Yes, it absolutely is... run Nvidia if you have anything related to GPU Compute! Its not one click setup, but it works and thats probably the key feature

            Comment


            • #7
              Originally posted by ms178 View Post

              If RDNA1 is an indicator how things will be handled for RDNA and CDNA going forward, things don't look good for Linux consumers who want to employ ROCm.
              It is not like ROCm have been targeted for consumer market at all. It so much P.I.T.A. to install it on any recent distro that most people who need compute just use nvidia instead.
              And that on top of very narrow hardware support list. Will probably see usable Mesa implementation of OpenCL sooner than any good changes in ROCm.

              Comment


              • #8
                They need to focus on their awarded contracts first. ROCm priority is to get it in shape for the first exascale deployment next year.
                Not coincidentally the GPU architecture for that deployment is not Navi based.
                You can also find many articles of several Labs getting their code ready for HIP as part of a "SW readiness program'.
                Additionally, part of El Capitan initial payment involves ROCm development.

                I doub't they will focus on anything else till after 4.0 which is when feature complete for 'Ultrascale Computing' will be achieved.

                Comment


                • #9
                  They also advertise OpenCL 2.2 but I cannot see that. clinfo still reports 2.0 and 1.2 for the OpenCL C and device respectively.

                  Comment


                  • #10
                    Originally posted by blacknova View Post

                    It is not like ROCm have been targeted for consumer market at all. It so much P.I.T.A. to install it on any recent distro that most people who need compute just use nvidia instead.
                    It is? I use upstream kernel and the tensorflow-rocm docker image on Arch Linux. Setup time : 30 seconds.

                    Comment

                    Working...
                    X