Announcement

Collapse
No announcement yet.

Radeon ROCm 3.0 Released With LLVM "AOMP" For Radeon OpenMP, FFT Updates

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

  • Radeon ROCm 3.0 Released With LLVM "AOMP" For Radeon OpenMP, FFT Updates

    Phoronix: Radeon ROCm 3.0 Released With LLVM "AOMP" For Radeon OpenMP, FFT Updates

    Announced last month at SuperComputing 19 in Denver was Radeon Open Compute 3.0 (ROCm 3.0) but it didn't end up shipping until last night. ROCm 3.0 is a big update to AMD's open-source Linux compute stack for ending out 2019...

    http://www.phoronix.com/scan.php?pag...m-3.0-Released

  • #2
    Funny I finally put the GTX 1080 Ti back into my Linux rig because the RX 5700 XT has no OpenCL when using it with Mesa (because KMS and stuff).
    Now Blender 2.81 with Cuda on Linux render https://download.blender.org/demo/te...27_2.blend.zip in 59 seconds.
    And yes Windows 10 Adrenalin Drivers does OpenCL when using the Red Devil RX 5700 XT and renders the same scene in silent mode in 1 minute and 30 seconds.

    So I have fixed some flaws while using Navi.

    But NVidia KMS is nearly not existent no matter if I load modules early on or compile custom kernel with dkms. And yes I set kernel parameter nvidia-drm.modeset=1

    That mean I can not switch between X11 sessions anymore while graphic pograms are running because NVidia will crash my applications when I begin to switch between this sessions.

    I already thought maybe I can use my old RX 480 to do X11 stuff while using the GTX 1080 Ti for Cuda only. Any hints?
    Last edited by Naquatis; 12-21-2019, 08:37 PM.

    Comment


    • #3
      Why not use the closed source opencl driver from the amdgpu-pro package?

      ROCm main target is AMDs professional card lineup, and Navi is only available as consumer cards atm, so it's in no hurry to support it until such cards emerge.
      Last edited by ObiWan; 12-21-2019, 03:48 PM.

      Comment


      • #4
        Instead of reinventing this stuff every few months I wish AMD would put more effort in making what they have usable for more people. It doesn't matter what features they add if no one can use it. It's nice that they have released this with support for end of life RH7 but when they fork RH it is with stuff that is 5 years behind the curve, 7 has been around for going on 10 years which makes it 15 years behind the curve. There are A LOT of improvement to the desktops, the graphics stacks etc that are very useful in 8 that aren't in 7. For example I don't have to log on and manually run xrandr to scale the desktop so I can read the fonts on my 4k display.. We are a week away from 2020 isn't that some thing that should just work or is that just to scifi?

        Comment


        • #5
          Originally posted by ObiWan View Post
          Why not use the closed source opencl driver from the amdgpu-pro package?

          ROCm main target is AMDs professional card lineup, and Navi is only available as consumer cards atm, so it's in no hurry to support it until such cards emerge.
          If I need to downgrade the kernel because the amdgpu-pro package does not work on latest kernel versions then I stay with NVidia.
          What works nicely with RX 480 was installing the userspace opencl-amd package while using Mesa .. but that does not work for Navi.

          AMDs professional card lineup .. pff .. look how this RX 5700 XT -> 7nm card is not even able to get close to an old NVidia GTX 1080 Ti when it comes to GPU calculations (.. and I did not add the openCL kernel compilation then it is 1 minute 45 seconds against 59 seconds).
          Last edited by Naquatis; 12-21-2019, 04:52 PM.

          Comment


          • #6
            Here on Arch the closed source opencl works fine with Navi and the current kernel

            Code:
            pacman -Q | grep opencl
            opencl-amdgpu-pro-comgr 19.30_934563-1
            opencl-amdgpu-pro-dev 19.30_934563-1
            opencl-amdgpu-pro-orca 19.30_934563-1
            opencl-amdgpu-pro-pal 19.30_934563-1
            Code:
            clinfo -l    
            Platform #0: AMD Accelerated Parallel Processing
             `-- Device #0: gfx1010
            Navis compute performance is lower then that of Vega, compute performance wasn't a focus for Navi, 3D rendering was.

            But I won't argue against NVidia getting way more performance out of their cards, then AMD, with their drivers.

            PS:
            Just did a render of BMW27 on my 5700XT with the closed source OpenCL, and it finished in 1:26.43
            Last edited by ObiWan; 12-22-2019, 07:14 AM. Reason: clarifying the post a bit

            Comment


            • #7
              Originally posted by ObiWan View Post
              Here on Arch it works fine with Navi and the current kernel

              Code:
              pacman -Q | grep opencl
              opencl-amdgpu-pro-comgr 19.30_934563-1
              opencl-amdgpu-pro-dev 19.30_934563-1
              opencl-amdgpu-pro-orca 19.30_934563-1
              opencl-amdgpu-pro-pal 19.30_934563-1
              Code:
              clinfo -l
              Platform #0: AMD Accelerated Parallel Processing
              `-- Device #0: gfx1010
              Navis compute performance is lower then that of Vega, compute performance wasn't a focus for Navi, 3D rendering was.

              But I won't argue against NVidia getting way more performance out of their cards, then AMD, with their drivers.

              PS:
              Just did a render of BMW27 on my 5700XT with the closed source OpenCL, and it finished in 1:26.43
              Does it mean that there is ROCm support for navi?

              Comment


              • #8
                Originally posted by Naquatis View Post

                If I need to downgrade the kernel because the amdgpu-pro package does not work on latest kernel versions then I stay with NVidia.
                What works nicely with RX 480 was installing the userspace opencl-amd package while using Mesa .. but that does not work for Navi.

                AMDs professional card lineup .. pff .. look how this RX 5700 XT -> 7nm card is not even able to get close to an old NVidia GTX 1080 Ti when it comes to GPU calculations (.. and I did not add the openCL kernel compilation then it is 1 minute 45 seconds against 59 seconds).
                You should never have to downgrade the kernel for AMD since there is no closed source kernel driver for AMD both the open and closed drivers use the same open in kernel driver on AMD hardware. Sometimes you might have to run an out of tree driver for recent stuff that hasn't made it into mainline yet.

                Comment


                • #9
                  Originally posted by caaperezag View Post

                  Does it mean that there is ROCm support for navi?
                  It's not ROCm, it the closed source OpenCL driver.

                  Comment


                  • #10
                    Originally posted by ObiWan View Post
                    Here on Arch the closed source opencl works fine with Navi and the current kernel

                    Code:
                    pacman -Q | grep opencl
                    opencl-amdgpu-pro-comgr 19.30_934563-1
                    opencl-amdgpu-pro-dev 19.30_934563-1
                    opencl-amdgpu-pro-orca 19.30_934563-1
                    opencl-amdgpu-pro-pal 19.30_934563-1
                    Code:
                    clinfo -l
                    Platform #0: AMD Accelerated Parallel Processing
                    `-- Device #0: gfx1010
                    Navis compute performance is lower then that of Vega, compute performance wasn't a focus for Navi, 3D rendering was.

                    But I won't argue against NVidia getting way more performance out of their cards, then AMD, with their drivers.

                    PS:
                    Just did a render of BMW27 on my 5700XT with the closed source OpenCL, and it finished in 1:26.43
                    Thank you! That helped a lot! Now the GTX 1080 Ti went back into the Windows rig.

                    Comment

                    Working...
                    X