Announcement

Collapse
No announcement yet.

AMD FreeSync Panel Replay Ready For Linux 6.6, Next-Gen GPU Enablement Started

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

  • AMD FreeSync Panel Replay Ready For Linux 6.6, Next-Gen GPU Enablement Started

    Phoronix: AMD FreeSync Panel Replay Ready For Linux 6.6, Next-Gen GPU Enablement Started

    AMD submitted more "new stuff" for their AMDGPU and AMDKFD kernel graphics and compute drivers for the upcoming Linux 6.6 kernel cycle...

    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

  • #2
    DC changes in 6.4 totally broke memory reclocking on Navi21 (memory clock stuck at 96Mhz). Astonishingly, this does only happen with some monitors. This bug has been reported for weeks, but only seems a low priority target as of now, such that it remains broken on 6.4.10…

    Comment


    • #3
      I thought that the AMDKFD driver was going to be merged into the AMDGPU driver. Why did that not happen?

      Comment


      • #4
        Originally posted by kiffmet View Post
        DC changes in 6.4 totally broke memory reclocking on Navi21 (memory clock stuck at 96Mhz). Astonishingly, this does only happen with some monitors. This bug has been reported for weeks, but only seems a low priority target as of now, such that it remains broken on 6.4.10…
        I switched from RX 580 to RX 7900 XTX and the new GPU exhibits the problem at 4k@120/144. Performance was similar or worse than with the RX 580. As a workaround, at first I keept booting 6.3 kernel, then just switched to 60Hz which also lowered idle desktop power from about 40W to 15W. I'm wondering if this FreeSync Panel Refresh is gonna lower the idle power at display's native 4k@120/144.

        Comment


        • #5
          Oh I wish this was possible on desktop! But I imagine you need some kind of special panel.

          Comment


          • #6
            I think it's part of Displayport for full monitors, not the raw LCD panel. Panel Replay won't benefit active LCD power requirements because LCD needs a minimum refresh to maintain the correct biasing.

            Where Panel Replay helps is lowering the power demands of having duplicate frame buffers and removing the unnecessary refreshes over performance long cables like Displayport.
            Last edited by user556; 13 August 2023, 02:10 AM.

            Comment


            • #7
              Originally posted by arboy84 View Post
              I thought that the AMDKFD driver was going to be merged into the AMDGPU driver. Why did that not happen?
              They have been merged for years.

              Comment


              • #8
                The amount of acronyms in this news serves as a test to check how much you are up-to-date with AMD GPU kernel development 😂

                Comment


                • #9
                  Originally posted by arboy84 View Post
                  I thought that the AMDKFD driver was going to be merged into the AMDGPU driver. Why did that not happen?
                  The two drivers (amdkfd and amdgpu) are built into a single module and function as a single driver but kept the code in the same separate sub-folders as before. All of the subfolders (like DC) also build into that single driver module.

                  I don't think there are any plans to combine the subfolders at this point, although it might make sense at some point in the future if we move some of kfd's functionality into drm as a common "single address space across multiple GPUs" implementation.
                  Last edited by bridgman; 16 August 2023, 12:30 AM.
                  Test signature

                  Comment

                  Working...
                  X