Announcement

Collapse
No announcement yet.

Linux 4.19.7 Getting Revised STIBP Code, Important PCIe Fix For Radeon GPUs

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

  • Linux 4.19.7 Getting Revised STIBP Code, Important PCIe Fix For Radeon GPUs

    Phoronix: Linux 4.19.7 Getting Revised STIBP Code, Important PCIe Fix For Radeon GPUs

    Greg Kroah-Hartman today published the initial patch series of work he intends to use as forming the Linux 4.19.7 point release in the coming days. With Linux 4.19.7 there are some important fixes...

    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
    Any news on EXT4 corruption?

    Comment


    • #3
      Seriously, it's right in the article...

      Comment


      • #4
        I'm really curious who would be running a modern GPU on a system old enough to have PCIe 1.0. Seems like a pretty odd mix, especially since it was only around for 2 years (I assume this problem doesn't affect PCIe 1.1?)

        Comment


        • #5
          Originally posted by jpg44 View Post
          Any news on EXT4 corruption?
          "Notably not included is any fix for the mysterious EXT4 file-system corruption on 4.19. That issue is still being investigated and looking like it could be a problem in the kernel outside of the EXT4 code itself but still not entirely clear."

          Comment


          • #6
            Originally posted by schmidtbag View Post
            I assume this problem doesn't affect PCIe 1.1?
            It even affects PCIe 2.

            //EDIT: To be more precise: The problem was that the driver incorrectly detected every PCIe generation as being 3.0.
            Last edited by V10lator; 04 December 2018, 11:16 AM.

            Comment


            • #7
              I wonder if ALC300 is supported with the same quality as ALC1220, where the mic is still unusable on pulseaudio and noisy on ALSA.

              Comment


              • #8
                Originally posted by schmidtbag View Post
                I'm really curious who would be running a modern GPU on a system old enough to have PCIe 1.0. Seems like a pretty odd mix, especially since it was only around for 2 years (I assume this problem doesn't affect PCIe 1.1?)
                For very specific workloads that are purely GPU-bound and CPU-agnostic.
                Probably GPGPU.

                - There are threads about people using modern PCIe3.0 Radeons (on older AMD FX-8xxx processors) to run the OpenCL Cycles renderer in Blender.
                - Runing neural nets in OpenCL would be another possibility.
                (And I'm sure the crypto currency crowd has a couple of example of stuff that runs perfectly on GPU. I'm just too lazy to check which ones are currently GPU-mined)

                Comment


                • #9
                  Originally posted by schmidtbag View Post
                  I'm really curious who would be running a modern GPU on a system old enough to have PCIe 1.0. Seems like a pretty odd mix, especially since it was only around for 2 years (I assume this problem doesn't affect PCIe 1.1?)
                  I have this setup at home, an Rx 460 card in a Supermicro H8SMI-2 motherboard. It's my old workstation, re-purposed into a low budget casual gaming rig. Its Opteron 1389 and 8 GB of DDR2-800 ECC is plenty for casual gaming, and the Rx 460 gives great frame rates on the old 19" 1280x1024 monitor.

                  Edit: In case you wonder why bother gaming on old hardware, this rig gives ~50 fps in Rise of the Tomb Raider and Far Cry 4 on Ultra settings - not too shabby for a machine that otherwise would be collecting dust in the basement.
                  Last edited by torsionbar28; 04 December 2018, 01:30 PM.

                  Comment


                  • #10
                    They seem to be making solid progress with a setup that reproduces in a couple minutes, someone on the ext4 bug report has bisected to blk-mq

                    Comment

                    Working...
                    X