No announcement yet.

Linux 4.19 Is Looking Good So Far, At Least On Intel Xeons

  • Filter
  • Time
  • Show
Clear All
new posts

  • Linux 4.19 Is Looking Good So Far, At Least On Intel Xeons

    Phoronix: Linux 4.19 Is Looking Good So Far, At Least On Intel Xeons

    This weekend I decided to fire up the current Linux 4.19 development kernel on the dual Intel Xeon Gold 6138 test platform based on the wonderful Tyan GT24E-B7106. At least for this system, it's really benefiting from the new kernel that will be released as stable in October...

  • #2
    It seems an excellent kernel. Some time ago some developer talked about I/O boost performance up to 400%.


    • #3
      Weird.. 4.19 on both rc3 and amd-staging-drm-next (which is based on 4.19-rc1) are both acting funny on my 144hz displays (regardless of joined or mirrored). The desktop has a weird distortion on the right edge. Changing it to 120hz fixes it. But I'm back on 4.18.7 for now. Not sure what the root cause could be.


      • #4
        I had a lot of issues with Ubuntu 4.19 RC's 1&2,drm-tip(s) dated 9/10 9/07; displays would fall over after suspend especially if suspended a second time. I'd be stuck at 640x480 unknown monitor and the suspended conf (?) file actually rejected as invalid even though it was correct. I even managed to trigger Speculation_Store_Bypass: thread vulnerable according to apport.

        If manually suspending by tapping the power button the 1st attempt would start to power down then spin back up, I needed to tap the power button a second time after allowing the spinup to complete.
        With MBab's 4.18.7 (Ubuntu flavor have yet to test non) the latest round of patches just about everything is fixed that I've noticed. Most were also fixed with drm-tip dated 9/08..I still have to test drm-tip dated 9/11 or RC3. I expect todays drm-tip to mirror MBab's spin,no Idea what to expect out of RC3.

        The last remaining resume from suspend error [drm:dc_link_detect [amdgpu]] *ERROR* No EDID read. From what I can see is caused by systemd/gnomeshell via syslog
        gnome-shell[1191]: Failed to set power save mode for output DP-1: Permission denied
        I'm pretty sure this is what caused all hell to break loose with the 4.19 RC1 RC2 and drm-tip prior to 9/08. All the recent code drops relating to display seems to mitigate the dreadful side effects.

        The other remaining error during boot [drm] dce110_link_encoder_construct: Failed to get encoder_cap_info from VBIOS with error code 4! Doesn't appear in the 4.15.33 kernel and not sure when it started. I started working backwards from 4.19 and stopped at 4.17.19 probably have to install some 4.16 kernels to see when it started.

        Time to check out RC3 I guess. Judging from the amd-staging branch 4.20/5.0 should rock!