Announcement

Collapse
No announcement yet.

Ubuntu 16.04 with AMD R9 285 (Tonga): no video

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

  • #11
    Thanks for advising, debianxfce & Hibbelharry! I've added the Padoka PPA and updated amdgpu et al. from there. I'm still using the default kernel (now at 4.4.0.-36). I was tempted to download the .deb kernel packages to update the kernel to 4.7.2 (http://kernel.ubuntu.com/~kernel-ppa/mainline/v4.7.2/), but I'm not sure... Is there not going to be any conflict between the Padoka's mesa/drivers and a newer kernel?

    Comment


    • #12
      Originally posted by debianxfce View Post
      Use kernel 4.7 and mesa from padoka ppa, because 4.4 is very old kernel. Your dmesg shows that you are using vesa driver. For amdgpu driver, you need to have firmware package and /etc/X11/xorc.conf. Also check that you not are blacklisting amdgpu. When using amdgpu driver you have this with dmesg:
      [drm] Initialized amdgpu 3.2.0 20150101 for 0000:00:01.0 on minor 0
      And this in /var/log/X0rg.0.log
      Sorry for the delay, I finally had more time to investigate this box. Even after installing Linux 4.8 (installed from the Ubuntu kerners) and using padoka to update the mesa stack, the problem persists. Here's a transcription of what happens:

      1) Cold boot starts
      2) Bios screen
      3) 1st password prompt to unlock Grub, which lives on an encrypted partition
      4) Grub screen, so far so good
      5) Normal boot process starts. At the time a password should be prompted for input to unlock the LUKS partition, the screen flashes and the video signal is cut. Screens puts itself in power save mode.
      6) Fans are speeding, but no clue as to what is going on. No SSH access, so the computer didn't finish booting. Blindly attempting to input the LUKS password after the video signal dies doesn't seem to do anything.

      I compared the problematic computer grub's configuration, and it is very much alike the one shipped by Ubuntu 16.04 except for a crypto module which is necessary given that Grub lives on an encrypted partition.

      I'm running out of ideas :/

      Comment


      • #13
        [ 0.000000] DMI: System manufacturer System Product Name/M2N-SLI DELUXE, BIOS ASUS M2N-SLI DELUXE ACPI BIOS Revision 1804 09/16/2010
        You have a motherboard with COM port. Any chances you can debug this issue via serial console? All you need is another system with COM port or USB-to-serial adapter.

        Comment


        • #14
          Originally posted by puleglot View Post
          You have a motherboard with COM port. Any chances you can debug this issue via serial console? All you need is another system with COM port or USB-to-serial adapter.
          Hi pugelot, thanks for your reply. That's a great idea. The system is remote though. Safest way to implement that would probably be by using an SSH enabled 'computer-in-the-middle'. A bit involved given the remote nature of the system. I'm thinking this could be his router (OpenWRT is installable on it), and then I'd only need an Ethernet to Serial adapter to be inserted between the router and the problematic computer.

          Comment


          • #15
            Another option is logging via netconsole. Kernel logs can be captured with netcat.

            Comment


            • #16
              Hardest part is to initialize the NIC card before GFX.

              Comment


              • #17
                Originally posted by puleglot View Post
                Another option is logging via netconsole. Kernel logs can be captured with netcat.

                https://git.kernel.org/cgit/linux/ke...netconsole.txt
                Interesting! This would still require a 'SSH-enabled-device-in-the-middle' to work securely (OpenWRT) but removes the extra hardware dependency (Ethernet-to-Serial adapter), at the cost of a little bit of software configuration. I will look into it!

                Comment


                • #18
                  Almost a year later, this issue is finally understood/resolved. The power supply was apparently not of a good quality enough. The PCIe express connectors were using a split cable (parallel) rather than two distinct cables coming from the PSU. For the reference, that power supply was Raidmax RX-530SS unit which on paper at least looked sufficient. When using this power supply the card would fail its init at boot, not outputting any video signal!

                  After trying it out in my old build which is using a Seasonic S12 power supply (500 W) with two distinct PCIe 6 pin connectors, the card was brought to life beautifully on Debian 9 (after installing linux-firmware from the non-free repo as explained here: https://linuxconfig.org/how-to-insta...-stretch-linux).
                  Last edited by Apteryx; 11 July 2017, 03:03 PM. Reason: Add link to linuxconfig for enabling amdgpu support on Debian 9

                  Comment


                  • #19
                    Arg. I spoke too soon. It did work beautifully on that specific test system, but not on the target system. See my new post here for the continued story: https://www.phoronix.com/forums/foru...u-black-screen

                    Comment

                    Working...
                    X