Announcement

Collapse
No announcement yet.

AMD Unleashes Initial AMDGPU Driver Support For GCN 1.0 / Southern Islands GPUs

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

  • #31
    I've just found the old email order confirmation, it was a 5450

    Comment


    • #32
      Originally posted by tajjada View Post
      The only comparable thing in this range from AMD is the R7 240, which is AMD's lowest-end GCN card, which is a lot more power hungry, not available with passive cooling, and significantly more expensive.
      AFAIK there are a number of passive R7 240's and 250's on the market, starting with XFX:


      Test signature

      Comment


      • #33
        Originally posted by bridgman View Post

        AFAIK there are a number of passive R7 240's and 250's on the market, starting with XFX:

        You should drop the small GPUs and go to 1x:2x:4x style like the old days, with reasonable prices like 100$:200$:400$.

        Comment


        • #34
          It's not ready for users. GFX and DMA are mostly working. DPM (power management) is implemented, but not working yet. UVD and VCE support have not yet been ported. It uses the same ucode as radeon just like CIK." Listed as working is the frame-buffer console, an unaccelerated X.Org Server, and basic OpenGL tests using GBM.
          Heh, with my intel/radeon hybrid gpu laptop I don't need an accelerated ddx, dpm, uvd or vce. Just render nodes for offloading with dri3.

          What's this about "basic" OpenGL tests? Is not everything implemented for full OpenGL 4.3 to be working yet?

          Comment


          • #35
            Originally posted by artivision View Post
            You should drop the small GPUs and go to 1x:2x:4x style like the old days, with reasonable prices like 100$:200$:400$.
            Yep... the market for dGPUs smaller than your typical iGPU is definitely shrinking... almost a "replacement" market these days (old board failed, need new one) rather than an upgrade market.

            The thing that has been keeping small dGPUs alive is the dual-GPU model in laptops -- which IMO is really an alternative to building a "big-ass APU". Personally I would rather see a bigger APU.

            Originally posted by haagch View Post
            What's this about "basic" OpenGL tests? Is not everything implemented for full OpenGL 4.3 to be working yet?
            "Basic" in the sense of "just OpenGL" I guess, ie not working with Glamor, interop etc.. yet.

            AFAIK there is also more testing & bug fixing required on the GL side.
            Last edited by bridgman; 14 May 2016, 11:51 AM.
            Test signature

            Comment


            • #36
              What is the current situation with DAL? As far as I remember the intention was to add it to 4.7? Will it be merged or still a lot to do to mainline it? Can somebody enlighten what is coming and when?

              Comment


              • #37
                Ah, now it's released: https://cgit.freedesktop.org/~agd5f/...ext-4.8-wip-si
                Interestingly: Last commit is drm/amdgpu: enable SI DPM so.. this is still the not working dpm support?
                The config option is config DRM_AMDGPU_SI.

                Now "Mesa and libdrm patches will be coming too". Are they not released yet?

                Comment


                • #38
                  From earlier discussions I thought that mesa and libdrm patches were also required, but Alex said that just the kernel code is needed for the functionality in the cover letter. There is other work going on but nothing tested yet.
                  Test signature

                  Comment


                  • #39
                    Hm, runpm doesn't work yet but as they say, it's early.
                    I first couldn't start X on my intel GPU, it was hanging on adding the radeon gpu I think:
                    Code:
                    [FONT=monospace][COLOR=#ff5454][B]INFO: task Xorg:724 blocked for more than 120 seconds.[/B][/COLOR]
                    [COLOR=#ff5454][B]  Tainted: G           O    4.6.0-rc7-mainline #13[/B][/COLOR]
                    [COLOR=#ff5454][B]"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.[/B][/COLOR]
                    Xorg            D ffff8807f823fd20     0   724      1 0x00000004
                    [COLOR=#000000][B]ffff8807f823fd20 0000000000008002 ffff88080b70d580 ffff880809e88e40[/B][/COLOR]
                    [COLOR=#000000][B]ffff8807f8240000 ffff880801f7da40 ffff880801f7d9e8 ffff88080867f088[/B][/COLOR]
                    [COLOR=#000000][B]ffff88080a1aa9c0 ffff8807f823fd38 ffffffff8162c60f ffff88080ac770e8[/B][/COLOR]
                    [COLOR=#000000][B]Call Trace:[/B][/COLOR]
                    [COLOR=#000000][B][<ffffffff8162c60f>] schedule+0x4f/0xa0[/B][/COLOR]
                    [COLOR=#000000][B][<ffffffffa0a5dfb7>] amd_sched_entity_fini+0x77/0x110 [amdgpu][/B][/COLOR]
                    [COLOR=#000000][B][<ffffffff810cd340>] ? wake_atomic_t_function+0x70/0x70[/B][/COLOR]
                    [COLOR=#000000][B][<ffffffffa0a05daf>] amdgpu_vm_fini+0x3f/0x210 [amdgpu][/B][/COLOR]
                    [COLOR=#000000][B][<ffffffffa0a08e85>] ? amdgpu_ctx_mgr_fini+0x75/0xb0 [amdgpu][/B][/COLOR]
                    [COLOR=#000000][B][<ffffffffa09dff20>] amdgpu_driver_postclose_kms+0x60/0xc0 [amdgpu][/B][/COLOR]
                    [COLOR=#000000][B][<ffffffffa00046ec>] drm_release+0x29c/0x4e0 [drm][/B][/COLOR]
                    [COLOR=#000000][B][<ffffffff8121fabe>] __fput+0xae/0x1f0[/B][/COLOR]
                    [COLOR=#000000][B][<ffffffff8121fc61>] ____fput+0x21/0x30[/B][/COLOR]
                    [COLOR=#000000][B][<ffffffff810a4a3b>] task_work_run+0x8b/0xb0[/B][/COLOR]
                    [COLOR=#000000][B][<ffffffff8100371d>] exit_to_usermode_loop+0xcd/0xd0[/B][/COLOR]
                    [COLOR=#000000][B][<ffffffff81003ca1>] syscall_return_slowpath+0x61/0x70[/B][/COLOR]
                    [COLOR=#000000][B][<ffffffff816308ba>] entry_SYSCALL_64_fastpath+0xa2/0xa4[/B][/COLOR][/FONT]
                    But after removing xf86-video-ati (forgot before rebooting) and xf86-video-intel, it starts fine.

                    As expected, without mesa/libdrm e.g. "DRI_PRIME=1 glxinfo" doesn't work. The message is funny though:
                    Code:
                    [FONT=monospace][COLOR=#000000]amdgpu: amdgpu_device_initialize failed. [/COLOR]
                    do_winsys_init: DRM version is 3.2.0 but this driver is only compatible with 2.12.0 (kernel 3.2) or later.[/FONT]
                    Should probably revise the "or later" part a bit.

                    I guess that's what's meant with "basic OpenGL".
                    Makes me hopeful that it will already just work with DRI_PRIME though.

                    EDIT: Also I can only try "DRI_PRIME=1 glxinfo" once without problems. The second time the command hangs in the terminal and if I then try to run it a third time while the other one hangs, X/the machine locks up with nothing in the logs.

                    Sensors work. Yay:
                    Code:
                    [FONT=monospace][COLOR=#000000]amdgpu-pci-0100 [/COLOR]
                    Adapter: PCI adapter
                    temp1:        +41.0°C  (crit = +120.0°C, hyst = +90.0°C)[/FONT]
                    rmmod'ing amdgpu doesn't work. Hangs at
                    Code:
                    [FONT=monospace][COLOR=#000000][drm] amdgpu: finishing device.[/COLOR][/FONT]
                    but that was something where radeon also had trouble a long time until at least didn't crash the machine anymore...

                    Tl;dr: It was the truth: It is an early preview for developers.
                    Last edited by haagch; 14 May 2016, 07:52 PM.

                    Comment


                    • #40
                      Originally posted by bridgman View Post
                      Yep... the market for dGPUs smaller than your typical iGPU is definitely shrinking... almost a "replacement" market these days (old board failed, need new one) rather than an upgrade market.
                      I guess appearance of hi-res LCDs going to help it a bit. To drive e.g. 4K @ native res one needs powerful dGPU, APU wouldn't do the trick. Looking on Michael doing 4K benchmarks it seems it no longer something exotic.

                      Personally I would rather see a bigger APU.
                      I wonder if putting HBM on APU would teach intel how to do their weak yet expensive Iris stuff right.

                      Comment

                      Working...
                      X