Help needed getting 7900 XTX working under linux

[stingray454]

I finally pulled the plug and got some nice upgrades for my Linux machine. I have, however, had issues getting the 7900 XTX working. What I've done so far:

- Installed Mesa 22.3.1 (and additional packages, like vulkan_radeon etc)
- Installed linux-firmware-git. Also tried manually downloading / copying the latest amdgpu firmware from the repo, just in case (no difference)
- Rebuilt initramfs

The GPU is recognised as a "VGA Compatible Adapter", so I do get an image / desktop, but it is ofc horribly slow as there is no gpu acceleration.

Can't say that I'm very good ad figuring out these things, so I would appreciate some help of what's needed. This is the amdgpu-related stuff from dmesg:

[    4.263635] [drm] amdgpu kernel modesetting enabled.
[    4.263703] amdgpu: Ignoring ACPI CRAT on non-APU system
[    4.263706] amdgpu: Virtual CRAT table created for CPU
[    4.263712] amdgpu: Topology: Add CPU node
[    4.263858] amdgpu 0000:03:00.0: vgaarb: deactivate vga console
[    4.263896] amdgpu 0000:03:00.0: enabling device (0006 -> 0007)
[    4.265811] amdgpu 0000:03:00.0: amdgpu: Fetched VBIOS from VFCT
[    4.265812] amdgpu: ATOM BIOS: 113-D7020100-102
[    4.265819] amdgpu 0000:03:00.0: [drm:jpeg_v4_0_early_init [amdgpu]] JPEG decode is enabled in VM mode
[    4.266009] amdgpu 0000:03:00.0: amdgpu: Trusted Memory Zone (TMZ) feature not supported
[    4.266034] amdgpu 0000:03:00.0: amdgpu: MEM ECC is not presented.
[    4.266035] amdgpu 0000:03:00.0: amdgpu: SRAM ECC is not presented.
[    4.266072] amdgpu 0000:03:00.0: amdgpu: VRAM: 24560M 0x0000008000000000 - 0x00000085FEFFFFFF (24560M used)
[    4.266074] amdgpu 0000:03:00.0: amdgpu: GART: 512M 0x0000000000000000 - 0x000000001FFFFFFF
[    4.266112] [drm] amdgpu: 24560M of VRAM memory ready
[    4.266113] [drm] amdgpu: 15911M of GTT memory ready.
[    4.268850] amdgpu 0000:03:00.0: amdgpu: CP RS64 enable
[    4.270073] amdgpu 0000:03:00.0: amdgpu: Will use PSP to load VCN firmware
[    4.552329] amdgpu 0000:03:00.0: amdgpu: RAP: optional rap ta ucode is not available
[    4.552332] amdgpu 0000:03:00.0: amdgpu: SECUREDISPLAY: securedisplay ta ucode is not available
[    4.552360] amdgpu 0000:03:00.0: amdgpu: smu driver if version = 0x00000032, smu fw if version = 0x00000034, smu fw program = 0, smu fw version = 0x004e4700 (78.71.0)
[    4.552363] amdgpu 0000:03:00.0: amdgpu: SMU driver if version not matched
[    4.718330] amdgpu 0000:03:00.0: amdgpu: SMU is initialized successfully!
[    4.746659] snd_hda_intel 0000:03:00.1: bound 0000:03:00.0 (ops amdgpu_dm_audio_component_bind_ops [amdgpu])
[    5.018406] amdgpu 0000:03:00.0: [drm:jpeg_v4_0_hw_init [amdgpu]] JPEG decode initialized successfully.
[    5.019769] kfd kfd: amdgpu: Allocated 3969056 bytes on gart
[    5.019870] amdgpu: sdma_bitmap: fff0
[    5.081058] amdgpu: HMM registered 24560MB device memory
[    5.081423] amdgpu: SRAT table not found
[    5.081423] amdgpu: Virtual CRAT table created for GPU
[    5.081976] amdgpu: Topology: Add dGPU node [0x744c:0x1002]
[    5.081980] kfd kfd: amdgpu: added device 1002:744c
[    5.081991] amdgpu 0000:03:00.0: amdgpu: SE 6, SH per SE 2, CU per SH 8, active_cu_number 96
[    5.082044] amdgpu 0000:03:00.0: amdgpu: ring gfx_0.0.0 uses VM inv eng 0 on hub 0
[    5.082045] amdgpu 0000:03:00.0: amdgpu: ring comp_1.0.0 uses VM inv eng 1 on hub 0
[    5.082045] amdgpu 0000:03:00.0: amdgpu: ring comp_1.1.0 uses VM inv eng 4 on hub 0
[    5.082046] amdgpu 0000:03:00.0: amdgpu: ring comp_1.2.0 uses VM inv eng 6 on hub 0
[    5.082046] amdgpu 0000:03:00.0: amdgpu: ring comp_1.3.0 uses VM inv eng 7 on hub 0
[    5.082046] amdgpu 0000:03:00.0: amdgpu: ring comp_1.0.1 uses VM inv eng 8 on hub 0
[    5.082047] amdgpu 0000:03:00.0: amdgpu: ring comp_1.1.1 uses VM inv eng 9 on hub 0
[    5.082047] amdgpu 0000:03:00.0: amdgpu: ring comp_1.2.1 uses VM inv eng 10 on hub 0
[    5.082048] amdgpu 0000:03:00.0: amdgpu: ring comp_1.3.1 uses VM inv eng 11 on hub 0
[    5.082048] amdgpu 0000:03:00.0: amdgpu: ring sdma0 uses VM inv eng 12 on hub 0
[    5.082049] amdgpu 0000:03:00.0: amdgpu: ring sdma1 uses VM inv eng 13 on hub 0
[    5.082049] amdgpu 0000:03:00.0: amdgpu: ring vcn_unified_0 uses VM inv eng 0 on hub 1
[    5.082050] amdgpu 0000:03:00.0: amdgpu: ring vcn_unified_1 uses VM inv eng 1 on hub 1
[    5.082050] amdgpu 0000:03:00.0: amdgpu: ring jpeg_dec uses VM inv eng 4 on hub 1
[    5.082051] amdgpu 0000:03:00.0: amdgpu: ring mes_kiq_3.1.0 uses VM inv eng 14 on hub 0
[    5.088179] amdgpu 0000:03:00.0: amdgpu: Using BACO for runtime pm
[    5.088380] [drm] Initialized amdgpu 3.49.0 20150101 for 0000:03:00.0 on minor 0
[    5.093439] fbcon: amdgpudrmfb (fb0) is primary device
[    5.287724] amdgpu 0000:03:00.0: [drm] fb0: amdgpudrmfb frame buffer device
​

Also attaching the relevant parts of lshw / lspci. Anyone have an idea why it's not recognised? Any pointers appreciated.

[stingray454]

Found the cause! In short

• For me, mesa-git (23.0 deveopment version) was required (22.3 did not work)
• mesa-git / lib32-mesa-git has to be compiled with llvm15+. The AUR package has a dependency for llvm14, so you need to clone the package and override the llvm version to build them
• After building / installing, I have working 3d acceleration / vulkan

[tyranids]

Found the cause! In short

• For me, mesa-git (23.0 deveopment version) was required (22.3 did not work)
• mesa-git / lib32-mesa-git has to be compiled with llvm15+. The AUR package has a dependency for llvm14, so you need to clone the package and override the llvm version to build them
• After building / installing, I have working 3d acceleration / vulkan

How do you change the compiler version to be used with an AUR package? I also tried and failed to get a 7900XT working under linux with kernels 6.0.9, 6.1-rc5, Mesa 22.3, and Mesa 23.0. I would always get back to having a desktop at the correct resolution, with terrible performance, and no mouse cursor. Actually the terrible performance was only in cases where the GPU would also not be recognized correctly, just AMD then some numbers.

I ended up giving up and installing Windows 11, but would really like to go back to a linux desktop. My current thought was in a month or so the Arch based distros should ship with 6.0+ kernel and Mesa new enough to work properly. Also I think the required firmware was pretty much never included by default, so I was grabbing it from the firmware git https://git.kernel.org/pub/scm/linux...x-firmware.git

[fakemoth]

Anyone doing some funky stuff with good results on Debian 11 to get the 7900 series working?

[DarkFoss]

It just works on Tumbleweed after swapping out my 6600xt for an Asrock Tachi RX 7900XTX. It was a longer process on Win 11 pro.