[Bug (solved)] mention of AMD linux/driver defaults

[Sx1Fl0NdqlkM6N4s2pswXKUI0k9FKTaOckLbNXu]

Git commit

It's just a documentation of encountered driver defaults failing at certain tasks.

OS:

debian testing

Hardware:

ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = AMD Radeon Vega 11 Graphics (RADV RAVEN) (radv) | uma: 1 | fp16: 1 | bf16: 0 | fp4: 0 | warp size: 64 | shared memory: 65536 | int dot: 0 | matrix cores: none
load_backend: loaded Vulkan backend from /home/guest/mnt/1/@ai/stable-diffusion/libggml-vulkan.so
load_backend: loaded CPU backend from /home/guest/mnt/1/@ai/stable-diffusion/libggml-cpu-haswell.so
Vulkan0 AMD Radeon Vega 11 Graphics (RADV RAVEN)
CPU AMD Ryzen 5 PRO 2400G with Radeon Vega Graphics

Issue overview:

• Tried using "SD 1.5", and "Flux 2" which worked completly fine,
  but while using "sefi" at 1024x1024, it would crash with: "device lost on Vulkan0", even tho it still produced a scrabled image at 512x512 (non-native resolution).
• what didn't work: GGML_VK_MAX_NODES_PER_SUBMIT=1, --max-vram 1, --disable-prefetch, --type f16, --diffusion-fa

Logs:

sdcpp-verbose.txt
journalctl.txt
modinfo_amdgpu.txt

Solution:

Default debian values of "amdgpu.lockup_timeout" are too low, so had to add custom grub/kernel paremeters of:
amdgpu.lockup_timeout=10000,60000,10000,10000
to resolve the issue.

• also same issue encountered elsewhere: Misc. bug: Vulkan: DeviceLost on AMD APUs (gfx90c) due to GPU job timeout from command batch size ggml-org/llama.cpp#21724

Operating System & Version

Debian GNU/Linux forky/sid (forky) x86_64 - 7.1.12+deb14-amd64

GGML backends

Vulkan

Command-line arguments used

Steps to reproduce

What you expected to happen

What actually happened

Logs / error messages / stack trace

No response

Additional context / environment details

No response