Loading GSP firmware from an AMD Strix laptop to a TB5 3090 eGPU causes instant reboot

Graphics / Linux Linux
1

I have a TUXEDO IBP-14 gen10 AMD laptop (HX370 APU, based on TongFang X4SP4NAL) and an RTX 3090 that I’ve been successfully connecting using various eGPU adapters:

…and it has all been working perfectly stable on my Debian-13 system with the DC drivers versions 575, 580 (from the D-12 repo) and 590 (from the D-13 repo), both open and proprietary flavors, across several kernel versions (6.16.1-6.18.x).

Recently I’ve purchased a new TB5 adapter: Minisforum DEG2: it is quite new, but a few reviews it received from Windows users praised it for perfect stability and flawless work.
However when I connected my 3090 via my new DEG2 to my laptop, it abruptly rebooted about 0.5-1 second after the driver was loaded. I’ve tried all the mentioned above driver versions and even Nouveau, different distros (live Fedora-43+Nouveau, live POP_OS-24.04+v580), but the result was always the same (I even wanted to try Windows-11, but it’s uninstallable on this laptop currently).

After some hair-pulling and extensive testing, I’ve managed to notice, that the operation that causes the reboots is somehow related to GSP firmware: I’ve verified that the following configurations allow the eGPU to work mostly stable:

  • Nouveau driver without firmware-nvidia-graphics package (containing GSP v570.144)
  • proprietary driver v580 and v590 with options nvidia NVreg_EnableGpuFirmware=0 (the open flavor still reboots even with this param set to 0)

Here is the reference bug report file when it works fine with proprietary v590 and no GSP firmware:
nvidia-bug-report-deg2-v590prop-noGsp.log.gz (1.1 MB)

It is impossible to get a bug report file with GSP firmware present due to the almost immediate reboot. However, if I remount my filesystem into sync mode and continuously stream /dev/kmsg to a file there (using dmesg --follow), then sometimes (like about every 3rd attempt) a split-second before a reboot the following amdgpu entries are caught:

[  203.722070] amdgpu 0000:65:00.0: amdgpu: Fence fallback timer expired on ring gfx_0.0.0
[  203.722081] amdgpu 0000:65:00.0: amdgpu: Fence fallback timer expired on ring sdma0
[  204.262063] amdgpu 0000:65:00.0: amdgpu: Fence fallback timer expired on ring comp_1.2.0

Here is the full /dev/kmsg contents from a “GSP boot”:
dmesg-deg2-x4sp4nal-debian13-6.18.5-v580prop-gsp.log (220.5 KB)

Not sure if amdgpu is to blame here or if it is just a victim of a general f*ck-up on the PCIe bus, but I will post it to their issue tracker just in case and will provide a link here later.
UPDATE: Making sure you're not a bot!

Corresponding egpu.io thread: WIP: 2025 14″ TUXEDO InfinityBook Pro 14 Gen10 (890M) [RAI3,12C,HX] + RTX 3090 @ 64Gbps-USB4v1>TB5 (Minisforum DEG2) + Linux Debian trixie // loading NV GSP firmware reboots the laptop, occasional falling off the bus, daisy-chaining fails consistent… | Work-In-Progress Builds

I will also post to linux-usb in case it’s a problem in the Linux thunderbolt module.