Hey all,
So I’ve been watching this thread and applied the Beelink 1.08 BIOS, Intel E610 1.30 firmware, built the ixgbe-6.2.5 drivers on CachyOS (kernel 6.17). And of course trying to use the Ethernet ports consistently resulted in NVM firmware crashes after a short period, so I decided to use the WiFi and wait patiently a little while longer for a fix.
That changed on Monday when the 10G network gear arrived and I really wanted to get it up and running, so decided to investigate the issue and found a solution that, for me at least, is reproducible and so far proving to be robust.
The issue strongly appears to manifest when the Ethernet controller/firmware cold boots. If an Ethernet port is linked during cold boot then the controller (memory) state is left buggy and will later cause instability and the NVM to crash, however if there is no link during cold boot initialization is correct (in this context at least) and the Ethernet link is stable.
After initial testing I’ve been running for a couple of days now non-stop without a single issue, I’m only using one port for now but since it’s stable I’ll order the hardware to cable the other port, hopefully given the nature of the fault using both ports won’t be an issue.
So try the following solution:
- Remove all power from the GTR9
- Remove any ethernet cables
- Add power to the GTR9
- Wait a few seconds
- Plug in the Ethernet cable
- Boot the computer
Don’t plug the Ethernet cable in as soon as the ports light-up, wait a couple of seconds, or boot the OS first (but that’s not necessary). I’ve tested this workaround on both ports and have achieved a stable Ethernet connection.
After applying the workaround you don’t need to do so again unless you power down/reboot the Ethernet controller, so you can reboot/shutdown the GTR9 without having to repeat the workaround.
I’ve not done specific performance testing, but have stress tested the system, I’ve also been hammering the 10G Ethernet link for a couple of days without issue (currently speeds up to 8.99 Gbits), whereas before the Ethernet would become unstable and crash within a few minutes.
For completeness here is my setup:
GTR9 in performance mode (140W + sane fan settings, ZFS filesystem)
Reserved GPU memory: 512MB
MikroTik CRS309-1G-8S+ (SwOS)
10Gtek 80-Meter 10GBase-T SFP+ to RJ45 Transceiver
10 metre Cat7 cable
CachyOS (6.17.1-2-cachyos) (KDE Wayland, sleep disabled)
Kernel params (for unified memory): amd_iommu=off amdgpu.gttsize=131072 ttm.pages_limit=33554432
❯ sudo dmidecode | grep -A 5 ‘BIOS Inform’
BIOS Information
Vendor: American Megatrends International, LLC.
Version: GTRP108
Release Date: 09/16/2025
Address: 0xF0000
Runtime Size: 64 kB
❯ sudo ethtool -i enp197s0f0np0
driver: ixgbe
version: 6.2.5
firmware-version: 1.30 0×8000e899 1.3863.0
expansion-rom-version:
bus-info: 0000:c5:00.0
supports-statistics: yes
supports-test: yes
supports-eeprom-access: yes
supports-register-dump: yes
supports-priv-flags: yes
You’ll need to patch the driver for the 6.17 kernel, you can use my quick patch attached. And the CachyOS kernel is using LLVM, so from the ‘src’ directory:
❯ make KSRC=/lib/modules/$(uname -r)/build LLVM=1
❯ sudo make KSRC=/lib/modules/$(uname -r)/build LLVM=1 install
Thanks to @jotapesse for the KWIN_DRM_NO_AMS=1 workaround, which so far is working flawlessly :)
Anyway, obviously I’m pretty happy with the results and hope it also works for you.
