Beware of using Ethernet port of GTR9PRO if you are running a homelab with multiple devices. When NIC malfunction happens, what I also experienced was jamming of local network; more specifically, the other devices that share same switch suffered from extremely unstable connection.
What caused this symptom was, NIC that fell into zombie-like state sent flood of abnormal frames to network, vandalizing LAN. It resolved only after I cutting the power supply for NIC by unplugging the device.
I suspect the core reason for Ethernet malfunctioning is a design failure on power supply system. Every time I load and run LLM with web search tools, which would trigger significant load on iGPU, SSD, NIC, Ethernet failure happened. I think they might share same power rail (probably 3.3v), and iGPU of Strix Halo’s abrupt power draw might be causing insecure power supply, such as sudden voltage drop. I remember seeing a comment on Reddit mentioning voltage drop.
