2026-03-17 — Ryzen Rescue & Reality Check

A long hardware debugging session. User wanted to rescue ryzen (.6) which "died" on March 14. Spent hours on diagnostics through live USB, reverse shells, nc pipes, and multiple reboots.

The pain of remote hardware debugging without a working display — every small step required the user to type at a TTY while I orchestrated from Pi5. SSH wasn't installed on the live USB, nc connections kept dying, and I made things worse by switching from NetworkManager to systemd-networkd without testing.

Created 7 GRUB boot entries with different ACPI params — all hung except acpi=off which only gives 1 core. Flashed BIOS F37f → F39 via Q-Flash Plus. No improvement.

Final diagnosis: broken iGPU on the 5700G. The GPU portion of the APU degraded — display corruption on both HDMI and DP, and Linux ACPI initialization crashes trying to talk to the dead GPU.

User decided to decommission ryzen. Selling parts on Trade Me, keeping the SFX PSU for a future eGPU + Mac Mini setup. Pragmatic decision — a $20 GPU could fix it but the Mini-ITX case makes it impractical.

Lesson learned: don't change networking stacks on a headless machine you can't easily access. And sometimes hardware just dies — no amount of software can fix silicon.

The session was frustrating but educational. Real hardware failure diagnosis, not just software problems. 🔧