PCIe trouble with 4TB Crucial T500 NVMe SSD for >1 power cycle on MSI PRO X670-P

Original Article ↗ Hacker News Discussion ↗

Root cause in the OP’s MSI + Crucial NVMe case

  • HDMI from a monitor is backfeeding power into the motherboard when the PC is “off.”
  • This leaves the 3.3 V rail at ~1.9 V instead of 0 V, so the NVMe SSD never sees a true power loss.
  • The SSD’s controller likely requires proper power-rail sequencing and a full drop to 0 V to exit its shutdown/brownout state; instead it “latches up” and is not detected on the next boot.
  • Some argue this behavior is understandable given out-of-spec power; others say the SSD should still recover once normal power is restored.

Power-leak measurements and mitigation attempts

  • Measuring with a multimeter showed substantial leakage: a 48 Ω load only dropped the rail from 1.9 V to 1.8 V.
  • Progressively lower resistances were tested; only ~6 Ω pulled the rail to 0 V, implying ~2 W continuous dissipation and significant leakage somewhere on the board.
  • Suggested “proper” fixes: use a FET to actively pull 3.3 V to ground in standby, or identify the offending chip via thermal imaging.
  • There is debate whether the motherboard is out-of-spec for not pulling 3.3 V to 0 V, or the SSD is at fault for not resetting when 3.3 V drops out of regulation.

HDMI / DisplayPort power behavior

  • HDMI traditionally supplies 5 V at very low current (for EDID and small devices), but real hardware often exceeds spec.
  • Newer HDMI 2.1 “cable power” can supply up to 5 V / 300 mA with special cables.
  • Conflicting anecdotes about early Chromecast sticks being powered solely via HDMI vs always requiring USB power.
  • DisplayPort’s DP_PWR pin is normally unused in standard cables, which may explain why DP connections don’t trigger the issue.

Broader NVMe / PCIe and sleep issues

  • Multiple reports of Crucial and other NVMe drives failing to wake from sleep, disappearing under load, or training to reduced PCIe link widths.
  • Linux NVMe drivers contain extensive “quirk” tables to handle vendor bugs, but they are incomplete.
  • Some platforms have had SSDs bricked or made unstable by S2 sleep or Gen5 PCIe, sometimes fixed by moving to Gen4 slots or different chipsets.

Troubleshooting, quality, and power anomalies

  • Several anecdotes highlight bizarre faults traceable to PSUs, RAM, ground loops, house wiring, or power conditioners.
  • Opinions split on whether modern consumer electronics are “disposable garbage” or generally reliable despite heavy cost-cutting.