Ethernet Link Drop on Raspberry Pi 5 with Official NVMe HAT + PSU + Heat-Sink/Fan

Start by reproducing the reported 3–4 minute Ethernet failure on the Raspberry Pi 5 setup, then review the supplied dmesg, journalctl, ethtool, and ip output for the macb eth0 link transitions. Compare the full system information at the linked paste with the reported hardware and firmware setup; done means identifying a reproducible cause and an appropriate kernel issue resolution.

Describe the bug

After approximately 3–4 minutes of uptime, the Ethernet link on my Raspberry Pi 5 drops unexpectedly. Sometimes it recovers after a few seconds, sometimes it does not reconnect at all without manual intervention. This cycle repeats indefinitely. The system is fully updated and uses only official Raspberry Pi hardware, including the NVMe HAT with the included drive, the official 27W power supply, and Raspberry Pi OS Bookworm. Static IP is set via nmtui, no DHCP is used. This behavior severely impacts headless server usage via SSH or VNC.

Steps to reproduce the behaviour

1. Set up a Raspberry Pi 5 with the official NVMe HAT and included NVMe drive.
2. Use the official 27W Raspberry Pi USB-C power supply.
3. Attach the official Raspberry Pi 5 heatsink and fan.
4. Install Raspberry Pi OS (Bookworm) to the NVMe drive.
5. Boot the system with a static IP configured via nmtui.
6. Connect the Ethernet port to a router or switch.
7. Observe that the link initially comes up as expected.
8. Wait ~3–4 minutes.
9. Observe that the Ethernet link drops, sometimes re-establishes, sometimes stays down.
10. Repeatable on every reboot.

Device (s)

Raspberry Pi 5

System

https://pastebin.com/Lst2LZZk

Logs

Logs

dmesg (Relevant Ethernet Entries)

[  210.765030] macb 1f00100000.ethernet eth0: Link is Down
[  213.838421] macb 1f00100000.ethernet eth0: Link is Up - 1Gbps/Full - flow control tx

journalctl -b | grep eth0

Consistently confirms the link down/up cycle following initial boot.

ethtool eth0

Speed: 1000Mb/s
Duplex: Full
Link detected: Yes
driver: macb

ip a

2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500
    link/ether 2c:cf:67:f7:fc:3d
    inet 192.168.0.50/24 brd 192.168.0.255 scope global eth0

vcgencmd get_throttled

throttled=0x0

No undervoltage or thermal throttling is present. System is stable in terms of power.

Additional context

[  210.765030] macb 1f00100000.ethernet eth0: Link is Down
[  213.838421] macb 1f00100000.ethernet eth0: Link is Up - 1Gbps/Full - flow control tx

This behavior occurs consistently around 3–4 minutes after boot. The Ethernet interface (eth0) drops without any user interaction and sometimes fails to recover without a reboot.

Additional context

• The system is intended to run as a headless application server, so stable Ethernet is critical.
• Wi-Fi connectivity appears stable, suggesting the issue is isolated to the wired interface.
• The issue persists even with up-to-date firmware and a fully upgraded Raspberry Pi OS Bookworm installation.
• All hardware in use is officially branded Raspberry Pi (Pi 5, NVMe HAT with SSD, heatsink/fan, and 27W power supply).
• The static IP was set using nmtui, and no DHCP is involved.
• The Ethernet drops even when idle, with no traffic load.
• No undervoltage or thermal throttling is reported via vcgencmd get_throttled.