A NIC failure means your machine has no working network path. Fix it by reseating the card and cable, confirming the adapter appears in Device Manager or ip link, reinstalling the vendor driver, resetting the TCP/IP stack, and testing on a second port. If none of that revives the link, replace the card or use a USB adapter.
Symptoms
- No link light on the Ethernet port, or the light stays amber and never turns green
- Adapter missing from Device Manager, or shown with a yellow warning triangle
- Windows reports "Network cable unplugged" even with a known-good cable
- Intermittent drops, huge packet loss, or speeds capped far below the negotiated link rate
- ipconfig shows APIPA (169.254.x.x) or no IPv4 address at all
Common Causes
Corrupted or mismatched driver
Windows Update often installs a generic driver that conflicts with Intel, Realtek, or Broadcom silicon. After a feature update the adapter can stop initialising or drop link under load.
Loose PCIe seating or bent pins
A card that shifted during shipping or a case move loses contact on the PCIe edge. Onboard NICs suffer the same fate when the motherboard flexes near the RJ45 jack.
Physical port or magnetics damage
Static discharge, a lightning-induced surge over Ethernet, or a bent RJ45 clip can fry the transformer on the card. The OS still sees the chip, but no link ever negotiates.
Disabled adapter or wrong power state
Group Policy, a VPN client, or aggressive power management can disable the NIC or park it in D3. The device shows up but refuses to carry traffic.
Firmware or PXE ROM fault
Server-class NICs (Intel X710, Mellanox ConnectX) can hang if firmware and driver versions drift apart. Symptoms include link flapping and PCIe bus errors in the event log.
Step-by-Step Fix
- Confirm the failure is really the NIC, not the cable or switch port
Swap the patch cable for a known-good Cat6 run and move to a different switch port. Test the same port with a laptop. If that laptop links up and yours does not, the NIC (or its driver) is the fault. Note the link LED behaviour: no light at all points to hardware; blinking without traffic points to driver or IP config. - Reseat the card and inspect the port
Power down, unplug mains, and open the case. Remove the PCIe NIC, check the gold fingers for oxidation, and press it firmly back into the slot until the retention clip clicks. For onboard NICs, look for cracked solder around the RJ45 jack and any bent pins inside the port. A torch and a magnifier help here. - Verify the adapter is present at the OS level
On Windows, open Device Manager and expand Network adapters. On Linux, run `ip link` and `lspci | grep -i ethernet`. On macOS, check System Settings > Network. If the adapter is absent entirely, boot into BIOS/UEFI and confirm the onboard LAN or PCIe slot is enabled. A device missing from lspci is almost always dead hardware. - Reinstall the vendor driver cleanly
Right-click the adapter in Device Manager and choose Uninstall device, ticking "Delete the driver software". Reboot, then install the current driver directly from Intel, Realtek, Broadcom, or the OEM support page, not a driver-updater utility. On Linux, reload the module: `sudo modprobe -r e1000e && sudo modprobe e1000e` and check `dmesg` for initialisation errors. - Reset the TCP/IP stack and Winsock catalog
A working NIC with a corrupted stack behaves like dead hardware. In an elevated prompt run: `netsh int ip reset`, `netsh winsock reset`, `ipconfig /flushdns`, then reboot. On Linux, `sudo systemctl restart NetworkManager` or flush with `sudo ip addr flush dev eth0` and re-request DHCP via `sudo dhclient -v eth0`. - Disable aggressive power management and offload features
In the adapter's Advanced properties, uncheck "Allow the computer to turn off this device to save power". Then temporarily disable Large Send Offload, TCP Checksum Offload, and Energy Efficient Ethernet. These features are common triggers for link flapping on Realtek and older Intel chipsets. - Test with an alternative NIC to isolate the fault
Plug in a USB 3.0 Ethernet adapter or a spare PCIe card. If that link comes up cleanly on the same cable and switch port, the original NIC is confirmed faulty. This step also gets the user back online while you order a replacement, which matters more than any diagnostic detail. - Replace the card or escalate to the motherboard
For a discrete NIC, fit a replacement in a different PCIe slot and install the driver before connecting the cable. For a dead onboard NIC, disable it in BIOS to stop resource conflicts and add a PCIe or USB adapter. On servers under warranty, capture the MAC, event logs, and `lspci -vv` output before opening a vendor case.
NIC failure signals mapped to likely cause and first action
| Signal | Likely cause | First action |
|---|---|---|
| No link light, adapter missing in Device Manager | Dead NIC or disabled in BIOS | Check BIOS, reseat card, test spare adapter |
| Yellow triangle on adapter, Code 10 or 43 | Driver corruption or resource conflict | Uninstall driver with deletion, reinstall vendor build |
| Link comes up then drops every few minutes | EEE / power management or bad cable | Disable EEE and power-save, swap patch cable |
| APIPA address 169.254.x.x | DHCP unreachable or stack corruption | Run netsh int ip reset, verify switch port VLAN |
| Speed capped at 100 Mbps on a gigabit link | Cable pair damage or forced duplex | Replace cable, set adapter to Auto Negotiate |
| Random BSOD referencing the NIC driver | Firmware/driver mismatch | Match firmware to driver release notes, update both |
Prevention
- Pin the vendor driver version in WSUS or Intune so Windows Update cannot swap it for a generic build
- Use surge-protected patch panels or inline Ethernet surge arrestors on runs that leave the building
- Label and torque PCIe retention screws after any hardware move to stop cards from working loose
- Monitor interface error counters (CRC, discards) via SNMP so a failing NIC is caught before it goes fully dark
FAQ
How do I tell a NIC failure apart from a router or ISP outage?
Test a second device on the same cable and port. If that device gets an IP and reaches the internet, the network path is fine and the NIC is the suspect. You can also boot a Linux live USB on the affected machine, if the adapter works there, the fault is software; if it does not appear even in `lspci`, the hardware is gone.
Is it worth replacing an onboard NIC, or should I just add a card?
For most desktops and workstations, adding a PCIe or USB Ethernet adapter is faster and cheaper than replacing the motherboard. Disable the failed onboard NIC in BIOS so Windows stops trying to bind to it. On laptops, a USB 3.0 gigabit adapter is the standard fix and performs close to line rate.
Why does my NIC work after a reboot but fail again within hours?
That pattern usually points to a driver leaking resources, an overheating controller, or Energy Efficient Ethernet renegotiating badly with the switch. Update to the current vendor driver, disable EEE and Green Ethernet in the adapter's advanced properties, and check the card's heatsink is seated. Persistent thermal failures need airflow fixes or a card replacement.
Need hands-on help isolating a NIC fault across your fleet? Contact our IT support team for a diagnostic session.