Yeah, this one's maddening — you're mid-download or on a call and the NIC drops, reconnects, drops again, and Windows keeps dinging at you like it's trying to be helpful. Let's kill it.
Jump straight to the fix
Nine times out of ten, it's power management, a trash cable, or a stale driver fighting your NIC. Do these in order and stop when the flapping stops.
- Turn off adapter power saving. This is the #1 cause on Realtek and Intel NICs. Open Device Manager (right-click Start, pick Device Manager), expand Network adapters, right-click your Ethernet adapter, choose Properties. Go to the Power Management tab and uncheck Allow the computer to turn off this device to save power. If that tab doesn't exist, skip to step 2.
- Disable Energy Efficient Ethernet (EEE) and Green Ethernet. Still in the adapter's Properties, go to the Advanced tab. Look for Energy Efficient Ethernet, Green Ethernet, EEE, or Power Saving Mode. Set it to Disabled. On Intel adapters it's sometimes called Gigabit Lite — kill that too.
- Swap the cable. Yes, actually swap it. A marginal Cat5e with a crushed kink or a broken retention clip will negotiate gigabit, then drop, then re-negotiate. Wash, rinse, repeat every 3-10 seconds. I had a client last month whose entire print queue died because of a $4 patch cable behind a filing cabinet. Replaced the cable, problem gone for six months and counting.
- Try a different switch port. Ports go bad. Especially on cheap unmanaged switches that have been running hot in a closet since 2018.
- Update or rollback the driver. Windows Update will happily push a Realtek driver from 2019 that breaks your 2023 motherboard. Grab the latest from your motherboard or laptop vendor — not from Windows Update. If you're already on the newest and the problem started right after an update, roll back in Device Manager: Properties → Driver → Roll Back Driver.
- Force the link speed to match your hardware. If you're on a gigabit-capable router but only have Cat5 running through a wall from 2004, force Speed & Duplex to 100 Mbps Full Duplex in the Advanced tab. It'll stop flapping because it's no longer trying to negotiate a speed the cable can't sustain.
If you want to confirm it's a link-layer problem before touching anything, open an admin Command Prompt and run this for a minute:
ping -t 192.168.1.1
(Swap in your gateway IP.) If you see Request timed out followed by Destination host unreachable in a repeating loop, you're watching the link physically drop and come back. That's hardware, power management, or cable. Not DNS, not your ISP.
Why this actually works
Windows has two separate power systems fighting over your NIC. There's the OS-level Allow the computer to turn off this device checkbox, and there's firmware-level EEE inside the chipset itself. Either one can put the PHY into a low-power state mid-transfer. When the NIC wakes back up, the link renegotiates, Windows re-runs DHCP or re-ARPs, and you see a disconnect-reconnect cycle that looks suspiciously like a software bug. It's not. It's power management doing exactly what it was told to do, just badly.
Cable issues are the second big culprit because gigabit Ethernet pushes all four pairs at 125 MB/s each. A cable that passes a basic continuity test can still fail at gigabit — crosstalk, impedance mismatch, a staple through the jacket. Windows will downshift to 100 Mbps, then try to renegotiate back up, then downshift again. That's your "reconnecting every few seconds" symptom.
Less common variations
If the standard stuff didn't fix it, work through these:
- USB Ethernet adapter on a hub. Bus-powered hubs can't supply the adapter's current draw when other devices spike. Plug the adapter directly into a rear USB port on a desktop, or a powered hub. Front-panel USB on cheap cases is notoriously flaky for this.
- Wake-on-LAN conflicts. Some BIOSes have a WoL setting that fights the OS. Disable Wake on Magic Packet and Wake on Pattern Match in the adapter's Power Management tab as a test.
- Duplicate IP or rogue DHCP server. If another device on the network grabs the same IP, you'll get a flapping loop. Check with
ipconfig /allandarp -a. Look for a second MAC with your gateway's IP. - Killer/Nahimic/Dragon software. Killer NICs ship with bandwidth-shaping drivers that are, politely, garbage. Uninstall the Killer Intelligence Center and use the plain manufacturer driver. I've fixed at least a dozen of these.
- BIOS PCIe ASPM settings. On some Intel boards, PCIe ASPM set to L1 makes the NIC drop under load. Set it to Disabled. This one's rare, but I've seen it twice on Gigabyte Z690 boards.
- Failing PoE injector or switch. If the NIC is fine on a direct router port but drops through the PoE run, the injector or the switch is dying.
Prevention
Disable power management on every desktop NIC you own. Desktops aren't laptops — there's zero reason to power-save the Ethernet port, and the setting adds nothing but problems. Buy decent cables: Belkin, Monoprice, or anything rated Cat6 from a real brand. Skip the $3 Amazon specials with no certification. Keep a spare 6ft patch cable in your bag, because you'll need it. And when Windows Update offers a "newer" driver for a NIC that's already working, let it age a couple weeks before you install it. The cure is usually worse than the disease.