The most common source of Ethernet connection problems is a faulty or damaged cable, but the full list includes damaged connectors, cables that are too long, electromagnetic interference, switching loops, duplex mismatches, and driver or hardware failures. Most of these issues share the same frustrating symptoms: dropped connections, slow speeds, or a link that works sometimes but not others. Here’s how to identify what’s actually going wrong.
Damaged Cables and Connectors
Physical damage is the single most frequent cause of Ethernet trouble because cables get stepped on, pinched under furniture, bent at sharp angles, and yanked out by the connector instead of the clip. Any of these can break one or more of the eight tiny copper wires inside the cable, creating intermittent drops that are hard to pin down. A cable can look perfectly fine on the outside while carrying a broken conductor inside the jacket.
The plastic clip on the RJ45 connector is another weak point. Once that clip breaks or weakens, the connector slowly works its way out of the port, just enough to cause random dropouts even though it still looks plugged in. This is especially common with older patch cables that have been plugged and unplugged many times. If your connection cuts out when you bump the cable or move your device, a loose connector is the likely culprit.
Environmental factors also degrade connectors over time. Humidity accelerates oxidation on the metal contact pins, temperature swings loosen tolerances, and dust buildup inside the port interferes with the electrical contact between the plug and jack. If you’re troubleshooting a connection that gradually got worse over months, dirty or corroded contacts are worth checking.
Cable Length Beyond 100 Meters
Standard Cat 5e and Cat 6 Ethernet cables have a maximum reliable length of 100 meters (328 feet). Beyond that distance, the electrical signal weakens enough to cause packet loss, slow speeds, or a connection that refuses to establish at all. This limit applies to the total cable run, including any patch cables at either end.
Cat 8 cables, designed for higher speeds up to 40 Gbps, have an even shorter maximum of just 30 meters (98 feet). If you’re running cable through a large building or between floors, it’s easy to exceed these limits without realizing it. The fix is adding a switch or repeater partway through the run to regenerate the signal.
Electromagnetic Interference
Most Ethernet cables in homes and offices are unshielded twisted pair (UTP). They’re flexible, affordable, and work perfectly in typical environments, but they’re vulnerable to electromagnetic interference from nearby equipment. Heavy-duty electrical motors, HVAC systems, fluorescent lighting, and power lines running parallel to network cables can all inject noise into the signal.
The telltale sign of interference is a connection that works fine most of the time but develops errors or slowdowns when specific equipment kicks on. If your Ethernet drops whenever the air conditioning starts or when you’re near a bank of fluorescent lights, interference is the likely cause. Running Ethernet cables away from power lines, or switching to shielded cable in noisy environments, solves the problem.
Duplex Mismatch
This one is deceptive. A duplex mismatch happens when one end of the connection (your computer’s network adapter, for example) is set to full duplex while the other end (the switch or router) is set to half duplex. The link comes up, pings work fine, and everything seems normal. Then the moment you try to transfer a file or load a heavy web page, the connection slows to a crawl or locks up entirely.
The reason is mechanical: the full-duplex side sends and receives data simultaneously, which is normal for it. But the half-duplex side interprets incoming data during its own transmission as a collision, so it stops, sends a jam signal, and retries. The result is massive packet loss and retransmissions that make the connection nearly unusable under load while appearing healthy during simple tests. Speed is also asymmetrical, performing far worse in one direction than the other.
Most modern equipment negotiates duplex settings automatically, but mismatches still happen when one side is manually configured or when auto-negotiation fails. Checking the link settings on both the adapter and the switch port usually reveals the problem quickly.
Switching Loops
If you accidentally connect two switches to each other with more than one cable, or connect a cable from one port on a switch back into another port on the same switch, you create a loop. The consequences are dramatic. A single broadcast frame enters the loop and circulates endlessly because, unlike internet traffic, local Ethernet frames have no built-in expiration. Each pass through the loop generates copies, and within seconds, the network is flooded with an exponentially growing storm of duplicate frames.
A broadcast storm can take down every device on the network, not just the ones near the loop. Symptoms include all devices losing connectivity simultaneously, switches with every port light blinking frantically, and a network that recovers briefly when you unplug a cable but crashes again moments after reconnecting it. Managed switches prevent this with a protocol called Spanning Tree, but unmanaged consumer switches have no such protection. If your entire network suddenly dies, check for any cable that connects a switch back to itself or creates a redundant path between two switches.
Driver and Software Problems
Your operating system communicates with the Ethernet hardware through a driver. When that driver is outdated, corrupted, or incompatible, you get symptoms that look like a hardware failure but aren’t: the connection drops randomly, speeds tank, or the adapter vanishes from your system entirely.
On Windows, a network adapter showing Error Code 10 (“device cannot start”) typically points to a driver issue. You can check for transmission errors by running a command in PowerShell that shows received errors and outbound errors on your adapter. If those numbers are climbing, something is wrong at either the driver or hardware level. Reinstalling or updating the driver is the first step. On Linux, proprietary chipsets (Broadcom is a common one) sometimes need drivers that aren’t included by default, which means a fresh install may not have network support until you manually add the correct driver package.
Failed Network Hardware
When cables, connectors, settings, and drivers have all been ruled out, the network adapter itself may have failed. Integrated Ethernet ports on motherboards and standalone PCIe network cards both fail over time, sometimes gradually.
The clearest sign of hardware failure is an adapter that repeatedly disappears from your device manager, reappears after a restart, and continues showing error codes even after multiple fresh driver installations. Error Code 43 on Windows (“device has stopped working”) frequently indicates the hardware itself is dead rather than a software problem.
For desktop users with a PCIe network card, it’s worth reseating the card. Power down, unplug the machine, open the case, remove the card, clean the PCIe connector with compressed air, and push it firmly back into the slot. If a different PCIe slot is available, try that. A card that works in one slot but not another means the motherboard slot failed, not the card. The simplest diagnostic of all: plug in a cheap USB Ethernet adapter. If it works perfectly while the built-in adapter fails, you’ve confirmed the onboard hardware is dead and can use the USB adapter as a permanent workaround or plan a replacement.