Summary
Now that Cat6, Cat6a, and multi-gigabit Ethernet are readily available and cheap, Cat5e feels like a cable category that should have retired years ago. My homelab still runs on Cat5e, mainly because I bought a spool years ago and don’t want to waste it. I recently cleaned out my shed and found several questionable Cat5e specimens lying around, including one that had spent years getting rained on. I honestly expected a clean Cat6a cable would embarrass Cat5e. Instead, I directly connected a Windows 11 laptop to a Linux PC with USB 3.0 gigabit adapters and watched all three cables hit 948 Mbps. Only one failed, showing that old and electrically broken aren’t the same thing. I gave Cat6 every chance to win The direct connection removed internet, storage, and router bottlenecks Ethernet category numbers aren’t just for marketing. Cat5e, Cat6, and Cat6a all contain the same twisted copper pairs, but the newer categories have increasingly better limits on interference and signal loss. Cat5e is rated to 100MHz, Cat6 pushes 250MHz, and Cat5a hits 500MHz. Cat6 typically uses tighter twists and often yourself. Cat6a can even control interference from neighboring cables, meaning it can allow 10 Gbps speeds across a full 100-meter cable. These numbers are impressive, but what do they mean for a typical homelab? To see what, if any, difference the categories made to my ordinary gigabit hardware, I connected my Windows 11 laptop directly to a Linux PC via two USB 3.0 gigabit adapters. I assigned 10.10.10.10/24 to Windows, and 10.10.10.11/24 to Linux, with no router, internet connection, or switch between them. The Linux PC ran the network tool iperf3 in server mode: iperf3 -s Windows generated network traffic directly into memory: iperf3 -c 10.10.10.11 -R -t 30 -i 0 I tested this way to avoid turning the Linux PC’s mechanical hard drive into a bottleneck. The -R option made the Linux PC send for 30 seconds, while -i 0 removed all the repetitive updates between intervals. Before testing each time, I checked the connection with ethtool . Every working cable needed to pass negotiation at 1000 Mbps and achieve full duplex before I could start. Since my network adapters were gigabit, they couldn’t actually reveal the top speed of these cables, but that was intentional. I personally wanted to know whether replacing Cat5e would actually improve the gigabit network that most people already own. So, the first test I did was to establish the speed ceiling by testing a clean Cat6a cable, which came in at 948 Mbps. The corroded Cat5e cable refused to be slower It matched Cat6 before surviving a ten-minute endurance test The clean Cat6a cable didn’t win. Neither did the new Cat5e cable. All three working cables reached the same practical gigabit ceiling when the Linux PC sent data to Windows. | Cable | Linux to Windows | Windows to Linux | Retransmissions | | Cat6a control | 948Mbps | 907Mbps | | | New Cat5e | 948Mbps | 906Mbps | | | Corroded Cable | 948Mbps | 906Mbps | The 1Mbps differences were pure measurement noise. Windows-to-Linux performance was consistently slower across both cable categories. That most likely pointed to either a USB adapter or its driver, rather than the cables themselves. A simultaneous bidirectional test exposed yet another limitation. Cat6a managed 906Mbps in one direction, then fell to 154Mbps in the other. The new Cat5e also produced a similar result. Since it was obvious the adapters couldn’t sustain equal traffic in both directions, I unfortunately had to exclude those results from the primary comparison. The 30-second burst testing that produced the results in the table wasn’t enough for the cable that honestly should have been destined for the trash. I reconnected the corroded and water-damaged cable and ran a longer test: iperf3 -c 10.10.10.11 -R -t 600 -i 60 To my absolute astonishment, every 60-second interval remained at exactly 948Mbps. Over 10 minutes, the cable had transferred an impressive 66.2GB without a single TCP retransmission. The Linux PC’s persistent error and drop counters didn’t increase either, and in the end, Windows reported zero dropped packets or errors. There was a hard limit to the resilience of Cat5e. Another cable I found while cleaning had been chewed badly enough to damage its copper conductors. This cable couldn’t establish a connection at any speed, with ethtool reporting Speed: Unknown! And Link detected: no . The point of even testing this broken cable was to separate cosmetic neglect from electrical damage. Broken clips, a bit of dirt, and surface corrosion didn’t slow down my water-damaged cable because its twisted pairs remained electrically intact. The cable the hungry rodent had chewed through had lost continuity through just one conductor (the solid-orange strand). No category rating in the world is going to repair that. Before replacing a Cat5e cable, check its negotiated link speed first. If two gigabit devices connect at only 100Mbps, inspect the connectors and terminations in the adapters as well. Gigabit Ethernet needs all four twisted pairs. One damaged pin may cause the speed to drop back to 100Mbps. Cat6a is still faster, just not on my gigabit network Faster cables help only when the surrounding hardware can use them This test wasn’t about proving Cat6a pointless, because it’s not. I used short patch cables, not a 100-meter run bundled up with AC power cables inside a wall. Cat6a provides the signaling bandwidth and crosstalk protection that those kinds of managed cable runs need. Cat6 and Cat6a also make sense when new wiring is intended to support 5GbE or 10GbE. My gigabit adapters prevented me from discovering the true maximum speed of any modern cable. I did end up giving the water-damaged cable a more realistic test. I placed an old unmanaged gigabit switch between the two computers and used the onboard gigabit Ethernet adapters. The cable from the Windows laptop to the switch was Cat6a, and the Cat5e cable ran from the switch to the Linux PC. The setup changed nothing. iperf3 still reported exactly 948 Mbps with zero retransmissions. Testing again with two Cat6a cables got me 1Mbps of speed improvement. Those test results identified the upgrade rule. Replacing functional Cat5e can’t improve a network when its adapters stop at 1Gbps. Faster Ethernet is more useful for local NAS transfers, backups, VM images, media editing, and server communication. It only accelerates internet access when the plan, modem, router, switches, and PC adapters all exceed gigabit speeds. Just one gigabit component in that mix will set the ~948 Mbps ceiling. I’m obviously going to throw out the rodent-damaged cable, and the corroded, water-damaged cable will probably return to retirement. However, I won’t be replacing working Cat5e just because Cat8a can theoretically hit higher speeds. On my gigabit network, the old cable was already as fast as everything else connected to it.