The cable tests fine, the laptop connects fine — but join two switches with it and the uplink is dead. This strange pattern has a logical cause, and it is almost always inside the cable's wiring. Here is the complete diagnosis, in the right order.
It sounds impossible the first time you meet it: one cable, two results. Between a switch and a laptop it links and browses normally. Between two switches — nothing, or a link that drops the moment traffic flows. In the video, Technical Anshu works through exactly this fault and finds the logical answer most installers were never taught: colour coding is not decoration. Ethernet cares not only that the same colours meet at both ends, but that the colours arrive in the standard order, because that order is what keeps the signal pairs twisted together. Work through the steps below in order — each one eliminates a suspect, so the fix at the end is a conclusion, not a guess.
Before touching anything, look at what the link is telling you. Plug the suspect cable between the two switches and note the port LED. No light at all means the two ports never agreed there is a live connection. A light at 100 Mbps on a gigabit uplink means the ports linked but fell back — a strong hint that some pairs are not usable. Then plug the same cable into a laptop and open the adapter status: if the laptop also shows 100 Mbps instead of 1.0 Gbps, the cable itself is limited, no matter how well it seems to work. The speed a link negotiates is the cheapest diagnostic you own, and it is free.
Take a short, factory-made patch cord — the kind that came in a box, not one crimped on site — and join the two switches with it, in the same two ports.
This single test splits the problem in half. Skipping it is how technicians burn an hour re-crimping a cable that was never the problem — or replacing a switch that was never faulty.
Hold both RJ45 plugs side by side, clip down, and read the wire colours through the clear plastic. Both ends must follow the same standard. The most common standard, T568B, is:
Pin 1 White-Orange Pin 2 Orange Pin 3 White-Green Pin 4 Blue Pin 5 White-Blue Pin 6 Green Pin 7 White-Brown Pin 8 Brown
T568A is equally valid — it swaps the orange and green pairs. What is not valid is inventing your own order. Matching colours pin-to-pin makes the cable "straight", but if the colour order itself is wrong, the twisted pairs inside get split — and that is exactly the trap this video exposes.
Ethernet does not just need eight connected wires; it needs the right wires twisted together. The signal on pin 1 is cancelled out by its partner on pin 2 because they are one twisted pair. If a cable is punched by grouping colours neatly — orange pair, then blue pair, then green pair, then brown pair — pins 3 and 6 end up carrying signals whose partners are not twisted with them. That is a split pair.
A cheap cable tester only checks that pin 1 reaches pin 1, pin 2 reaches pin 2 — it beeps happily on a split-pair cable. A PC at 100 Mbps over a short run often survives it, because 100 Mbps uses only two pairs and forgives a lot. A gigabit uplink uses all four pairs at once, in both directions, and simply refuses. This is why "the tester passed it" and "it works on the laptop" prove far less than they seem to.
If the cable checks out, test the ports. Move the uplink to a different free port on each switch, one switch at a time. On a managed switch, also confirm the port is not shut down, not locked to a VLAN that excludes the uplink, and has no port-security limit blocking the second switch. On an unmanaged switch there is nothing to configure — which itself is useful information, because it removes settings from the suspect list. Also clean the obvious basics: a plug that does not click is not fully seated, and a port full of dust in a site office is more common than anyone admits.
Leave both uplink ports on auto speed and auto duplex. Gigabit over copper was designed to negotiate — the two ends even elect a clock master during negotiation — so forcing one side to "1000 full" while the other side stays on auto can produce a link that flaps, negotiates down, or carries heavy errors while looking up. If someone has manually set speed or duplex on either switch in the past, set it back to auto and retest before blaming hardware. A full case guide in this set is devoted to this forced-settings trap.
In the video's building example, switches sit on different floors of a 12-storey building and the uplink is a long fixed run, not a patch cord. A hand-crimped end can be perfect and the run can still fail — a rodent bite mid-run, a kink behind a ceiling panel, or a cable stapled too tightly can damage one pair. Patch a short factory cable directly between the switches to prove them, then test the fixed run section by section with a spare cable where you can. The 100 metre channel limit (90 metres of fixed cable plus patch cords) is also real: a run that "should be about 80 metres" is sometimes not.
When split pairs are the finding, there is no setting that rescues the cable — re-terminate it. Cut both plugs off, and crimp fresh ones following T568B at both ends (or your site's documented standard), keeping the twists intact right up to the plug and untwisting as little as physically possible. Retest in this order: continuity tester, laptop link speed, then the switch-to-switch uplink at 1000 Mbps. An uplink that links at gigabit and stays clean under a large file copy is fixed; one that links at 100 is not, whatever the lights say.
See the live troubleshooting in the video — the same cable tested with a laptop, then between two switches, and the wiring fault found on camera.
▶️ Watch Full VideoBecause the two jobs ask different things of the cable. A laptop link at 100 Mbps uses only two of the four pairs and tolerates sloppy wiring over a short run. A gigabit switch uplink uses all four pairs at once, in both directions, so a cable with split pairs — connected correctly pin to pin but with the wrong wires twisted together — passes a PC and fails the uplink.
Yes. A basic continuity tester proves each pin reaches the same pin at the other end. It cannot detect split pairs, where pins 3 and 6 carry signals whose twisted partners are elsewhere. That fault passes the tester, often works at 100 Mbps, and breaks gigabit links. Comparing the plug colours against T568B at both ends is the simplest way to spot it.
On modern switches, almost never — gigabit ports use auto MDIX and sort out the transmit and receive pairs themselves, and gigabit needs all four pairs wired correctly anyway. A crossover cable cannot rescue a split-pair cable. Reserve crossover thinking for very old 10/100 equipment without auto MDIX.