RJ45 Wiring Table

One swap: the orange and green pairs exchange positions. In T568A, pin 1 is white/green and pin 2 green (green pair on pins 1-2); in T568B, pin 1 is white/orange and pin 2 orange, with the green pair moved to pins 3 and 6. The blue pair (pins 4-5) and brown pair (pins 7-8) never move. Electrically the two are mirror-image equals - signal pairs are balanced, so which physical pair carries which channel is irrelevant as long as both ends agree. The confusion is purely procedural: mixing the standards in one cable produces a crossover, and mixing them across a building produces links that sometimes work at 100 Mbps (which ignores two of the four pairs) and mysteriously fail at gigabit (which does not). The house rule that resolves everything: B is the de facto installed base in North America, A is preferred in US federal contracts, and whichever your existing jacks use, use everywhere.

Because gigabit was designed to end the argument. Old 10 and 100 Mbps Ethernet had separate transmit and receive wire pairs, so connecting two like devices (PC to PC, switch to switch) required a crossover cable to route one end's transmit into the other's receive. T568A-vs-B on opposite ends was exactly that swap, taken advantage of by the wire order. 1000BASE-T changed the physics: it uses all four pairs, each one bidirectional simultaneously, so there is no fixed TX/RX pair to swap - and the standard added Auto MDI/MDI-X, where each port detects the wiring and swaps its role electronically in milliseconds. Every gigabit port for the last two decades negotiates this automatically, so a crossover cable connects nothing that a normal patch cable would not. The one legacy use left is forcing a role on ancient 100 Mbps gear, and even there a modern switch's uplink handling makes it unnecessary. Keep one in the bag for museum work; wire your building to a single standard regardless.

Depends on speed, and the table's pair column shows why miswires fool people. 10 and 100 Mbps Ethernet use two pairs: pins 1-2 and 3-6 (the orange and green in T568B) - the blue and brown pairs sit idle, which is why a cable with two broken conductors can still pass 100 Mbps and why 10/100 systems could share cable with phone lines. 1000BASE-T and faster use all four pairs, simultaneously and bidirectionally, with phantom circuits derived across them - so every conductor matters, and so does the pairing: each data channel rides the physical twist of one pair, and a split pair (wires from two different pairs terminated on a channel's pins) destroys the noise cancellation even though every pin shows continuity. This is also why Power over Ethernet works so cleanly: 802.3af Alternative A powers over the same two pairs that carry 10/100 data (pins 1-2 and 3-6), and Alternative B powers over the spare pairs (pins 4-5 and 7-8) - the standard certified both splittings, and which one your injector uses is invisible to the link.

Almost certainly a split pair. The failure signature is exact: a basic continuity tester shows all eight conductors connected in the right pin order, but gigabit negotiates slowly, drops to 100 Mbps, or floods errors - because the data channels depend on each twisted pair's cancellation, and if the two wires of a pair land on pins belonging to different channels (the classic error: blue and white/blue placed by color order instead of pair order), every transmission couples into its neighbor. The color table prevents it if followed as pairs, not as individual colors: sort white/orange with orange as a twisted unit, not as the second and third wires of a row. Cheap testers that only verify continuity cannot see the fault - a wire-map tester that reports pairs, or a certifier that measures NEXT (near-end crosstalk), is the only instrument that catches it. The second suspect is the strip-back: more than about 25 mm of untwisted cable at the termination starts degrading crosstalk performance on its own, which is why the standard tolerates at most 13 mm at the plug.