An Ethernet cable is a balanced twisted-pair copper cable, normally terminated with RJ45 (8P8C) connectors, that carries Ethernet frames between devices such as computers, switches, routers, wireless access points and patch panels. In network hardware, it is the most common physical medium for wired local area networks, defined by performance categories such as Cat5e, CAT6 and Cat6A and by the IEEE 802.3 family of physical layer standards.
Within networking, the cable is far more than a passive wire. Its category, conductor type, shielding and installed length decide which speeds a link can sustain, how much Power over Ethernet it can deliver safely and how well it resists interference. A poor cable choice is one of the most frequent hidden causes of slow or unstable links.
How an Ethernet cable carries data
An Ethernet cable contains four twisted pairs of insulated copper conductors, usually 22 to 24 AWG. Each pair carries a differential signal: the transmitter drives opposite voltages on the two wires, and the receiver reads only the difference, so noise that affects both wires equally is canceled. The twist rate differs between pairs, which reduces crosstalk, the coupling of signal from one pair into another.
Older 10BASE-T and 100BASE-TX links use only two pairs, one per direction. Gigabit Ethernet (1000BASE-T) and faster standards use all four pairs in both directions at once, relying on echo cancellation and advanced encoding. This is why a damaged pair that goes unnoticed at 100 Mbps can make a gigabit link fail or fall back to a lower speed.
The electrical parameters below determine whether a cable channel can support a given speed.
| Parameter | What it measures | Effect on the link |
|---|---|---|
| Insertion loss | Signal attenuation along the cable | Limits maximum length and speed |
| NEXT | Crosstalk between pairs at the near end | Causes bit errors at high frequency |
| Alien crosstalk | Coupling between neighboring cables | Critical for 10GBASE-T in bundles |
| Return loss | Reflections from impedance mismatches | Degrades signal quality and margin |
| DC resistance | Resistance of each conductor | Drives heat and voltage drop under PoE |
Ethernet cable categories and speeds
Ethernet cable categories, defined in ANSI/TIA-568 and the related ISO/IEC 11801 classes, specify the frequency bandwidth a cable must support. Higher bandwidth allows faster Ethernet standards over the standard 100-meter channel, which consists of up to 90 meters of permanent link plus patch cords.
| Category | Bandwidth | Typical maximum speed | Notes |
|---|---|---|---|
| Cat5e | 100 MHz | 1 Gbps at 100 m, 2.5 Gbps often | Minimum for new installations |
| CAT6 | 250 MHz | 1 Gbps at 100 m, 10 Gbps to about 55 m | Common choice for offices |
| Cat6A | 500 MHz | 10 Gbps at 100 m | Better alien crosstalk control |
| Cat7 and Cat7A | 600 to 1,000 MHz | 10 Gbps at 100 m | ISO class F, rarely used with RJ45 |
| Cat8 | 2,000 MHz | 25 or 40 Gbps up to 30 m | Data center short links only |
The multigigabit standards 2.5GBASE-T and 5GBASE-T, often marketed as 2.5 GbE and 5 GbE, were designed to run over existing Cat5e and CAT6 cabling, which makes them attractive for upgrading wireless access points without recabling. Speed always depends on the weakest component of the channel, including jacks, patch panels and patch cords.
Shielding, conductors and construction
Beyond the category, an Ethernet cable is described by its shielding code, its conductor type and its jacket rating. Shielding is written in the ISO/IEC format, such as U/UTP for fully unshielded cable or S/FTP for an overall braid plus individually foiled pairs.
| Option | Description | Best use |
|---|---|---|
| U/UTP | No shield, pairs only | Most office and home installations |
| F/UTP | Overall foil around all pairs | Moderate interference, Cat6A runs |
| S/FTP | Braid plus foil on each pair | Industrial sites and dense bundles |
| Solid conductor | One copper wire per conductor | Permanent in-wall horizontal cabling |
| Stranded conductor | Many thin wires per conductor | Flexible patch cords |
| Plenum or LSZH jacket | Low smoke, fire-rated sheath | Air-handling spaces and public buildings |
Shielded cable only works when the shield is bonded and grounded end to end through shielded jacks and patch panels. An interrupted or ungrounded shield can act as an antenna and make interference worse. The 568A and 568B wiring schemes are electrically equivalent; what matters is using the same scheme at both ends of a permanent link.
Ethernet cable compared with other media
Twisted-pair copper competes with fiber optics and direct attach copper in modern networks. Each medium has a distinct balance of distance, cost, power delivery and immunity to interference.
| Medium | Typical reach | Power delivery | Main advantage |
|---|---|---|---|
| Twisted-pair Ethernet cable | Up to 100 m | Yes, through PoE | Low cost and universal RJ45 ports |
| Multimode fiber | About 300 to 550 m at 10 Gbps | No | Electrical isolation and higher speeds |
| Single-mode fiber | Kilometers | No | Long-distance campus and WAN links |
| Direct attach copper | Up to about 7 m | No | Cheap SFP+ links inside a rack |
Copper remains dominant at the access layer, where endpoints need power and short runs, while fiber patch cables and SFP modules take over for uplinks, backbones and long distances.
Power over Ethernet and common cabling faults
Ethernet cables also carry DC power under the IEEE 802.3af, 802.3at and 802.3bt standards, delivering up to about 90 W at the power source with Type 4 PoE. Higher currents heat the cable, especially in large bundles, so thicker 23 AWG conductors and proper bundle sizes matter. Cheap copper-clad aluminum (CCA) cable has higher resistance, fails standards compliance and should never be used for PoE.
- Untwisting pairs at termination beyond about 13 mm increases crosstalk and causes certification failures.
- Exceeding the bend radius or crushing cable with tight ties alters impedance and raises return loss.
- Running parallel to power cables introduces interference; maintain separation or cross at right angles.
- Mixing categories in a channel limits it to the lowest category present.
- Skipping certification testing leaves marginal links that fail only under load or at higher speeds.
The short version
An Ethernet cable is the wired connection that links devices in a network, such as a computer to a router. Inside it are eight thin copper wires twisted into four pairs, which keeps signals clean and stops them from interfering with each other.
Cables come in categories with higher numbers supporting faster speeds. For most homes and offices, CAT6 cable is plenty, and the same cable can even power devices such as cameras and wireless access points.
Where it makes a difference
A school replacing its wireless network installs new access points that support 2.5 Gbps uplinks and draw about 25 W each. The existing cabling is a mix of certified Cat5e and unbranded patch cords bought over several years. During testing, several links negotiate only 1 Gbps and two access points reboot under load.
The installer certifies every channel and finds that the unbranded cords are copper-clad aluminum with high resistance, causing voltage drop under PoE. After replacing them with compliant solid-copper CAT6 patch cords, all access points link at 2.5 Gbps and run stably. The school avoids a full recabling project and spends only a fraction of the expected budget.