If your switchport is set to half duplex, what is happening is that it is unable to receive and send traffic at the same time.
If it sends traffic while the wire is busy, you get a collision. When that happens, it waits for a random period of time before retrying again. This makes for an inefficient link.
When you set it to full duplex, then the switchport uses different wires for transmit (TX) and receive (RX) signals. It is then capable of receiving data whilst it is sending data, with no collisions on the wire.
Cisco kit can tell if there's a duplex mismatch - if the device on the other end of the wire is set with half instead of full, or vice-versa. It will alert you to these mismatches through error logging and/or to the console (depending on how your device is configured).
Graham is almost correct however there is a mistake in his explanation regarding pairs used in 10/100BaseT.
Back in the old days Ethernet was presented as a Coaxial cable - either Thick Ethernet (10Base5) or Thin Ethernet (10Base2). It was effectively a single cable up to 500-meters (10Base5) or 200-meters (10Base2) in length, terminated at either end with 50-ohm resisters (Terminators). All devices connected to this cable in parallel usually using 'T' pieces or 'Taps'.
Coaxial cable obviously only has one pair - the inner core (signal) and the outer sheath (ground), obviously in this scenario only a single device can transmit at a time. If two devices attempt to transmit at the same time there is a collision. In Ethernet NICs there is collision detection circuitry (CSMA/CD) that can detect when a collision occurs and forces the NIC to back off and wait a random time before attempting to transmit again.
Fast forward a bit and we have 10BaseT Ethernet that runs over UTP cabling (Cat3, Cat5 & Cat6). Since Cat5/6 cabling is star-wired - i.e. all outlets come back to a common wiring closet and each link is an individual cable, this new (at the time) standard included a transmit pair and a separate receive pair. If a 10BaseT hub is used then the operation is as before - i.e. half duplex, however if a switch is used where each port is its own collision domain and only two devices ever exist on the domain (the switchport and the attached device) then the collision detection circuitry can be disabled and the link configured for full-duplex operation. Regardless of the duplex for 10/100BaseT two pairs are used.
There is no explicit standard for 10BaseT at full-duplex, however when 100BaseT appeared it included a standard for auto-negotiation that allowed NIC's to negotiate speed & duplex with the switchport (802.3u).
The switch receives on one pair and transmits on another regardless of the duplex setting. If the interface is set to full-duplex then the collision detection circuitry is disabled and the interface can transmit & receive simultaneously. If the interface is set to half-duplex then the collision detection circuitry is enabled, if traffic is being received on the receive pair the interface cannot transmit until the reception has finished. If a collision occurs - i.e. both ends try and send at the same time then a collision will be detected by both parties (the host and the switchport) and they will back off and wait a random time before attempting to transmit again.
All this information is out there with very detailed descriptions. Why not try google? This will stop this forum filling up with repeated posts.....
I cannot understand, If the switch/pc uses 2 different pairs for tx and rx then how can a collision occur? and im still unsure what difference half / full duplex will make if tx and rx are on different pairs
We are pleased to announce availability of Beta software for 16.6.3.
16.6.3 will be the second rebuild on the 16.6 release train targeted
towards Catalyst 9500/9400/9300/3850/3650 switching platforms. We are
looking for early feedback from customers befor...
Introduction Featured Speakers Luis Espejel is the Telecommunications
Manager of IENova, an Oil & Gas company. Currently he works with Cisco
IOS® and Cisco IOS XE platforms, and NX to some extent. He has also
worked as a Senior Engineer with the Routing P...
In this session you can learn more about Layer 3 multicast and the best
practices to identify possible threats and take security measures. It
provides an overview of basic multicast, the best security practices for
use of this technology, and recommendati...