Full duplex means a link can send and receive at the same time. Half duplex means it can do both, but only one direction at a time, so the two ends take turns. Every switched Ethernet link today runs full duplex, and Wi-Fi is still half duplex.
On a wired network the difference matters mostly when it goes wrong: a duplex mismatch, where one end runs full and the other half, gives a link that passes a ping and crawls under load.
- Full duplex sends and receives simultaneously. Half duplex takes turns
- Simplex is one direction only, such as a broadcast
- Switch ports run full duplex. Hubs ran half duplex, with collisions
- A duplex mismatch shows as late collisions on one end and CRC errors on the other
- Leave autonegotiation on at both ends. Forcing one side causes the mismatch
On this page
The modesWhat half duplex and full duplex mean
Duplex describes the direction of communication on a link between two connected devices, and full duplex vs half duplex is a question of whether both can transmit data at once. There are three transmission modes, and the question of simplex vs half duplex vs full duplex comes down to who may talk and when.
Simplex. Data flows in one direction only. The receiver has no way to answer on the same channel. A broadcast radio station or a TV signal is simplex.
Half duplex. Both devices can transmit, but not at the same time. A walkie-talkie is the standard example: press to talk, release to listen. Communication is bidirectional, and each direction waits for the other.
Full duplex. Both devices transmit and receive simultaneously, over separate paths. A telephone call is full duplex: both people can speak at once and still hear each other.
What is full duplex worth in numbers? A 100 Mbps link in full duplex mode carries 100 Mbps in each direction at once. The same link in half duplex shares 100 Mbps between both directions, minus time lost to collisions. That is why a full duplex port is sometimes described as 200 Mbps in total.
EthernetHow Ethernet moved from half duplex to full duplex
Early Ethernet was half duplex by design. Every device shared one cable, or one hub, so only one could transmit at a time. Devices listened before sending and backed off when two transmitted together. That method is CSMA/CD, carrier sense multiple access with collision detection, and the shared segment is a collision domain.
A hub keeps that model alive in a box: it repeats every frame out of every port, so all devices behind it still share one half duplex collision domain.
A switch ended it. Each switch port is its own segment with exactly two devices on it, the switch and the computer.
On twisted pair cabling, 10BASE-T and 100BASE-TX send on one wire pair and receive on another, so both ends can transmit at once without interfering. Full duplex Ethernet was standardized as IEEE 802.3x, and with it CSMA/CD is turned off, because a collision can no longer happen.
Gigabit Ethernet over copper uses all four pairs of a UTP cable in both directions at the same time, separating the signals with echo cancellation. 10 Gigabit Ethernet is defined for full duplex only. In practice, a modern wired link is either full duplex or broken.
This is a physical layer and data link matter. Nothing above it, not IP and not TCP, knows or cares which duplex mode the link uses, until the link starts dropping frames.
Duplex mismatchDuplex mismatch is the problem you will actually meet
The full duplex vs half duplex question rarely comes up as a design choice anymore. It comes up as a fault. A duplex mismatch is a link where one end runs full duplex and the other runs half duplex, and it is one of the classic causes of a network that works and is slow.
The usual cause is well meant. Someone forces a port to 100 Mbps full duplex, on the switch or on a server, and leaves the other end on autonegotiation.
The autonegotiating end hears no negotiation from its partner. It can still detect the speed from the signal, but not the duplex setting, and the Ethernet standard tells it to fall back to half duplex for compatibility with hubs.
The result is a mismatch that nobody configured on purpose. Both ends report the link as up at the right speed, so it looks healthy.
Then traffic flows in both directions. The full duplex end sends whenever it has data. The half duplex end sees incoming data while it is transmitting, treats it as a collision, stops and retries later.
The frames it abandons arrive at the full duplex end as damaged or undersized frames, which fail the checksum and are discarded.
DiagnosisHow to spot a duplex mismatch
The symptoms are distinctive once you know them.
Ping works. Small packets a second apart rarely meet traffic going the other way, so a ping test looks clean.
Transfers crawl. A large file copy, a backup or a video call starts and then collapses. TCP sees the lost frames as congestion, shrinks its window and retransmits, and the retransmissions are lost the same way.
One direction is worse. Throughput is usually asymmetric: much poorer in one direction than the other.
The counters disagree. This is the proof. On the half duplex end, the interface shows late collisions, which should never occur on a correctly working link. On the full duplex end, it shows CRC or FCS errors and runt frames.
To check, compare both ends of the same cable.
- On a managed switch, the port status shows the negotiated speed and duplex, and the interface counters show collisions, late collisions and CRC errors.
- On Windows, Get-NetAdapter in PowerShell lists each adapter's LinkSpeed and a FullDuplex field. Get-NetAdapterAdvancedProperty shows the configured value, which many drivers name Speed & Duplex. Get-NetAdapterStatistics shows received and outbound packet errors.
- On Linux, ethtool followed by the interface name shows the speed, the duplex and whether autonegotiation is on.
If one end says full and the other says half, you have found it. The slowness looks like packet loss from the outside, which is why it gets blamed on the internet provider.
AutonegotiationWhy autonegotiation stays on at both ends
Autonegotiation lets two Ethernet interfaces advertise what they support and agree on the best common speed and duplex. It works when both ends run it. It fails predictably when only one end does.
Forcing one end to full duplex creates the mismatch. A forced port stops advertising, so its partner falls back to half duplex. If you must fix a setting, fix it identically on both ends, and document it, because the next person will replace one side and leave the other.
Autonegotiation problems belong to the past. The habit of hard coding 100 full dates from early autonegotiation implementations that did not interoperate well. Current switches and adapters ship with it on by default, and gigabit copper links use it as part of bringing the link up.
A wrong result is a clue, not a setting to override. If a link negotiates 100 Mbps or half duplex between two gigabit devices, suspect the cable, a damaged pair or a bad patch panel port. Forcing the speed hides the fault.
Wi-FiWhy Wi-Fi is half duplex
Wi-Fi runs half duplex. A radio cannot listen on a channel while transmitting on that same channel, because its own signal drowns everything else out. So on each channel, one device transmits at a time and everyone else waits.
The access point and all its clients share that channel. Wi-Fi uses CSMA/CA, collision avoidance: devices listen, wait a random interval and then send, because a radio cannot detect a collision the way wired Ethernet could.
Two practical consequences follow for an office.
Airtime is shared. The rate a laptop shows is the rate for its turn, shared with every other device on that access point and channel, uploads and downloads together.
Slow devices cost everyone. A client far from the access point transmits at a low rate and occupies the channel longer for the same data. More access points on separate channels, and wired connections for desktops, help more than a faster headline rate.
PitfallsWhere people go wrong
Hard coding one side. Setting a server NIC or a switch port to 100 full while the other end autonegotiates is the most common cause of a duplex mismatch. Change both or neither.
Trusting a ping test. A clean ping proves the link passes small packets. It says nothing about a link that fails under sustained two-way traffic.
Reading only one end. The full duplex side never counts a collision. If you check only that end, you see CRC errors and blame the cable.
Forcing speed to fix a bad link. A gigabit link that negotiates down is telling you about the cable or the port. Replace the cable before overriding the negotiation.
Expecting Wi-Fi to behave like a switch port. Wi-Fi is shared and half duplex. A busy conference room on one access point will never match a wired desk.
Keeping a hub around. An old hub, or a cheap device with a broken autonegotiation, drags its links into half duplex. Replace it with a small switch.
ComparisonFull duplex and half duplex, and what each end reports in a mismatch
| Criterion | Full duplex | Half duplex |
|---|---|---|
| Direction | Both directions at once | One direction at a time |
| Capacity of a 100 Mbps link | 100 Mbps each way | 100 Mbps shared |
| Collisions | Cannot occur | Normal, handled by CSMA/CD |
| Devices on the segment | Exactly two | Two or more sharing |
| Where you find it | Switch ports, fiber, modern NICs | Hubs, Wi-Fi, walkie-talkies |
| Latency under load | Low and steady | Rises with contention |
| Mismatch counter on this end | CRC errors and runts | Late collisions |
The row that matters for troubleshooting is the last one. In a half duplex vs full duplex mismatch, each end records a different kind of error, so checking only one side of the link can hide the problem. Read the counters on both ends before touching a setting.
FAQFrequently asked questions
What is the difference between full duplex and half duplex?
Full duplex lets both ends of a link send and receive at the same time. Half duplex lets both ends transmit, but only one at a time, so they take turns. Full duplex gives each direction the full link speed and has no collisions.
What is full duplex?
Full duplex is a transmission mode in which two devices send and receive data simultaneously over separate paths. A telephone call is the everyday example. Switched Ethernet links run full duplex, which is why collisions no longer occur on a wired office network.
What is half duplex?
Half duplex is a transmission mode in which both devices can transmit, but not at the same time. A walkie-talkie works this way. Ethernet hubs and Wi-Fi operate in half duplex, with rules for taking turns and handling collisions.
What is simplex vs half duplex vs full duplex?
Simplex sends in one direction only, such as a broadcast. Half duplex sends in both directions, one at a time. Full duplex sends in both directions at once. Almost all modern data links are full duplex, with Wi-Fi the main exception.
Is full duplex faster than half duplex?
Yes, for real traffic. A full duplex link carries its rated speed in each direction at once, while half duplex shares the speed between directions and loses time to collisions and waiting. The difference is largest when data flows both ways.
What is a duplex mismatch?
A link where one end runs full duplex and the other runs half duplex. It usually happens when one side is forced to a fixed setting and the other autonegotiates. The link comes up and passes pings but performs very poorly under load.
How do I fix a duplex mismatch?
Set both ends of the link the same way. The standard fix is to enable autonegotiation on both the switch port and the device. If one end must be fixed, configure the other end with exactly the same speed and duplex.
Should I set my network card to 100 Mbps full duplex?
No. Leave it on autonegotiation unless the other end is also forced to the same value. Forcing only the network card makes the switch fall back to half duplex, which creates a duplex mismatch.
Is Wi-Fi full duplex or half duplex?
Half duplex. On each channel only one device transmits at a time, and the access point and its clients share that airtime. That is one reason wired connections perform more consistently than wireless ones in a busy office.
Is Ethernet full duplex?
Switched Ethernet is. Each switch port connects exactly two devices with separate transmit and receive paths, so both send at once. Ethernet on hubs and shared coaxial cable was half duplex, which is where collisions and CSMA/CD came from.
Does full duplex use CSMA/CD?
No. CSMA/CD exists to handle collisions on a shared, half duplex medium. On a full duplex link collisions cannot happen, so the mechanism is switched off. Some articles say otherwise, but a collision on a full duplex link indicates a fault.
What are late collisions?
Collisions detected after the first 64 bytes of a frame have already been sent. On a properly working network they should not happen. They usually point to a duplex mismatch, seen on the half duplex end, or to a segment that exceeds cable length limits.
Keep readingRelated concepts
Read next · Cabling and connectivity What UTP Cable Is, and Why the Twist Does the Shielding The twisted pair cable itself: categories, pairs and what each one supports. Open this next9 min- Diagnostics · 11 min Packet Loss vs Latency, and Why They Feel the Same to Users How to tell dropped packets from slow ones, and what each one does to calls and transfers.
- Network fundamentals · 10 min Layer 1 Networking, the Physical Layer, and Why It Is Where Most Faults Start What the physical layer covers, from signals and cables to link negotiation.