TCP and UDP are both transport-layer protocols that move data across the internet, but they work in fundamentally different ways. TCP (Transmission Control Protocol) is a connection-based protocol that guarantees reliable, ordered delivery of data. UDP (User Datagram Protocol) is connectionless and prioritizes speed over reliability, with no guarantee that data arrives at all. Understanding what’s true about each protocol comes down to how they handle connections, error checking, speed, and overhead.
TCP Establishes a Connection Before Sending Data
TCP requires a formal handshake between two devices before any data is exchanged. This is called the three-way handshake, and it works in three steps. First, the client sends a SYN (synchronize) packet to the server, requesting a connection and sharing its initial sequence number. Second, the server responds with a SYN-ACK packet, acknowledging the client’s request while also sending its own synchronization request. Third, the client sends a final ACK packet confirming the server’s request, and the connection is established.
This process ensures both devices agree on sequence numbers so they can track every piece of data sent between them. If a packet gets lost or arrives out of order, TCP detects the problem and retransmits the missing data. That makes TCP a “reliable” protocol in networking terms: it guarantees that data arrives complete and in the correct order.
UDP skips all of this. It sends data (called datagrams) directly to the recipient with no prior setup, no acknowledgment, and no tracking. There’s no notification if a packet fails to arrive, no mechanism to reorder packets that show up out of sequence, and no protection against duplicate messages. The protocol simply makes a best-effort attempt at delivery.
TCP Controls the Flow of Data
One of TCP’s key features is flow control, which prevents the sender from overwhelming the receiver with more data than it can process. The receiver advertises a “window size” that tells the sender how much free buffer space it has available. The sender is limited to having no more than that amount of unacknowledged data in transit at any time. As the receiver processes data and frees up buffer space, it updates the window size, and the sender adjusts accordingly.
TCP also manages congestion on the network itself. Each connection tracks a congestion window, a variable that limits how fast data can be sent based on network conditions. The actual sending rate is capped at whichever is smaller: the receiver’s advertised window or the congestion window. This means TCP automatically slows down when the network is congested and speeds up when conditions improve.
UDP has neither of these mechanisms. The sender transmits data at whatever rate it chooses, regardless of whether the receiver can keep up or whether the network is congested. This lack of flow control is part of what makes UDP faster, but it also means applications using UDP need to handle congestion and packet loss on their own if they care about those things.
UDP Has Far Less Overhead
Every packet sent over a network includes a header containing protocol information. TCP headers are a minimum of 20 bytes and can grow up to 60 bytes with optional fields. UDP headers are only 8 bytes per packet. That difference matters when you’re sending millions of small packets per second, as gaming servers or streaming platforms often do.
The smaller header isn’t the only source of UDP’s efficiency. TCP also generates extra traffic through its acknowledgment system: the receiver must confirm receipt of data, and the sender must track and potentially retransmit anything that goes unacknowledged. All of that adds latency and bandwidth overhead that UDP simply avoids.
Each Protocol Suits Different Applications
TCP is the right choice when data integrity matters more than speed. Web browsing, file downloads, and email all use TCP because missing or out-of-order data would corrupt a webpage, break a file, or scramble a message. When you load a website, every byte of HTML, CSS, and JavaScript needs to arrive correctly for the page to work.
UDP dominates in real-time applications where a slight loss of data is acceptable but delay is not. Video conferencing tools like Zoom, VoIP phone calls, online multiplayer games, and live video streaming all rely on UDP. In a video call, if a single packet of audio is lost, you might hear a tiny glitch, but the conversation keeps moving. With TCP, the protocol would pause the stream to request retransmission of that lost packet, creating a noticeable lag that feels far worse than a momentary audio blip. The same logic applies to online gaming, where a fraction of a second of input delay can make the experience feel sluggish and unresponsive.
UDP Is Powering Newer Web Protocols
TCP has a long-standing limitation called head-of-line blocking. When multiple streams of data share a single TCP connection (as they do in HTTP/2), a single lost packet holds up everything behind it until it’s retransmitted. This can cause visible stalls when loading a complex webpage.
HTTP/3, the latest version of the protocol that powers the web, solves this by replacing TCP with a protocol called QUIC, which is built on top of UDP. QUIC adds its own reliability and encryption layer while avoiding head-of-line blocking entirely. If one packet is lost, only the specific data stream it belongs to is affected, not the entire connection. QUIC also enables faster connection setup (sometimes with zero round trips for returning visitors) and can survive network changes, like switching from Wi-Fi to cellular data, without dropping the connection. This is a significant shift: even web traffic, long the domain of TCP, is increasingly riding on UDP as a foundation.
Quick Comparison
- Connection: TCP requires a three-way handshake; UDP sends data immediately with no setup.
- Reliability: TCP guarantees delivery and correct ordering; UDP offers no delivery guarantees.
- Flow control: TCP adjusts sending speed based on receiver capacity and network congestion; UDP has no built-in flow control.
- Header size: TCP headers are 20 to 60 bytes; UDP headers are 8 bytes.
- Speed: UDP is faster due to less overhead and no acknowledgment delays; TCP is slower but more dependable.
- Use cases: TCP for web pages, file transfers, and email; UDP for gaming, video calls, and live streaming.