UDP Throughput MTU Calculator

UDP Throughput, MTU & Payload

UDP is a connectionless transport protocol, and its header is just 8 bytes — much leaner than TCP’s 20-byte minimum. To find the biggest UDP payload that fits in one IP packet without fragmenting, you subtract the headers: UDP_payload = MTU - IP(20) - UDP(8) = 1472 bytes on standard Ethernet (MTU 1500). Bump up to jumbo frames (MTU 9000) and each datagram can carry 8972 bytes.

UDP, unlike TCP, gives you no acknowledgements, no retransmissions, no flow control, and no congestion control. That puts reliability and rate control squarely on the application. The best throughput you can hope for is link bandwidth minus header overhead. On a 100 Mbps Ethernet link, effective UDP goodput lands around 100 × (1472 / 1518) ≈ 96.9 Mbps once you account for the 18-byte Ethernet frame header plus FCS.

Applications

You'll find UDP behind VoIP (RTP), live streaming (HLS-over-UDP, WebRTC), DNS queries, NTP time sync, gaming netcode, SNMP monitoring, and newer protocols like QUIC and HTTP/3. Whenever low latency beats perfect delivery, UDP is the natural pick.

FAQ

Why 1472 and not 1500? 1500 is the link MTU, and that number already includes the 20-byte IP header and the 8-byte UDP header. Take those out and you're left with 1472 bytes of usable UDP payload in each packet.

Does UDP do congestion control? Not on its own. Applications have to build their own, which is exactly what QUIC does with CUBIC/BBR over UDP and what WebRTC does with GCC. Flood UDP carelessly and you can take a whole network down with it.

Can UDP fragment? It can. Send a UDP datagram bigger than the path MTU and IP will split it into fragments. You generally don't want that, since losing one fragment throws away the whole datagram. Keep payload ≤ 1472 bytes, or down to 508 bytes if you need delivery guaranteed across the Internet, per RFC 791.