Protocol Overhead Calculator
Work out how much of a link is actually payload once Ethernet framing, IP, TCP or UDP and any tunnel headers are accounted for.
Result
Breakdown
| Layer | Bytes |
|---|---|
| Application payload | 1460 |
| TCP header | 20 |
| IPv4 header | 20 |
| Ethernet header + FCS | 18 |
| Preamble + inter-frame gap | 20 |
| Total on the wire | 1538 |
Efficiency by payload size
| Payload | Efficiency | Goodput | Packets/s |
|---|---|---|---|
| 64 B | 45.1% | 450.7 Mbit/s | 880,282 |
| 128 B | 62.1% | 621.4 Mbit/s | 606,796 |
| 256 B | 76.6% | 766.5 Mbit/s | 374,251 |
| 512 B | 86.8% | 867.8 Mbit/s | 211,864 |
| 1024 B | 92.9% | 929.2 Mbit/s | 113,430 |
| 1460 B | 94.9% | 949.3 Mbit/s | 81,274 |
| 8960 B | 99.1% | 991.4 Mbit/s | 13,830 |
About Protocol Overhead Calculator
Line rate is not payload rate. Every packet carries transport, network and Ethernet headers, plus 20 bytes of preamble and inter-frame gap that never appear in a packet capture but still occupy the wire.
When jumbo frames actually help
Raising the MTU to 9000 bytes lifts efficiency from about 95% to over 99% and cuts the packet rate by six, which matters for storage replication and backup traffic. It only works if every device in the path agrees: one switch left at 1500 bytes silently drops oversized frames, and because ICMP is often filtered the symptom is a hang rather than an error. Jumbo frames are a data-centre and dedicated-link technique, not something to enable across the internet.
Why packets per second matters more than bits
Forwarding hardware and CPUs process packets, not bytes. A 10 Gbit/s link carries about 812,000 full-size packets per second but over 14.8 million minimum-size ones - which is why vendors quote small-packet forwarding rates and why a DDoS of tiny packets can exhaust a router long before it saturates the link.
Common use cases
- Explaining why a link shows 940 Mbit/s throughput on a gigabit port.
- Sizing VoIP capacity, where overhead exceeds payload.
- Checking whether a device can handle the packet rate, not just the bit rate.
Edge cases and gotchas
- Preamble and inter-frame gap are on the wire but invisible to tcpdump - captures always understate utilisation.
- Jumbo frames improve efficiency but only if every device in the path agrees on the MTU.