TPS

TPS stands for transactions per second, a throughput measure describing how many transactions a network or application processes during a time interval. It is calculated by dividing transaction count by elapsed seconds. The number is meaningful only when “transaction,” processing stage, workload, and measurement conditions are clearly defined.

A blockchain may count transactions included in blocks, executed successfully, or finalized beyond a chosen threshold. A rollup might report layer 2 user transactions while settling one batch on layer 1. Some systems bundle many actions into one transaction, while others count protocol votes or internal operations. These differences make headline comparisons unreliable without methodology.

TPS matters because limited capacity contributes to congestion, high fees, and delayed execution. Payments, games, exchanges, and social applications need different combinations of throughput and latency. A chain processing many simple transfers may perform poorly with complex smart contracts. Gas or compute per second can provide more context, although it also requires a defined workload.

Peak TPS is not sustained production performance. A laboratory test may use powerful nodes, few geographic regions, empty state, simple transactions, and relaxed decentralization. Real networks face variable message delay, state access, failed transactions, upgrades, attacks, and growing history. Higher block limits can increase TPS while raising hardware requirements and reducing the number of users able to verify independently.

Finality and latency must be separated from throughput. A system can accept thousands of transactions quickly but take longer to provide irreversible settlement. Sequencer confirmation on a rollup is not identical to proof verification and base-layer finality. Network outages, reorganizations, and censorship resistance also affect whether quoted capacity is useful when users actually need it.

When comparing TPS, ask what is counted, whether results are peak or sustained, which hardware and network were used, and when transactions become final. Review fees, reliability, node cost, data growth, and application complexity alongside throughput. Capacity under adversarial spam and recovery after overload provide more useful evidence than an uncontested benchmark. Independent reproduction strengthens performance claims. TPS is a useful engineering measurement when normalized carefully, but it is not a complete score for blockchain scalability, security, or real-world product quality.