Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Transactions per second (TPS) measures blockchain throughput: the number of transactions a network processes during one second. A meaningful TPS figure must state which transactions were counted, the measurement interval, and the test or operating conditions. Valid transactions committed by the network are generally more useful than raw submission attempts.
TPS describes volume over time, not the waiting time for one transaction. To judge performance, read TPS alongside transaction latency, confirmation or finality, workload, security, and decentralization. A larger TPS number by itself does not prove that users get faster, safer, or cheaper transactions.
As an Amazon Associate I earn from qualifying purchases.
What TPS measures
The basic calculation is:
TPS = transactions counted ÷ elapsed seconds
If a system commits 18,000 qualifying transactions during a 60-second interval, its measured throughput for that interval is 300 TPS. That arithmetic is simple; deciding what counts as a transaction is the difficult and essential part.
ITU-T Recommendation F.751.1 (2020) calls TPS a standard performance indicator and defines it using processed transaction count and processing time. The recommendation also says that the topology and test environment should be reported. Without those details, a number cannot be interpreted or reproduced responsibly.
#1 Best Overall
Hyperledger’s performance-metrics definition is more specific: “Transaction throughput is the rate at which valid transactions are committed by the blockchain SUT in a defined time period.” In that approach, invalid transactions are subtracted rather than counted as successful work. The measurement is taken across the system’s committed nodes at a stated network size.
Throughput is a rate, not a total
A chain that processes 100,000 transactions in ten minutes has an average of about 166.7 TPS during that window. The total transaction count and the rate answer different questions. A short burst can produce a high average while the network is quiet before and after it; a longer average can hide spikes and congestion.
Every TPS claim needs a counting rule
Reports may count submitted transactions, attempted transactions, successful transactions, valid transactions, committed transactions, or finalized transactions. Those categories are not interchangeable. A benchmark that counts every request sent to nodes can report a larger number than one that counts only valid state changes committed by the network.
When reading a figure, identify the exact rule: Are failed or invalid transactions excluded? Does “processed” mean included in a block, accepted by a consensus layer, or finalized under the protocol’s strongest assurance? Is the figure for one base layer or for several layers combined?
Why TPS is not the same as transaction speed
TPS answers, “How much work can the system complete per unit of time under this definition?” Latency answers, “How long did a particular transaction take to reach a specified network state?” Finality answers, “How strongly does the protocol treat that result as irreversible?” These are related but separate measurements.
| Metric | What it measures | Why it matters |
|---|---|---|
| Throughput (TPS) | Qualifying transactions completed during a time interval | Capacity for sustained or burst workload |
| Latency | Elapsed time from submission to a stated confirmation threshold | How long a user waits for a usable result |
| Confirmation | Progress toward the protocol’s accepted inclusion or agreement state | Whether an application can act on the transaction |
| Finality | The protocol’s strongest stated point of irreversibility | How much reversal risk remains under the protocol model |
Hyperledger describes transaction latency as the time from submission until the effect is usable across the network at a specified confirmation threshold. That interval can include propagation and consensus settling. A high-throughput system can therefore still have a long wait for inclusion or finality when demand exceeds available capacity.
Confirmation and finality can occur at different times
Solana documentation treats confirmation time and finality as distinct terms. A transaction may be observed as confirmed before it reaches the protocol’s strongest final state. Applications should define which state they require instead of using “confirmed” as a universal synonym for irreversible.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Ethereum’s transaction lifecycle provides another example. A transaction is broadcast and waits for a validator to include it in a block. The containing block can later become justified and then finalized. Inclusion is an earlier event than finality, so a TPS figure based on included transactions does not tell you how long finality takes.
A block interval is not a TPS number
Ethereum’s documentation says blocks are normally created and committed once every 12 seconds, with slots 12 seconds apart. That is a block timing statement, not a throughput result. The number of transactions in each block depends on the resource use of those transactions, so multiplying a block interval by an assumed transaction count would not produce a reliable TPS claim without a stated workload.
Transactions are not identical units of work
A simple transfer and a smart-contract or program interaction can consume very different amounts of computation, storage access, bandwidth, and consensus resources. Counting them as identical units can make two workloads look comparable when they are not.
Rank #3
Resource limits constrain transactions per block
Ethereum documentation describes state-changing operations and gas, while its block documentation describes gas-bounded blocks. A block can hold many low-gas transfers or far fewer high-gas contract calls. The resulting TPS therefore changes with the mix of transactions even when the block schedule is unchanged.
Workload details belong next to the number
A useful benchmark identifies transaction type, payload size, read and write behavior, signature or validation requirements, and the proportion of successful versus failed calls. If a test uses only minimal transfers, its result should not be presented as the expected rate for a decentralized application whose calls consume much more block capacity.
Why blockchain TPS numbers differ
Different counting rules
One report may count every submitted request; another may count only valid transactions committed by consensus; a third may count finalized transactions. Failed or invalid requests can inflate a headline number when they are included. Hyperledger’s definition specifically calls for invalid transactions to be subtracted from valid throughput.
Different workloads
Transfer-heavy tests, contract execution, token minting, and storage-intensive operations place different demands on nodes. A network can show a high TPS result for a lightweight transfer test and a much lower result for a computation-heavy application without either measurement being erroneous.
Different measurement windows and demand
A controlled test may drive the network continuously until it reaches a limit. A live-network figure may average actual activity during a quiet or busy period. The former is a capacity result under stated conditions; the latter is observed traffic. They should not be compared as if they represented the same thing.
Free tools Windows power users keep installed
One-click scans. No signup required.
Rank #4
Different test environments
Hardware, node count, geographic distribution, network topology, software version, consensus configuration, and workload generation all affect results. ITU-T and Hyperledger both call for these conditions to be documented. A result from a small, centrally located test network cannot automatically describe a geographically distributed production network with many independent operators.
Different layer boundaries
A base-layer TPS figure counts work settled directly by that chain. An ecosystem total may add transactions handled by one or more layer-2 systems. Ethereum.org identifies rollups as a primary scaling approach and distinguishes on-chain from off-chain scaling. A base-layer result and a multi-layer aggregate count different work and must be labeled accordingly.
How to evaluate a TPS claim
Use this checklist before treating a number as evidence of performance:
- Identify the result type. Determine whether it is a theoretical ceiling, a controlled benchmark, or observed live-network throughput.
- Read the counting definition. Check whether the result counts submitted, attempted, valid, committed, confirmed, or finalized transactions.
- Inspect the workload. Find out whether the test used simple transfers, contract calls, large payloads, or a mixture.
- Check the interval. Note the duration, whether the value is an average or peak, and whether the network was continuously saturated.
- Locate the layer boundary. Establish whether the figure covers a single base layer or aggregates additional execution layers.
- Review the environment. For tests, record hardware, software, node count, geography, topology, and consensus settings.
- Pair throughput with latency and finality. Ask how long transactions took to reach the confirmation or finality threshold that matters to the application.
- Assess the trade-offs. Consider how the design affects security assumptions, validator participation, decentralization, fees, and operational reliability. TPS alone cannot establish those properties.
Comparing two blockchains without misleading yourself
Put both systems on the same measurement basis before comparing their numbers. The following axes identify the minimum information needed.
| Comparison axis | Question to answer | Reason |
|---|---|---|
| Result category | Is each number observed traffic or a controlled or theoretical maximum? | Capacity and real demand are different measurements |
| Transaction workload | Are the operations and resource demands equivalent? | One transaction may represent far more work than another |
| Validity rule | Are failed or invalid transactions excluded? | Counting unsuccessful requests can inflate throughput |
| Time window | What interval and averaging method were used? | Peaks, bursts, and long-run averages are not interchangeable |
| Network and layer scope | Does the result cover one chain or an aggregate across layers? | Different boundaries count different work |
| Test conditions | What hardware, topology, node population, and software were used? | Environment can materially change a benchmark |
| User-facing performance | What are inclusion latency and finality time? | Users experience waiting time, not TPS directly |
If a source does not disclose enough information to align these axes, report the figures as non-comparable rather than ranking the networks. The absence of a shared methodology is itself a limitation of the comparison.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Worked TPS examples
Example: valid committed throughput
Suppose a test runs for 120 seconds and the blockchain commits 24,000 valid transactions. The measured throughput is 24,000 ÷ 120 = 200 TPS. If 2,000 additional requests were rejected as invalid, they do not belong in the numerator under the valid-committed definition.
Example: submission rate versus committed rate
An application might submit 500 requests per second while the network commits only 350 valid transactions per second. The submission rate describes demand placed on the system; the committed rate describes completed work. Calling both figures “TPS” without labels would obscure the backlog and the network’s actual throughput.
Example: latency despite steady throughput
A network can sustain 200 committed TPS while a particular transaction waits through several blocks before inclusion. The system is processing substantial volume, but that does not reveal the transaction’s submission-to-confirmation time. A separate latency measurement must record the transaction’s timestamps and the confirmation threshold used.
Recommended Free Tools
What TPS can—and cannot—tell you about a blockchain
What it can tell you
- How many transactions were completed per second under a stated definition.
- Whether a workload is approaching a measured capacity limit.
- How much aggregate transaction volume a design handled in a specified test or live interval.
What it cannot establish by itself
- How quickly an individual transaction will be included or finalized.
- Whether a transaction is cheap, because fees depend on demand, resource pricing, and application behavior.
- Whether the network is secure or decentralized. Those properties depend on protocol design, validator requirements, adversary assumptions, and participation—not on a single throughput number.
- Whether an application will perform well, because its workload may differ from the benchmark.
- Whether a multi-layer aggregate represents the same work as a base-layer measurement.
Choosing the right performance question
Different users need different measurements. An exchange or payment service may require sustained valid throughput plus a predictable confirmation target. A trading application may care more about inclusion and finality latency during bursts. A smart-contract platform should examine throughput for representative contract calls, not only simple transfers. A researcher comparing protocols needs the full workload and environment description before drawing conclusions.
For capacity planning, measure demand and committed throughput over a window long enough to reveal backlog growth. For user-experience analysis, collect submission-to-confirmation and submission-to-finality times. For protocol evaluation, preserve the security and decentralization context alongside both metrics. The same TPS value can be adequate for one use case and insufficient for another.
Common mistakes and their corrections
- Mistake: “Higher TPS means faster transactions.” Correction: TPS is throughput; measure latency and finality for waiting time and irreversibility.
- Mistake: Treating a headline maximum as normal production activity. Correction: Label theoretical, controlled-test, and observed figures separately.
- Mistake: Comparing simple transfers with complex contract calls as equal work. Correction: Match the workload or explain the resource difference.
- Mistake: Counting failed submissions as successful throughput. Correction: State the validity rule and subtract invalid transactions when using a valid-commit definition.
- Mistake: Adding layer-2 activity to base-layer TPS without explanation. Correction: State the layer boundary and what each number includes.
- Mistake: Using TPS as a proxy for decentralization, security, fees, or overall quality. Correction: Evaluate those properties with their own evidence and assumptions.
Bottom line
TPS is a throughput measurement: qualifying transactions completed divided by elapsed seconds. It becomes useful only when the report defines the counted transactions, workload, time window, network or layer scope, and test conditions. Read it with latency and finality, then consider security and decentralization trade-offs. A high TPS figure is evidence of high measured volume under particular conditions—not a complete verdict on blockchain performance.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
