For years, transactions per second (TPS) has been the go-to metric for blockchain developers to promote their new networks. It's often used to showcase superiority over competitors. However, as blockchain technology evolves, many industry leaders are questioning whether TPS remains a meaningful measure of network performance.
This article explores the relevance of TPS in today's blockchain ecosystem, examining its limitations and potential alternatives that might better represent true network capabilities.
The Evolution and Limitations of TPS
In cryptocurrency's early days, TPS was indeed crucial. Blockchains like Bitcoin and Litecoin primarily handled simple value transfers between addresses. During this period, TPS helped users understand how different technical choices—such as block size or cryptographic algorithms—affected transaction processing speed.
The scaling debate largely centered around how many transactions a blockchain could handle, which was vital if cryptocurrencies were to become global digital currencies. When Ethereum launched in 2015 with programmable smart contracts and approximately 13 TPS, many operations became far more computationally intensive than simple transfers.
Account Abstraction and Transaction Bundling
Years later, account abstraction emerged, allowing users to group multiple operations (called User Operations or UserOps) through transaction bundling. This development further exposed TPS's growing blind spots.
As Steven Goldfeder, co-founder of Offchain Labs, explains: "Focusing on raw TPS numbers is somewhat like counting the number of bills in your wallet while ignoring that some are singles, some are twenties, and some are hundreds."
Some transactions carry significantly more computational value than others, yet TPS treats them all equally.
The Solana TPS Debate: Exaggeration or Misunderstanding?
Solana has faced particular scrutiny regarding its TPS claims. The network promotes benchmark figures of 65,000 TPS, while its whitepaper suggests a theoretical maximum of 710,000 TPS with a 1 GB network connection. Meanwhile, solana.com currently displays real-time transactions closer to 3,000 TPS.
Critics argue that 80-90% of these transactions consist of non-user activity, primarily consensus votes. Austin Federa, Strategic Lead at Solana, defends these figures: "Votes are real transactions on Solana that pay fees. If people don't want to count votes, let's look at the 'real TPS' numbers."
Solana Compass, a staking tool provider, currently lists the network's "real TPS" (excluding consensus votes) at approximately 704. Federa notes this figure is still "10 times what most recent blockchains can do."
The Complexity of Modern Transactions
Federa highlights another challenge with TPS metrics: "Solana is now probably at least five times faster than when I joined. But you don't see it because transaction complexity has increased dramatically."
Simple transactions like consensus votes or sending SOL have low computational costs, while arbitrage trades or NFT minting can be "100 times more computationally intensive." TPS metrics treat both types identically, despite their vastly different resource requirements.
Better Alternatives to TPS
Given TPS's limitations, several alternative metrics have emerged that might better represent blockchain performance.
User Operations Per Second (UOPS)
Anthony Rose, Senior Vice President of Technology at Matter Labs (developers of zkSync), suggests User Operations per second (UOPS) might be a more meaningful metric. Ethereum's account abstraction standard (ERC-4337) introduces "pseudo-transaction objects" called UserOperations that function like instructions telling smart accounts what actions to perform on a user's behalf.
For example, a simple token swap on a decentralized exchange typically requires two separate transactions. With account abstraction, these UserOps batch together, yet TPS still counts them as a single transaction.
"As we see more account abstraction adoption and improved user experiences, TPS becomes increasingly less useful as a metric," Rose explains.
However, UOPS has its own limitations. The metric doesn't apply well outside the Ethereum ecosystem, and Federa argues it "rewards overly complex products" similarly to "counting website visitors based on the number of HTTP requests."
Gas Per Second (GPS)
Avihu Levy, Chief Product Officer at StarkWare, suggests: "We will all evolve to some kind of gas per second."
Users pay gas to process smart contracts on Ethereum, making this metric a measure of computational work. Gas per second (GPS) would consider transaction size and type, plus block size and block time, to measure throughput. Levy believes this comes closest to measuring network performance in terms of computational output or resource consumption.
The challenge is that different networks measure computation differently. Ethereum uses gas, Starknet uses Cairo steps, Solana uses "compute units," and Aptos uses gas units. Creating a "canonical benchmark" to properly convert these various computation measures presents significant difficulties.
The Future of Blockchain Performance Metrics
Despite its flaws, TPS likely isn't disappearing soon. Rose acknowledges that while TPS is "a flawed metric with no strong rebuttal," it remains what "many people rely on."
He notes: "There's a trade-off between interpretability, understandability, and metric quality. People understand TPS in ways they don't yet grasp more meaningful metrics."
Davies suggests TPS will maintain its prominence until blockchains move beyond simple "value storage" to more sophisticated applications: "If you're speculating in blockchains, TPS only has value if it helps 'sell' your 'investment'."
Federa envisions a future where TPS becomes irrelevant to most users, similar to how computer specifications concern only developers today: "The success of the PC industry is that you don't have to care too much about what computer is in front of you."
He adds: "We haven't achieved this with blockchains yet, but I truly hope we get to where all blockchains are fast enough that only developers need to worry about this."
👉 Explore advanced blockchain metrics
Frequently Asked Questions
What does TPS mean in blockchain?
TPS stands for "transactions per second," a metric that measures how many transactions a blockchain can process in one second. It has traditionally been used to compare blockchain performance, though its relevance is increasingly questioned as transaction types diversify.
Why is TPS considered a flawed metric?
TPS treats all transactions equally regardless of complexity. Simple value transfers require minimal computation, while smart contract interactions or NFT minting might be 100 times more computationally intensive. This makes raw TPS numbers misleading for comparing network capabilities.
What are the main alternatives to TPS?
User Operations per second (UOPS) and Gas per second (GPS) are emerging alternatives. UOPS measures user actions rather than transactions, while GPS measures computational work performed. Both attempt to provide more meaningful performance indicators than simple transaction counts.
How does account abstraction affect TPS relevance?
Account abstraction allows multiple user operations to be bundled into single transactions. This means a blockchain might show low TPS while actually processing many user actions, making TPS increasingly inadequate for measuring true user-facing performance.
Do different blockchains measure TPS differently?
Yes, measurement methodologies vary significantly. Some networks include consensus-related transactions in their counts, while others exclude them. Some report theoretical maximums from ideal conditions, while others report real-world averages, making direct comparisons problematic.
Will TPS remain important in the future?
While TPS will likely remain a commonly cited metric due to its simplicity, most experts believe more sophisticated measures will gain prominence as blockchain technology matures and transaction types continue to diversify.