A detailed guide to blockchain speed | TPS vs. time to finality | Solana, Aptos, Fantom & Avalanche compared — which chain has sub-second finality?
When it comes to blockchain speed, time to finality is more important than transactions per second (TPS). Find out which network has the fastest finality as we compare speed-focused Solana, Aptos, Avalanche, and Fantom.
TPS vs. Time to Finality
Calculating TPS (transactions per second)
What is the fastest blockchain? It all depends on how you define speed. Projects usually advertise their high processing capacity, or “throughput”, measured in transactions per second (tps). This measures the number of transactions that all of the network’s nodes can commit in a second. You can approximately calculate TPS this way:
TPS = (number of tx in a block)/(block time in seconds)
For example, a Bitcoin block contains around 1,700 transactions and takes 10 minutes or 600 seconds to mine. From this, we get a TPS of 1,700/600 = 2.83. Some blocks reach 2,500 transactions, though, so the peak TPS would be 2,500/600 = 4.17
Ethereum churns out a block every 13 seconds, but the gas limit is 30 million gas per block (15m being the target). The minimum gas per transaction is 21,000, so the maximum you can fit is 30,000,000/21,000 = 1428. In reality it’s much fewer, because smart contract interactions require much more gas; the average works out around 11.8 tps.
A blockchain that crams lots of transactions into a single block and commits blocks quickly can have a very high TPS, but that does not necessarily mean faster processing for the end user.
Time to Finality
What really defines a blockchain’s processing speed is time to finality, or the amount of time needed to fully confirm a transaction. In many blockchains, including Bitcoin and Ethereum, when a new Block A is added to the chain, it can still theoretically be reversed, though it takes a lot of resources. Once the next Block B is attached just after A, rolling back A becomes even costlier — and after a few more blocks, it is just economically impractical. At this point you can consider the transactions in Block A final. This is known as probabilistic finality: the probability of a reversal is never zero, but it decays fast as more blocks are added.
The number of blocks required to make sure that a transaction won’t be rolled back is called ‘blocks to finality’. The time needed to create these blocks is known as ‘time to finality’. For Bitcoin with its 10-minute blocks, it’s considered to be 6 blocks and 60 minutes (6*10 minutes). For Ethereum, which has an average block time of 13 seconds, it is also 6 blocks and 13*6=78 seconds.
Time-to-finality (TTF) for a specific transaction depends on when you send it. Let’s say that block time is 15 seconds and it takes just one block after the current one to reach finality. If you’re lucky to send your transaction just 2 seconds before the current block is committed, your payment will be finalized in 2*15=30 seconds.
Apart from probabilistic finality, there is also deterministic finality: a condition where a transaction, once confirmed, cannot be rolled back. In blockchains with deterministic finality, such is Aptos, just one block is enough for an operation to become final, so the time-to-finality can be really short.
Crypto media tends to focus on the high TPS when talking about the speed of different blockchains. But really, what regular users should focus on is the time to finality, because it’s only once a transaction is final that you’ll see the results — be it a deposit on a crypto exchange, a minted NFT, a payment, etc.
For this article, we have selected several blockchains with a TTF below 5 seconds, which can be considered extremely fast. We will compare both their TPS and the time to finality to understand how speed is measured (in theory and in practice), how it is presented to the public, and if speed alone is enough to win big in this industry.
Solana: a 2,000 tps network, but prone to blackouts
Summary: Solana processes an average of 2,000 tps, while the theoretical limit is 710k tps. However, it suffers from frequent outages that have lasted over 24 hours.
Of all major blockchains with active mainnets, Solana has the highest actual TPS, processing 2 thousand transactions per second — more than all the other major networks combined.
The first lab tests on a single node in May 2018 showed an amazing 400,000 tps. In October 2019 — ahead of the testnet launch — the devs ran a series of tests where nodes operated high-end GPUs and got 93k tps at the peak and an average of 50k with 100 nodes.
Next, they did a CPU-only test to check Solana’s minimum capacity, or the throughput that the mainnet was definitely able to deliver. 200 nodes running standard retail-grade CPUs delivered 47,838 transactions per second and an average time to finality of 2.34 seconds.
In early 2020, the testnet went live and soon reached 56,000 tps — and by May 2020, lab tests with 50 nodes showed 111k tps.
We are listing these tests in such detail because they allow us to compare Solana with the blockchains that don’t have a fully functioning mainnet yet, such as Aptos and Fantom (read on to find out how fast they are).
The Solana White Paper states that the theoretical limit for a 1-gigabit (equal to 125 MB per second) network is 710k tps. You can already get ‘gigabit Ethernet’ from providers like AT&T for around $75 a month, but it’s hard to imagine that all of Solana’s 1,700+ nodes will suddenly switch to 1 gigabit networks, so 710k tps will probably remain a purely theoretical limit for quite a while.
Let’s get back to reality, though: how fast is Solana, really? The explorer shows an average of 1,800–2,500 tps.
What about our key value, time to finality? Here, opinions and results vary wildly:
- the White Paper states that 0.5 second is possible;
- Reddit users report that swaps on Radium get confirmed in a couple of seconds, but others say that confirmation takes around 10 seconds;
- Ava Labs’ Kevin Seqniki got 21 to 46 seconds in his own test.
The reason for such differences is probably the fact that Solana’s performance is inconsistent. The network is prone to congestion. Sometimes it simply blackouts for hours on end, causing mayhem and outrage.
Some quote these incidents as the reason SOL’s summer rally ended. Indeed, the first big SOL dump (from $171 to $142) happened in September after the network was down for 17 hours. The reason? The IDO of Grape Protocol, which was overrun by bots trying to buy out all the tokens. The load reached 400,000 tps, and the validators couldn’t process it.
In January 2022, there were 6 more congestion episodes of over 8 hours each. For example, on January 21–22, the outage lasted for 30 hours due to arbitrage bots spamming the network with duplicate transactions, which the validators failed to filter out.
Those who suffered most were the borrowers on Solend who urgently needed to repay loans. Because of the congestion, they couldn’t pay on time and got liquidated (though the platform later reimbursed them for part of the penalty).
One could argue that congestion is a sign of high user demand for this fast and very cheap blockchain. But, these stories also showcase the paradox of incredibly low fees on Solana: it can be worthwhile for trading bots to spam the network, because the profits outweighs the expenditure on fees. If Solana’s developers don’t find a solution, the network risks losing its audience to newcomers such as Aptos.
Aptos: a 160,000 TPS blockchain with sub-second finality
Summary: Aptos is an exceptionally safe and scalable chain capable of 160,000 tps with sub-second finality, currently in testnet.
Back in June 2019, Facebook launched a blockchain project called Libra, later renamed Diem. The original Libra Association was supposed to include 100 members investing at least $10 million each ($1 billion in total). However, faced with scrutiny from the regulators, association members left one by one, and while Facebook eventually piloted a wallet called Novi (designed mostly for remittances), in March 2022 it sold off the Diem assets to the Silvergate crypto bank for $182 million.
The engineers working on Diem created some of the most exciting blockchain tech out there. This includes the Move smart contract language and Move VM (virtual machine). These are far faster, cheaper, and more secure than Solidity and EVM, which are used by Ethereum and other major networks.
Luckily, Move and Move VM are open source, so they are not owned by Silvergate. Anyone can build with Move, and a few of the original Diem devs took advantage of this to create a new blockchain called Aptos. In March 2022, Aptos closed a $200 million funding round led by Andreessen Horowitz, Three Arrows Capital, BlockTower Capital, and many others.
Aptos already has a partnership with Google Cloud that will allow users to deploy nodes in as little as 15 minutes. It’s also working with BNB Chain to offer cross-chain features and BUSD support on Aptos.
Aptos is billed as ‘Layer 1 for everyone’ and ‘the safest and most scalable’ chain in the industry. We’ve already covered its outstanding security and scalability aspects in our previous articles. In a few words, many of the common bugs that plague Ethereum dApps and allow for massive exploits are simply impossible with Move and Aptos.
Here, though, we will focus on speed. Aptos has a number of amazing tricks up its sleeve:
- Optimistic concurrency control: transactions are executed in parallel and then validated; in case of a validation issue, a transaction is aborted and re-executed, and all those that depend on it (higher transactions) are re-validated.
- Collaborative scheduling: a way to schedule transaction execution in a way that no resources are wasted.
- Lazy commit: all the transactions in a block are committed together.
- On-chain reputation: an algorithm that automatically minimizes the negative effect of validators going down.
The bottom line is that parallel execution and the other tools have allowed Aptos to reach speeds of 130,000 tps, with potential speeds as high as 160,000 tps. Even more importantly, the network should have sub-second finality, meaning that a transaction will take less than 1 second to be fully confirmed (compare this to 2.3 seconds on Solana).
A developer in the Aptos Discord even reported that it took him less than an hour with Aptos to build something that would take days with other networks. To learn more about Aptos’ approach to speed, check out this excellent Medium article.
For now, Aptos runs on a testnet, and starting on May 13, node operators and developers will be able to register for the incentivized testnet. The testnet is designed to try out things like staking rewards with test tokens, delegation, transaction costs, voting and governance, and more in order to prepare for the mainnet launch.
The incentivized testnet stage should last until August 2022, and hopefully we’ll see a mainnet soon after. Only then will we find out if Aptos can really deliver 100k+ tps and sub-second transaction finality. However, we can already say that the extra-strong security gives the network a serious advantage over Solana with its outages. The combination of speed and safety can drive adoption among mainstream users.
We at Pontem clearly see the massive potential of Aptos’ technology. That’s why we have become the first team building open-source Aptos-first dApps using Move. Pontem’s years of experience with Move and Move VM put us in a unique position to build foundational tools for Aptos.
Summary: Avalanche was one of the first chains to achieve sub-second finality — but watch out for the gas fee spikes caused by popular play-to-earn games.
Evaluating Avalanche’s speed is a bit complicated, because it includes 3 subnets:
- the X-chain (from ‘exchange’),for issuing and trading assets;
- the P-chain,where the Snowman consensus ‘lives’ and where validators are coordinated;
- the C-chain, where smart contract interactions happen. It is compatible with the Ethereum Virtual Machine (EVM), and if you use MetaMask to work with Avalanche, you’re working with the C-chain.
Each subnet boasts an ‘official’ throughput of up to 4,500 TPS. This may not seem like much compared to Solana’s 100k+ TPS, but where Avalanche really shines is the time to finality.
This is possible thanks to the unusual consensus mechanism that the C-Chain uses. Validators run 25 rounds of ‘polls’ among themselves to see if other nodes agree with their decision to accept or reject certain transactions. Essentially it’s the wisdom of the crowd in action. This requires very little time or computation, and transactions are finalized instantly, without the need for any confirmations.
The sub-second results were obtained under optimal testing conditions. But Avalanche also performs well in real life: developer Kevin Siqniki, whom we cited above, got 1.3–3.4 seconds in his independent test.
In addition to the 3 chains we’ve discussed, any node can create a new subnet, thus taking the load off the C-Chain. In the past few months, this has become an urgent issue, as the C-Chain has been suffering from congestion and high fees. The main reason? Play-to-earn games.
Avalanche is a hub for NFT-games like Crabada, Yield Hunt, Play2Moon, and others. As thousands of players stake and unstake NFTs and collect rewards, the network gets clogged, and gas fees rise. What used to be a few cents last summer increased to $3–5 and even $14 on some occasions. Some DeFi projects on Avalanche have also caused gas fee spikes in the past, including Hatter and Wonderland.
Crabada, the main culprit of the high gas fees on Avalanche. Credit: Crabada Twitter
Crabada alone accounts for 16% of all the gas spent on the network. On May 4, the project announced that it’s finally moving to a subnet of its own. Hopefully, this will bring some relief to Avalanche users.
Fantom: the only top-10 chain with 1 second finality
Summary: Fantom delivers an average time to finality of just one second and consistently low fees, though it lost some of its popularity after star developer Andre Cronje quit DeFi.
Of all the blockchains in the top 10 by DeFi TVL, only Fantom and Avalanche can deliver sub-2 seconds finality. With the introduction of the Go-Opera network upgrade in April 2021, Fantom’s average time to finality was reduced to 1 second.
However, the 300k tps claims have apparently been dropped, and the official website now simply states ‘thousands’. The site also features an article about why TTF is a much more important metric than TPS. On Twitter, you’ll find wildly varying estimates, from 1,500 tps to 80,000 tps. The highest actual average value over 24 hours was around 20.8 tps in September 2021 (1.8 million transactions in 24 hours).
Apart from the fast finality time, Fantom’s key advantage is low fees. If you were to choose where to allocate DeFi capital based on the network fees, Fantom would win over Avalanche, as even during the largest fee spikes (e.g. in November 2021) a transaction would cost less than $1.50. The average fee is below 10 cents.
A possible reason is that Fantom hasn’t become a hub for NFT games, which consume lots of gas due to the complexity of transactions with non-fungible assets. Instead, Fantom is home to multiple forks of Tomb Finance, where people also claim and re-stake rewards daily (like in P2E games) — but with no NFTs involved, gas consumption remains reasonable.
In Q4 2021 and at the start of 2022, Fantom suddenly became the focus of media attention and even reached no.3 by TVL, below Ethereum and Solana but ahead of BSC. The number of unique addresses soared as users flocked to Fantom. The rally was partly fueled by the hype around star developer Andre Cronje’s Solidly project.
Unfortunately, Fantom was dealt a blow when Cronje announced he was quitting DeFi. The network lost 70% of the TVL between late January and May, and the Solidly saga basically ended in a flop. Still, this doesn’t detract from Fantom’s strength as the fastest of the major chains.
Beware of ‘the fastest ever’ blockchains
Dubious projects sometimes lure in liquidity by advertising “lightning-fast” processing speeds without providing any evidence. A very telling example is the recent aggressive (suspiciously aggressive, actually) marketing campaign for a chain called BitGert.
If you google ‘the fastest blockchain’, you’ll see a barrage of news articles claiming that it’s ‘the fastest in the industry’ at 100k tps. However, BitGert’s block time is 15 seconds, so the average time to finality would be at least 7 seconds.
By the way, BitGert’s creators were found to have generated their pictures for the Team page using AI. Plus, in spite of the advertised $0.000000001 transaction fee, every transfer of $BRISE is slapped with a 12% fee, part of which is used for marketing (thus the constant stream of paid articles). Interestingly, the only address that doesn’t pay the 12% fee is the ‘owner address’ hard-coded into the contract!
After BitGert repeatedly got called out for these and other shenanigans, the same paid news websites began promoting another project, called TechPay Coin. This reportedly has 300,000 TPS and is ‘the fastest blockchain ever’. It even advertises a time to finality of just 1.15 seconds and claims to be ‘fully compatible’ with both EVM and Cosmos SDK (no mean feat).
However, the only information about TechPay’s smart contracts and blocks comes from the project’s website. There is no detailed documentation, the team is anonymous, and the blog on Medium features posts like ‘Why cryptocurrency is so important’.
These examples show that users should take any claims of sky-high TPS with caution. A bona fide network explorer, results of lab tests, independent assessment by analysts — all this is needed before one calls a network ‘the fastest’.
Pontem is a product studio working toward global financial inclusion powered by blockchains. We are partnered with Aptos to build foundational dApps, development tools,, AMMs, and more.The Move IntelliJ IDE is just the beginning.”