Elysium vs HyperEVM
–
Independent project. Not affiliated with, endorsed by, or sponsored by Kinetiq or Hyperliquid. Testnet data only.
Loading benchmark data…
Latency
How long until a transaction is in a block. Primary samples in the shared window, in milliseconds from the moment our client started submitting.
Chain-side inclusion p50 per hour
No data yet.
Data
Chain-side inclusion distribution
No data yet.
Data
Every transaction over time
No data yet.
Data
Gas and fees
What the same transaction costs on each chain. Both chains price gas in HYPE, so fees compare directly. Testnet fee parameters, not market prices.
Where the gas goes
No data yet.
Data
Fee per transaction over time
No data yet.
Data
Gas price per hour
No data yet.
Data
Recent transactions
Newest first. Each hash opens the chain's own explorer.
Methodology and caveats
Load. The same MinutePulse contract is deployed on both chains. One client host submits 10 new pulse() transactions per UTC minute to each chain, at the same time, with the same code, fee strategy and polling. This is a fixed offered load, not a throughput test.
Headline latency: chain-side inclusion. The block timestamp of the including block minus the moment our client started submitting. Block timestamps are whole seconds, so each block is placed at the middle of its second (timestamp + 0.5 s). That keeps the median and mean unbiased; each single transaction is only accurate to about ±0.5 s, which is why a transaction whose estimate lands below zero is shown as “< 0.5 s”.
Clock correction. Chain-side inclusion compares the chain's clock with our host's clock, so any error in the host clock would shift every value. The client measures its own clock offset against three public NTP servers (time.google.com, time.cloudflare.com, pool.ntp.org) every 5 minutes, takes the median, and corrects each transaction with the offset measured when it was sent. Samples without a fresh offset are left out of chain-side figures. The correction row in the latency table shows how large the corrections were; NTP measurement itself is accurate to roughly half the network round trip, typically under 100 ms.
RPC accept is the round trip of eth_sendRawTransaction to each chain's public RPC. Observed inclusion is when our client saw the block, polling every 500 ms, so it includes RPC round-trip time and polling delay. The RPC endpoints are in different places, so compare the watcher round-trip row before reading anything into observed inclusion.
Gas and fees. The fee is gasUsed × effectiveGasPrice from each receipt, in HYPE on both chains. Execution gas is the EVM work, and it should match on both chains because the bytecode is identical. Elysium also reports gasUsedForL1, the gas charged for posting the transaction's data to its parent chain; we show it as posting gas. Testnet gas prices are chain parameters: Elysium's base fee is a fixed floor and HyperEVM's testnet base fee moves. Mainnet prices and posting cost at scale can differ.
What "included" means differs. HyperEVM runs on HyperBFT and a block is final when it is produced. Elysium is an Arbitrum Orbit (Nitro) L2 with AnyTrust data availability and a single Conduit sequencer, and it settles to HyperEVM testnet. Inclusion on Elysium is a sequencer confirmation; settlement on the parent chain is not measured here.
Two confirmations are not meaningful on either chain. Elysium produces a block only when there is a transaction, so waiting for two more blocks measures other traffic. HyperEVM is final at inclusion. The 2-conf row is shown only for completeness.
Conditions. Both are public testnets. During sampling, our benchmark was a large share of Elysium's traffic, so Elysium figures describe a lightly loaded chain. HyperEVM carries other traffic, and its public RPC rate limits apply to our client. Kinetiq documents about 300 ms perceived receipts on Elysium.
Samples. Only transactions measured with the current method (version 4) count, and only from the window in which both chains were running. Recovered or rebroadcast transactions are excluded from percentiles and counted separately. The latency difference carries a 95% bootstrap interval (10,000 resamples, fixed seed). No comparison is shown until each chain has at least 200 samples, and results are provisional until the shared window reaches 24 hours.
Privacy. This page does not print the benchmark account, but it is visible on-chain from any linked transaction.
About
An independent, open measurement of how Elysium, a Layer 2 on Hyperliquid, compares with HyperEVM, the Layer 1 it settles to. The goal is a fair, reproducible answer to one question: is the L2 actually faster and cheaper for the same transaction?
Not affiliated with Kinetiq or Hyperliquid. Testnet only; nothing here reflects mainnet performance or prices.