Ethereum’s next upgrade turns a 2-second block bottleneck into a roughly 9-second window

Ethereum node operators now have an active public network for rehearsing Glamsterdam’s coordinated execution- and consensus-client upgrade. Platåberget has an open, permissionless validator set, turning the upgrade into a practical test of whether client pairs, validators, builders, and supporting infrastructure continue to work together.

The ethPandaOps-maintained network launched Aug. 13 and is expected to remain available until Glamsterdam reaches mainnet.

For application work, the official page directs decentralized application, smart contract, and other EVM testing to Sepolia. Platåberget focuses on client software, staking setups, builders, and infrastructure.

What Ethereum operators need to test

The Glamsterdam roadmap says node operators will need compatible releases for both sides of an Ethereum node.

The execution client processes transactions, while the consensus client tracks validators and agrees on the chain. Platåberget provides client-pair images and resources for checkpoint syncing and deposits, plus tools for testing builder and fork-choice behavior under load.

The largest architectural change is EIP-7732, or enshrined proposer-builder separation. The Review-stage proposal separates consensus-block validation from execution-payload validation, adds staked builders and creates a Payload Timeliness Committee to report whether a builder revealed its payload on time.

Ethereum says the design expands the roughly two-second critical path: the next proposer gets six seconds to validate a payload, while other validators get nine seconds. Operators need to exercise builder onboarding, payload timing, and fork-choice behavior under the new division of duties alongside basic uptime checks.

A diagram outlines the Glamsterdam operator test, covering two-client upgrades, ePBS validation timing, and key validator testing areas.

Review-stage EIP-7928 adds block-level access lists that record the accounts and storage locations touched by a block. The map lets clients parallelize disk reads, transaction validation, state-root calculation and state reconstruction.

Its immediate value is a dependency map for parallel work, but realized throughput will still depend on client implementation.

EIPs 2780, 8037, and 8038 are listed as scheduled and are being tested for Glamsterdam, all three remain in Review.

Platåberget’s public page gives an 18,000-gas example for a plain ETH transfer. The current EIP-2780 reference case instead totals 21,000 execution gas for a transfer to an existing externally owned account: a 12,000 base, 3,000 cold-recipient access charge, and 6,000 value charge, making a blanket promise of cheaper ordinary transfers premature.

The state-creation proposal charges 183,600 state gas for creating a new account, while the state-access proposal raises a storage write from 2,800 to 10,000. Teams need to recheck hardcoded subcall gas, cached estimates, refund-dependent flows, and workloads that create accounts or storage, including mints, airdrops, and factory contracts.