Coin Brief ENDE

Simulation: EIP-7999 fee markets deliver more execution than one-dimensional designs

A paper summarised on the Ethereum Research forum on 12 November compares two ways of pricing Ethereum blockspace after the Glamsterdam upgrade: a single base fee for all resources, and the multidimensional fee market proposed in EIP-7999. The author, Fei Wu, did part of the work during an internship at the Ethereum Foundation, and the post says the results can be reproduced from a published repository.

The study starts from what it calls the normalised benchmark, the strongest one-dimensional design from earlier posts. It adjusts the transaction floor price to each propagation allocation and uses EIP-8372 to scale the cost per state byte and the state gas limit by a common factor, so that regular gas and state gas hit their targets at the same base fee. With EIP-1559's target-to-limit ratio of one half, that design delivers 151.9 to 177.9 million execution gas per block in simulation, but state growth reaches only 86.0 to 88.1 GiB a year against a 120 GiB target. Congestion is measured as the share of blocks that use at least 98% of a hard limit.

Raising the target ratio to four fifths increases execution to 199.0 to 215.0 million gas, but state growth falls to 56.7 to 68.7 GiB a year and near-limit blocks rise from about 4% to between 13.2% and 20.6%. The paper traces this to the EIP-1559 update rule. A higher target clears at a base fee of only 2.4 to 7.2 wei, but under the integer arithmetic of EIP-1559 the fee cannot be lowered once it is below 8 wei, while every block above target raises it by at least one wei. In the simulations the fee averaged 18.3 to 27.2 wei, four to eight times its equilibrium. State, as the most price-sensitive resource, shrinks the most.

Swapping in the exponential update rule used for blobs in EIP-4844 lets fees fall and lifts execution to 214.8 to 244.2 million gas with state growth of 94.0 to 96.0 GiB, but 38.8% to 40.8% of blocks then sit near a limit. Against these variants, conservative EIP-7999 configurations deliver 4% to 34% more execution than the one-half benchmark at a similarly low congestion level, with state growth at its target, and maximum-throughput settings deliver 19.5 to 56.2 million more gas than the EIP-4844-style design with fewer near-limit blocks.

Simulation: EIP-7999 fee markets deliver more execution than one-dimensional designs
Simulation: EIP-7999 fee markets deliver more execution than one-dimensional designs — Coin Brief

Why it matters

The debate over whether Ethereum needs separate prices for computation, state and data has often been argued in principle. This paper puts numbers on the trade-off and identifies a concrete mechanical limit in the current update rule. The results are simulations under modelled demand, and they come from one research group, so they are an input to the EIP discussion rather than a verdict.

Written by Victoria Shinder.