EIP-7923: A Sparse-Page-Based Memory Model for the EVM
2026-11-03 –, Lightning Stage

The EVM’s quadratic memory pricing caps usable memory at roughly 3MB per transaction. EIP-7923 replaces it with a sparse-page, linear-cost model enabling 64MB per transaction — implemented via POSIX mmap and benchmarked on real hardware.


EIP-7923 enables 64MB of memory allocation per transaction through a sparse-page-based memory model with linearized gas costs. Today, the EVM’s quadratic memory pricing — designed to prevent DoS attacks — limits memory usage to about 3MB per transaction and lacks modern memory management features like stack and heap allocation, resulting in real inefficiencies for compute-heavy contracts. EIP-7923 replaces quadratic pricing with a linear model, unlocking up to 64MB of memory per transaction. We explain how this is achieved through a virtual memory system implemented using POSIX mmap, present gas-cost measurements from real-world hardware benchmarks, and justify the 64MB cap as a security-critical balance that prevents DoS risk, particularly in concurrent eth_call scenarios.

Qi Zhou is a prominent Ethereum researcher, engineer, and protocol contributor with deep expertise in Data Availability, Layer 2 infrastructure, and decentralized storage. He is an active member of the Ethereum community, regularly contributing to Ethereum Magicians and ethresear.ch, and has authored or co-authored multiple Ethereum Improvement Proposals (EIPs), including EIP-4804, EIP-6150, EIP-7650, EIP-7907, and EIP-7923.