2025-09-12 –, Ethreactor Stage
Language: English
This talk presents EIP-7923, which introduces a sparse-page memory model with linear gas pricing—enabling up to 64MB memory per transaction. It modernizes EVM memory management while maintaining DoS resistance.
The EVM’s current quadratic memory pricing limits smart contracts to about 3MB of usable memory per transaction, making it hard to support data-heavy use cases. In this talk, we introduce EIP-7923, which proposes a sparse-page-based memory model with linearized gas costs. This enables up to 64MB of memory per transaction and introduces modern memory features more in line with conventional computing systems.
We’ll walk through the virtual memory design based on POSIX mmap, show how memory pages are lazily allocated, and demonstrate gas cost benchmarks based on real hardware. The new model enables advanced applications such as AI inference, high-throughput simulations, or large Merkle trees—while preserving security through a conservative 64MB cap and safe concurrent eth_call handling. This upgrade modernizes the EVM and expands its capability without compromising network stability.
Founder of QuarkChain
Dr. Qi Zhou is the founder of QuarkChain and EthStorage. As an active Ethereum ecosystem contributor, he authors/co-authors several EIPs/ERCs including ERC-4804/EIP-7902/etc. His interest is in researching and developing the future Web3 infrastructure, including data availability, layer 2, decentralized storage, and token economics. He has received multiple grants from EF/OP for data availability sampling, proof of L2 storage, and instant fault proofs.