🔥 Buterin: cache-first tables beat big RAM for crypto speed
posted 5 Oct 2025
Ethereum co-founder Vitalik Buterin published a technical post on Oct. 5, 2025, stating that memory access time grows with the cube root of memory size rather than remaining constant. The post appeared on his personal site at vitalik.eth.limo.
Buterin's research shows memory access latency scales as O(N^1/3) instead of O(1). He presented a physical model explaining why access slows as memory grows and included measurements across cache, RAM and other memory tiers. The measurements matched the cube-root scaling pattern.
The post examines how this affects cryptographic engineering. Developers often create large precomputed tables for operations like elliptic-curve calculations. Buterin's findings show that once these tables grow beyond CPU cache size, they become slower than smaller tables that fit in cache. The cost of fetching data increases with memory size, making bigger tables counterproductive.
Buterin explained that physical layout drives this behavior. Processors and memory sit at different distances, and more capacity requires more physical space. Greater distance means longer access times as systems scale. He wrote that developers should account for memory distance and hierarchy when designing algorithms rather than assuming uniform access time.
Also read: Buterin says PeerDAS will make L2 faster
Also read: Buterin says PeerDAS will make L2 faster
The post covers practical applications in blockchain workloads and zero-knowledge systems that use heavy precomputation. Buterin's analysis addresses how engineers can choose table sizes and data layouts to reduce cache misses in proof systems and signature verification.
Buterin regularly publishes technical notes on cryptography and Ethereum performance. The Oct. 5 post adds to that work with a model and measurements focused on memory-aware design. The research targets blockchain clients and cryptographic libraries that rely on precomputation strategies.
