The most recent proof system, Boojum, introduced by zkSync Era, offers a higher throughput compared to the current rate of 100 transactions per second (TPS) and provides cost reductions in the long run. Boojum is a Scalable Transparent Argument of Knowledge (STARK)-based proof system that operates on consumer-grade general processing units (GPUs).
ZkSync Era is among a small number of Ethereum scaling solutions that employ zero-knowledge rollups (ZK-rollups) to enhance capacity, speed, and lower fees. These ZK-rollups primarily rely on two popular proof systems: zk-STARKs and Zero-Knowledge Succinct Non-Interactive Argument of Knowledge (zk-SNARKs).
Boojum, the new proof system, builds upon a Rust-based cryptographic library developed by zkSync. It incorporates an upgraded version of arithmetic circuits for zkSync Era and its ZK developer stack. Notably, Boojum enables everyday personal computers to act as provers, eliminating the need for powerful hardware and servers.
This upgrade allows computers with as little as 16 gigabytes (GB) of GPU random-access memory (RAM) to participate in network activity. In comparison, zkSync currently operates on a cluster of 100 GPUs, each with 80 GB of RAM.
Prior to the launch of Boojum, zkSync primarily relied on zk-SNARKs, which were effective but less transparent compared to zk-STARK-based systems. The previous system processed approximately 100 transactions per second, while Boojum promises to provide superior processing capabilities.
During the final stage of implementation, the new proof system will combine STARK proofs with a non-transparent pairing-based SNARK. This will result in a slightly upgraded version of the current SNARK-based proof system. The upgraded proof requires less storage and incurs lower verification costs, thereby reducing the overall cost of the proof system and the transactions themselves.
Boojum is currently operational on the zkSync Era mainnet for testing purposes, generating and verifying “shadow proofs.” Developers are actively testing these shadow proofs using real production data before proceeding with a complete migration and wider adoption. The system will be upgraded without any regenesis.