BNB Chain Achieves 88% Throughput Surge via BEP-675 Upgrade

According to Crypto Briefing, BNB Chain deployed the BEP-675 upgrade on the BSC Testnet on August 7, pushing throughput from 1,237 TPS to 2,324 TPS — an 88% jump achieved without altering the 450ms block interval or the 100M gas ceiling. The change eliminates a long-standing redundancy in BSC's block production pipeline, where validators were re-executing transactions that block builders had already executed. For protocol architects and MEV-aware operators, the move signals a deeper shift: BNB Chain is moving toward trust-minimized pre-execution at the consensus layer, not merely at the application layer.

The redundancy problem on the critical path

Before BEP-675, BSC's block production followed a two-stage model — block builders assembled and executed a set of transactions, then validators re-executed the same bundle to verify state correctness. That verification step consumed more than 125ms of every 450ms block interval, roughly a quarter of the available budget, and held median gas utilization per block at 29.49M out of the 100M limit. In effect, the chain was leaving throughput on the table not because of consensus constraints but because of duplicated computation across the builder-validator boundary. The redundancy was, in part, an artifact of trust assumptions baked into MEV-aware architectures, where validators could not accept a builder's output without re-checking the resulting state transitions.

SendBidBlock and the shift to pre-executed blocks

BEP-675 introduces SendBidBlock, a submission mechanism that lets builders deliver fully executed blocks directly. Validators can then skip the re-execution step, relying on the pre-computed results while the underlying security model is preserved. Legacy SendBid flows remain supported for backward compatibility, which lowers migration risk — but operators that want to use SendBidBlock must run a full node, an infrastructure bar that effectively gates participation to professional builders. BNB Chain evaluated the upgrade on an internal cross-region QANet topology designed to mirror mainnet conditions, a notable methodological choice that makes the 2,324 TPS figure a closer approximation of production performance than synthetic benchmarks typically allow.

What changes for operators, and what comes next

The practical effect is sharp: validator execution on the critical path fell from approximately 125ms to 15ms, dropping that step's share of the block interval from over 25% to roughly 3%. Median gas usage per block rose from 29.49M to 98.99M, meaning blocks that previously filled less than a third of their capacity now fill almost completely. For MEV searchers and block builders, the immediate implication is that infrastructure tuned around redundant execution windows will need to be retooled around pre-executed delivery; for validators, the calculus is whether the expected bid uplift justifies the full-node operational cost. The proposal was first drafted on April 10 and shipped roughly four months later — a relatively short review cycle for a consensus-adjacent change, and one worth tracking for governance pacing on future upgrades. BEP-675 sits within BNB Chain's broader H2 2026 technical roadmap, which also calls for FOCIL — a forced inclusion list mechanism designed to prevent censorship at the block production level — and Block-Level Access Lists for further execution efficiency. Following the testnet phase, the next concrete step is mainnet-scale validation, with a stated target of doubling mainnet throughput and a longer-term goal of a 10x improvement over the current baseline. Watch for the mainnet proposal date, the full-node specifications required for SendBidBlock participation, and any governance signal on whether FOCIL will ship before or after the throughput mainnet cutover — the order will determine whether censorship resistance or raw scale reaches the base layer first.