BEP-675 on BSC Testnet: 88% Higher Throughput by Cutting Redundant Execution

TL;DR

• BEP-675 lets block builders submit a fully executed block instead of a bid, so validators skip re-executing it before sealing.
• On a cross-region testnet mirroring mainnet, throughput rose from 1,237 to 2,324 transactions per second (TPS) at the same 450ms block interval and 100M gas limit.
• Blocks that averaged under half full now pack to the gas limit on most blocks and finality lag didn't move.
• Builders adopting BEP-675 need to run a fullnode instead of a fastnode and legacy bids still work but lose out on packing time.

Most blocks on BNB Smart Chain are produced by specialized block builders competing to assemble the most valuable set of transactions.

Under the existing process, builders execute a block before submitting it, then validators execute it again before sealing. This duplicate execution sits on the critical path and reduces the time available to fill the next block, especially as gas limits increase.

BEP-675 removes that bottleneck.

What BEP-675 Changes

Block builders construct and submit bids to the validator responsible for producing the next block. Validators then re-execute every transaction in the bid before signing it. This protects the network from invalid blocks, but it means each block is executed twice within a 450ms interval, once by the builder and once by the validator.

As blocks grow, the validator’s execution time grows too. That time comes directly from the builder’s window to pack transactions, causing some blocks to be submitted partially filled or nearly empty.

With BEP-675, builders submit a fully executed block rather than an unverified bid. Validators check it against consensus rules, sign and broadcast it, then complete full verification afterward. This removes roughly 90 to 100ms of execution from the critical path, reducing the validator reserve from 125ms to 15ms.

Builders can use that additional time to include more transactions.

How the Test Was Run

The benchmark was conducted on QANet, an internal testnet designed to mirror BSC’s cross-region validator topology.

Both configurations used:

• A 450ms block interval
• A 100M gas limit
• The same mixed transaction workload
• Order flow routed through builders and sentries rather than the public mempool

The test compared the legacy SendBid flow with BEP-675’s SendBidBlock flow.

Results

Throughput increased by 88%, from 1,237 to 2,324 TPS, without changing the gas limit or block interval.

The improvement came mainly from fuller blocks. Under the legacy flow, around one-quarter of blocks were submitted nearly empty because builders ran out of time. With BEP-675, most blocks were packed close to the gas limit. Finality lag remained unchanged, while execution performance held up under the higher block usage.

These are testnet results, and handling multiple competing bids per slot becomes more complex under the new flow.

What This Means for BSC

BEP-675 reduces the tradeoff between higher gas limits and shorter builder windows, giving BSC more room to scale throughput.

It is part of the H2 2026 roadmap alongside FOCIL and Block-Level Access Lists.

Builders need to run a full node to use the new flow. Legacy bids will remain supported, but they will have less packing time. Validators can adopt BEP-675 through a client upgrade and configuration change.

For users, the experience stays the same, with more transactions able to clear in each block.

What's Next

The next step is testing whether the same gains hold at mainnet scale and alongside the other H2 2026 upgrades.

By removing one redundant execution step, BEP-675 delivered an 88% throughput increase on testnet without affecting finality.