RELEASE NOTES

v1.6.5 is a maintenance release that allows validators to set an optional cap on transaction gas limits.

Changelog

FEATURE

#3488 miner: optional transaction gas limit cap

IMPROVEMENT

#3483 feat: remove the handshake from the bsc protocol
#3486 feat: update filtermaps checkpoints

Notice regarding upgrade from v1.5.x to v1.6.5

Snapshot Regeneration

The upgrade will trigger snapshot regeneration. Please expect reduced node performance until this process finishes. On our reference hardware (8-core AMD EPYC 7R13, 32GB RAM, SSD), regeneration took approximately 5 hours.

Log Indexing

v1.6.x introduces a new log indexing mechanism. For v1.6.5, the index is constructed starting from block 71,126,721. This operation consumes significant CPU resources, and the node may appear unresponsive during startup. If you do not need the full index, add the flag --history.logs 345600 to limit indexing to the last 3 days.

Assets

Description

This is an experimental feature release based on v1.6.4, with binary version: v1.6.102.

This feature release implemented the none consensus based Block-Access-List. It is based on EIP-7928 and using the non-consensus method, which similar to BEP-592
Deviations from EIP-7928:

• Consensus-level changes were removed, include BAL hash in block header.
• BALs now transfer only via the block propagation payload (ETH/BSC P2P messages) to help peers import blocks faster.
• If the BAL-driven parallel execution path fails (missing data, validation mismatch, etc.), the importer falls back to the legacy serial execution pipeline to guarantee correctness, so enabling BAL never jeopardizes block processing.

The feature is disabled by default, to try it, you may add this flag: --experimental.bal.

Kindly note that our current implementation shows an approximate 18.6% performance improvement in average mgas/s within our local test environment. However, the actual benefits are contingent upon the complete upgrade of the entire network to support BAL; otherwise, your node will not receive blocks with BAL, resulting in no performance gain. We believe there is significant potential to further enhance this feature and achieve even greater performance improvements in the future. For more information, pls refer: https://www.bnbchain.org/en/blog/boosting-bnb-smart-chain-performance-with-block-access-list

Change Log

• feature: non-consensus based block-access-list(EIP-7928)

Assets

v1.6.4 is the release for the BSC Mainnet Fermi hard fork, which is expected to be activated at 2026-01-14 02:30:00 AM UTC. All BSC Mainnet nodes must upgrade to v1.6.4 before the hard fork time. Upgrading from v1.6.3 only requires a binary replacement.

v1.6.4 includes two BEPs, primarily aimed at reducing the block interval from 0.75 seconds to 0.45 seconds:

• BEP-619: Short Block Interval Phase Three: 0.45 Seconds
• BEP-590: Extended Voting Rules for Fast Finality Stability

In addition to the block interval reduction, several other key parameters will be updated.
Please refer to BEP-619: Parameter Changes for details.

Beyond the hard-fork-related changes, this release also includes multiple improvements and bug fixes.
For the full list, please refer to the changelog below.

Changelog

FEATURE

• #3466 config: update BSC Mainnet hard fork time: Fermi
• #3454 eth: support fetching commit ID information from extra data

BUGFIX

• #3448 TxPool: revert transaction lifetime back to 3 hours
• #3457 cmd/geth: remove support for the multidatabase flag
• #3467 go.mod: downgrade bls-eth-go-binary to restore Windows compatibility

IMPROVEMENT

• #3445 eth: improve prefetching by using cached reader
• #3459 feat: implement auto-unsubscribe for transactionReceipts
• #3468 metric: add metric for vote count

Notice regarding upgrade from v1.5.x to v1.6.4

Snapshot Regeneration

The upgrade will trigger snapshot regeneration. Please expect reduced node performance until this process finishes. On our reference hardware (8-core AMD EPYC 7R13, 32GB RAM, SSD), regeneration took approximately 5 hours.

Log Indexing

v1.6.x introduces a new log indexing mechanism. For v1.6.4, the index is constructed starting from block 59,484,738. This operation consumes significant CPU resources, and the node may appear unresponsive during startup. If you do not need the full index, add the flag --history.logs 345600 to limit indexing to the last 3 days.

Assets

Description

v1.6.3 is a maintenance release, which fixes 2 RPC issues and with several enhancements, pls refer change log for detail.

ChangeLog

FEATURE

NA

BUGFIX

#3429 build(deps): bump github.com/consensys/gnark-crypto
#3433 internal/ethapi: fix eth_simulateV1
#3431 eth/tracers: fix crasher in TraceCall with BlockOverrides

IMPROVEMENT

#3436 revert: revert the nano check in parlia
#3435 eth: fix stuck when handleBlockBroadcast

Assets

Description

v1.6.2 is for BSC testnet Fermi hardfork: Fermi. Fermi hard fork is a critical upgrade to further reduce block interval from 750ms to 450ms, so users will have even better experience. Here is the list of Fermi BEPs:

• BEP-590: Extended Voting Rules for Fast Finality Stability
• BEP-619: Short Block Interval Phase Three: 0.45 Seconds
• BEP-592: Non-Consensus Based Block-Level Access List
• BEP-593: Incremental Snapshot
• BEP-610: Implement EVM Super Instruction

Here is the timeline of BSC Fermi hard fork

• Testnet: 2025-11-10 02:25:00 AM UTC
• Mainnet: TBD

Besides the BEPs, v1.6.2 also include several improvements for miner, MEV and performance, and also includes two new features. Here is the description of the two new features:

• new websocket RPC filter API for transaction receipts. (#3363)
• new configuration "ProxyedNodeIds" to improve block propagation efficiency.(#3417)

Compatibility

• flag --light.** was removed, include: --light.egress, --light.ingress, --light.maxpeers, --light.nopruning, --light.nosyncserve, --light.serve

ChangeLog

FEATURE

#3363 websocket: add transactionReceipts for receipts notification
#3367 BEP-619: Short Block Interval Phase Three: 0.45 Seconds
#3368 BEP-590: Extended Voting Rules for Fast Finality Stability
#3374 Implement BEP-592: Non-Consensus Based Block-Level Access List
#3372 core/systemcontracts: define fermiUpgrade
#3390 feat: implement incremental snapshot
#3395 feat: EVM execution opcode level optimization
#3400 consensus/parlia: set kAncestorGenerationDepth to 3 in BEP-590
#3397 consensus/parlia: fix updateAttestation&improve assembleVoteAttestation
#3417 p2p: define ProxyedNodeIds in Config

BUGFIX

#3373 ethapi: reject oversize storage keys before hex decode

IMPROVEMENT

#3388 miner/minerconfig: update config to adapt 100M gaslimit
#3404 miner: validator not inturn backoff before mining
#3407 fix: change lock to read lock in legacy pool
#3415 eth: broadcast votes to evn peers regardless of deltaTdThreshold
#3416 cmd/geth: improve config for sentry nodes when init network
#3419 miner: use latest block as pending block for simplicity
#3426 eth: increase the delta td threshold to broadcast votes

Assets

Description

This is an experimental feature release based on v1.6.1-beta, with binary version: v1.6.101.

v1.6.1-beta-feature-ScalableDB introduces a multi-database architecture (scalaledb) that divides chaindata into four separate stores which includes BlockDataBase(with meta), TrieDataBase, SnapDataBase and IndexDataBase according to data schema . Scalaledb provides stronger storage scalability. Different DBs can be allocated customized resources and tuned configurations. In particular, databases can compact independently, avoiding interference and improving read performance under heavy I/O.

Run
Run geth from genesis with "multidatabse" flag:
./geth --config config.toml --datadir data-seed init --multidatabase genesis.json

If you run geth from snapshot, you need to use migrate tool to convert an existing snapshot to the scalaledb layout with
./geth -datadir <your-data-dir> db migrate . You can directly start after migration , no additional flags are required if your snapshot has already been migrated.

Performance Test Results:
In large-scale simulated benchmarks, Multi-DB improves Geth block sync performance by ~3.8%. Starting from the same block height, after seven days of syncing, the Multi-DB node was ahead by ~200,000 blocks.

Change Log

• ethdb: support multidb mode with dedicated snapshot DB and index DB

Assets

Description

This is an experimental feature release based on v1.6.1-beta, with binary version: v1.6.100. The suffix "-BAL" stands for "Block Access List.".

v1.6.1-beta-feature-BAL implements BEP-592: Non-Consensus Based Block-Level Access List to enhance block execution performance. This implementation introduces a Block Access List (BAL) feature that tracks account and storage access during transaction execution, enabling efficient prefetching of state data.

Try BAL:
As an experimental feature, BAL is disabled by default. It can be enabled by running with the flag --enablebal, or by updating the config.toml file as follows:

...  
[Node]  
EnableBAL = true  
...  

Since validators produce the BAL, they need to be upgraded first. The connected full nodes can then retrieve the BAL data to accelerate their synchronization performance.

Performance Test Results:

• Block import performance increased by approximately 15%, with potential improvements of nearly 20% for I/O-heavy blocks.
• CPU usage reduced by ~30%, as Tx prefetch would be replaced BAL.
• The size of the BAL is about 40% of the block size. For a block with a 75M GasLimit, the block could be around 270KB, while the BAL would be approximately 110KB.

Change Log

• Implement BEP-592: Non-Consensus Based Block-Level Access List

Assets

Description

v1.6.1-beta is a maintenance preview release, which fixes several issues of the v1.6.0-alpha, it is more reliable, so mark it as beta stage.

This release has no new feature, no breaking changes, only some bugfix and improvements, pls refer change log for detail.

ChangeLog

FEATURE

NA

BUGFIX

#3336 miner: avoid to commit a bid twice
#3347 fix: discovery AyncFilter deadlock on shutdown
#3340 core: rework fast node

IMPROVEMENT

#3337 eth/pebble: use NoSync as write mode
#3332 FilterMap: update bsc checkpoint file
#3324 eth/downloader: remove InsertHeaderChain to improve sync speed
#3319 core/rawdb: remove func AncientOffSet and ItemAmountInAncient
#3346 cmd/geth: remove subcmd hbss2pbss and insecure-prune-all
#3354 freezer: add debug log for out of bounds access
#3340 core: rework fast node
#3348 core/rawdb: align more code with upstream
#3361 cmd/utils: set journalfile default to true

Assets

Description

This is an experimental feature release based on v1.6.0-alpha, the "-incr-snapshot" suffix stands for "incremental snapshot".
v1.6.0-alpha-feature-incr-snapshot implements BEP-593: Incremental Snapshot

Incremental Snapshot mechanism makes it easier for users to start and run BSC nodes with reduced data requirements. By implementing this feature, users will be able to bootstrap their nodes more efficiently while maintaining the necessary functionality for participating in the BSC network.

• For enabling incremental snapshot generation:

geth --config config.toml --datadir ./data/ --cache 8000 --rpc.allow-unprotected-txs --history.transactions 0 --syncmode full --db.engine=pebble --state.scheme=path --incr.enable --incr.block-interval=806400 --incr.datadir ./incr/ --incr.state-buffer=1073741824 --incr.kept-blocks=1024 --mainnet --history.blocks=360000  

• For downloading and merging incremental snapshot:

geth --config config.toml --datadir ./data/ --cache 8000 --rpc.allow-unprotected-txs --history.transactions 0 --syncmode full --db.engine=pebble --state.scheme=path --incr.use-remote --incr.remote-url https://download.snapshots.bnbchain.world/mainnet-geth-pbss-incr --incr.datadir ./store-incr/ --mainnet --history.blocks=360000  

Changelog

• feat: implement incremental snapshot

Assets

Description

v1.6.0-alpha is a preview release for upstream code sync, it catches up with go-ethereum release [v1.16.1] and also includes several bug fix.

Notice:
a.upgrade to v1.6.0-alpha from v1.5.x would trigger snapshot regeneration, which would downgrade your node's performance until the regeneration process complete. The regeneration cost ~5 hours on our node with 8 core AMD EPYC 7R13, 32GB, SSD, but it could be different in your node.

b.for FastNode(--tries-verify-mode none), pls don't upgrade to v1.6.0-alpha, there is a data compatible issue.

Changelog

Code Sync

• upstream: merge geth-v1.16.1

Key changes from the code sync include:

• Archive Mode: Added history indexing and RPC querying interface
• Enhanced Log Filtering: Introduced Filtermap as a faster replacement for bloombit-based log queries
• Log Timestamps: Added timestamp metadata to derived event logs
• Performance Optimizations: Multiple improvements to enhance overall performance
• Code Refactoring: Major cleanup and improvements to components including:
  • PBSS snapshot system
  • Blockchain configuration
  • EVM internals
  • Removal of EOF (Ethereum Object Format) code

BUGFIX

• core/filtermaps: stop indexing if target block is pruned
• freezer: slow down freeze when live sync
• worker: fix a trie prefetch corner case
• consensus/parlia: ignore client version warning when in history sync
• fix: only enable EVN feature after node get synced
• core/fitermaps: fix final block logic
• fix: set all chain tables to be prunable
• api.go: add retry for snapshots stale error
• core/types: disable EIP-7594 in BSC

Others

• docs: update readme for release types
• prefetch: lower prefetch threshold from 100 to 50

Assets