diff --git a/cmd/geth/chaincmd.go b/cmd/geth/chaincmd.go
index 6589fb35d0..f078c5ab9c 100644
--- a/cmd/geth/chaincmd.go
+++ b/cmd/geth/chaincmd.go
@@ -439,7 +439,7 @@ func showMetrics() {
 	blockWriteTimer := metrics.GetOrRegisterResettingTimer("chain/write", nil)
 
 	blockPrefetchExecuteTimer := metrics.GetOrRegisterResettingTimer("chain/prefetch/executes", nil)
-	mgaspsHist := metrics.GetOrRegisterHistogram("chain/execution/mgasps", nil, metrics.NewUniformSample(2000))
+	chainMgaspsMeter := metrics.GetOrRegisterResettingTimer("chain/mgasps", nil)
 
 	// not important
 	fmt.Println("accountReadSingleTimer", accountReadSingleTimer.Total())
@@ -475,14 +475,16 @@ func showMetrics() {
 	fmt.Println("PrefetchBALtime:  ", core.PrefetchBALTime)
 	fmt.Println("PrefetchMergeTime:", core.PrefetchMergeBALTime)
 	fmt.Println("ParallelExeTime: ", core.ParallelExeTime)
-	fmt.Println("PostMergeTime:    ", core.PostMergeTime)
 	fmt.Println("PrefetchTrieWallTime: ", core.PrefetchTrieTimer)
 	// fmt.Println("PrefetchTrieCPUtime:  ", state.PrefetchTrieCPUTime)
 
+	// Sync state time
+	fmt.Println("PrefetchChTime: ", core.PrefetchChTime)
+
 	// total
 	fmt.Println("blockInsertTimer", blockInsertTimer.Total())
 
-	mgasps := mgaspsHist.Snapshot()
+	mgasps := chainMgaspsMeter.Snapshot()
 	fmt.Println("mgasps,mean,max,min:", int64(mgasps.Mean()), mgasps.Max(), mgasps.Min())
 }
 
diff --git a/cmd/utils/cmd.go b/cmd/utils/cmd.go
index 6931dd332c..1671175858 100644
--- a/cmd/utils/cmd.go
+++ b/cmd/utils/cmd.go
@@ -52,7 +52,7 @@ import (
 )
 
 const (
-	importBatchSize = 1
+	importBatchSize = 10
 )
 
 // ErrImportInterrupted is returned when the user interrupts the import process.
diff --git a/core/block_validator.go b/core/block_validator.go
index 591e472bc1..d074ff96a1 100644
--- a/core/block_validator.go
+++ b/core/block_validator.go
@@ -20,7 +20,6 @@ import (
 	"errors"
 	"fmt"
 
-	"github.com/ethereum/go-ethereum/consensus"
 	"github.com/ethereum/go-ethereum/core/state"
 	"github.com/ethereum/go-ethereum/core/types"
 	"github.com/ethereum/go-ethereum/params"
@@ -108,12 +107,13 @@ func (v *BlockValidator) ValidateBody(block *types.Block) error {
 	}
 
 	// Ancestor block must be known.
-	if !v.bc.HasBlockAndState(block.ParentHash(), block.NumberU64()-1) {
-		if !v.bc.HasBlock(block.ParentHash(), block.NumberU64()-1) {
-			return consensus.ErrUnknownAncestor
-		}
-		return consensus.ErrPrunedAncestor
-	}
+	// Don't check intermediate blocks for N-blocks
+	// if !v.bc.HasBlockAndState(block.ParentHash(), block.NumberU64()-1) {
+	// 	if !v.bc.HasBlock(block.ParentHash(), block.NumberU64()-1) {
+	// 		return consensus.ErrUnknownAncestor
+	// 	}
+	// 	return consensus.ErrPrunedAncestor
+	// }
 	return nil
 }
 
@@ -148,19 +148,20 @@ func (v *BlockValidator) ValidateState(block *types.Block, statedb *state.StateD
 		return fmt.Errorf("invalid receipt root hash (remote: %x local: %x)", header.ReceiptHash, receiptSha)
 	}
 	// Validate the parsed requests match the expected header value.
-	if header.RequestsHash != nil {
-		reqhash := types.CalcRequestsHash(res.Requests)
-		if reqhash != *header.RequestsHash {
-			return fmt.Errorf("invalid requests hash (remote: %x local: %x)", *header.RequestsHash, reqhash)
-		}
-	} else if res.Requests != nil {
-		return errors.New("block has requests before prague fork")
-	}
+	// if header.RequestsHash != nil {
+	// 	reqhash := types.CalcRequestsHash(res.Requests)
+	// 	if reqhash != *header.RequestsHash {
+	// 		return fmt.Errorf("invalid requests hash (remote: %x local: %x)", *header.RequestsHash, reqhash)
+	// 	}
+	// } else if res.Requests != nil {
+	// 	return errors.New("block has requests before prague fork")
+	// }
 	// Validate the state root against the received state root and throw
 	// an error if they don't match.
-	if root := statedb.IntermediateRoot(v.config.IsEIP158(header.Number)); header.Root != root {
-		return fmt.Errorf("invalid merkle root (remote: %x local: %x) dberr: %w", header.Root, root, statedb.Error())
-	}
+	// Commented for N-blocks
+	// if root := statedb.IntermediateRoot(v.config.IsEIP158(header.Number)); header.Root != root {
+	// 	return fmt.Errorf("invalid merkle root (remote: %x local: %x) dberr: %w", header.Root, root, statedb.Error())
+	// }
 	return nil
 }
 
diff --git a/core/blockchain.go b/core/blockchain.go
index cbf56f4690..1d3852ad59 100644
--- a/core/blockchain.go
+++ b/core/blockchain.go
@@ -108,8 +108,6 @@ var (
 	blockPrefetchTxsInvalidMeter = metrics.NewRegisteredMeter("chain/prefetch/txs/invalid", nil)
 	blockPrefetchTxsValidMeter   = metrics.NewRegisteredMeter("chain/prefetch/txs/valid", nil)
 
-	mgaspsHist = metrics.NewRegisteredHistogram("chain/execution/mgasps", nil, metrics.NewUniformSample(2000))
-
 	errInsertionInterrupted = errors.New("insertion is interrupted")
 	errChainStopped         = errors.New("blockchain is stopped")
 	errInvalidOldChain      = errors.New("invalid old chain")
@@ -1712,7 +1710,8 @@ func (bc *BlockChain) InsertChain(chain types.Blocks) (int, error) {
 	}
 	defer bc.chainmu.Unlock()
 
-	_, n, err := bc.insertChain(chain, true, false) // No witness collection for mass inserts (would get super large)
+	// _, n, err := bc.insertChain(chain, true, false) // No witness collection for mass inserts (would get super large)
+	_, n, err := bc.insertChainN(chain, true, false)
 	return n, err
 }
 
@@ -1946,6 +1945,9 @@ type blockProcessingResult struct {
 	procTime time.Duration
 	status   WriteStatus
 	witness  *stateless.Witness
+	// For merged N-blocks
+	receipts types.Receipts
+	logs     []*types.Log
 }
 
 // processBlock executes and validates the given block. If there was no error
diff --git a/core/blockchainN.go b/core/blockchainN.go
index 4ce6521986..78d8a917f4 100644
--- a/core/blockchainN.go
+++ b/core/blockchainN.go
@@ -20,6 +20,7 @@ package core
 import (
 	"errors"
 	"fmt"
+	"sync"
 	"sync/atomic"
 	"time"
 
@@ -35,6 +36,13 @@ import (
 	"github.com/ethereum/go-ethereum/log"
 )
 
+var (
+	PrefetchBALTime   = time.Duration(0)
+	PostMergeTime     = time.Duration(0)
+	PrefetchTrieTimer = time.Duration(0)
+	PrefetchChTime    = time.Duration(0)
+)
+
 func (bc *BlockChain) insertChainN(chain types.Blocks, setHead bool, makeWitness bool) (*stateless.Witness, int, error) {
 	// If the chain is terminating, don't even bother starting up.
 	if bc.insertStopped() {
@@ -133,125 +141,228 @@ func (bc *BlockChain) insertChainN(chain types.Blocks, setHead bool, makeWitness
 		return nil, it.index, err
 	}
 	// Track the singleton witness from this chain insertion (if any)
-	var witness *stateless.Witness
+	var (
+		witness      *stateless.Witness
+		wg           sync.WaitGroup
+		allReceipts  types.Receipts
+		allLogs      []*types.Log
+		totalGasUsed uint64
+		headerTime   time.Duration
+	)
 
-	for ; block != nil && err == nil || errors.Is(err, ErrKnownBlock); block, err = it.next() {
-		// If the chain is terminating, stop processing blocks
-		if bc.insertStopped() {
-			log.Debug("Abort during block processing")
-			break
+	// All blocks share the same stateDB to simulate commiting after processing multiple blocks
+	startBlock := block
+	endBlock := chain[len(chain)-1]
+	parent := it.previous()
+	if parent == nil {
+		parent = bc.GetHeader(startBlock.ParentHash(), startBlock.NumberU64()-1)
+	}
+	statedb, err := state.New(parent.Root, bc.statedb)
+	if err != nil {
+		log.Crit("failed to initailzied state", "error", err, "root:", parent.Root.Hex())
+	}
+	// Prefetch pre-N-blocks state with merged pre-block BALs for N-blocks
+	// Here pre-block BALs are not merged, but we skipped allready fetch state to simulate merged operations.
+	prefetchStart := time.Now()
+	for _, blk := range chain {
+		statedb.PrefetchStateBAL(blk.NumberU64())
+	}
+	PrefetchBALTime += time.Since(prefetchStart)
+
+	// pre-block state for block N
+	prestateCh := make(chan *state.StateDB, len(chain))
+
+	wg.Add(2)
+	// process post-N-blocks state with merged post-BALs for N-blocks
+	go func() {
+		defer wg.Done()
+
+		mstart := time.Now()
+		// Stop prefetcher cause we'll directly fetch tries in parallel
+		statedb.StopPrefetcher()
+
+		for _, blk := range chain {
+			// We don't need to worry about blockNumber is not changed, because it'll be set during process block
+			prestateCh <- statedb.CopyState()
+			statedb.MergePostBalStates(blk.NumberU64())
 		}
-		// If the block is known (in the middle of the chain), it's a special case for
-		// Clique blocks where they can share state among each other, so importing an
-		// older block might complete the state of the subsequent one. In this case,
-		// just skip the block (we already validated it once fully (and crashed), since
-		// its header and body was already in the database). But if the corresponding
-		// snapshot layer is missing, forcibly rerun the execution to build it.
-		if bc.skipBlock(err, it) {
-			logger := log.Debug
-			if bc.chainConfig.Clique == nil {
-				logger = log.Warn
-			}
-			logger("Inserted known block", "number", block.Number(), "hash", block.Hash(),
-				"uncles", len(block.Uncles()), "txs", len(block.Transactions()), "gas", block.GasUsed(),
-				"root", block.Root())
+		PostMergeTime += time.Since(mstart)
 
-			// Special case. Commit the empty receipt slice if we meet the known
-			// block in the middle. It can only happen in the clique chain. Whenever
-			// we insert blocks via `insertSideChain`, we only commit `td`, `header`
-			// and `body` if it's non-existent. Since we don't have receipts without
-			// reexecution, so nothing to commit. But if the sidechain will be adopted
-			// as the canonical chain eventually, it needs to be reexecuted for missing
-			// state, but if it's this special case here(skip reexecution) we will lose
-			// the empty receipt entry.
-			if len(block.Transactions()) == 0 {
-				rawdb.WriteReceipts(bc.db, block.Hash(), block.NumberU64(), nil)
-			} else {
-				log.Error("Please file an issue, skip known block execution without receipt",
-					"hash", block.Hash(), "number", block.NumberU64())
+		// Prewarm the trie for future committing
+		pstart := time.Now()
+		statedb.PrefetchTrie()
+		PrefetchTrieTimer += time.Since(pstart)
+	}()
+
+	go func() {
+		defer wg.Done()
+
+		hstart := time.Now()
+		// Write all headers then parallel executing to validate post-tx BALs
+		bc.writeNBlockHeaders(chain)
+		headerTime = time.Since(hstart)
+
+		for ; block != nil && err == nil || errors.Is(err, ErrKnownBlock); block, err = it.next() {
+			// If the chain is terminating, stop processing blocks
+			if bc.insertStopped() {
+				log.Debug("Abort during block processing")
+				break
 			}
-			if err := bc.writeKnownBlock(block); err != nil {
-				return nil, it.index, err
+			// If the block is known (in the middle of the chain), it's a special case for
+			// Clique blocks where they can share state among each other, so importing an
+			// older block might complete the state of the subsequent one. In this case,
+			// just skip the block (we already validated it once fully (and crashed), since
+			// its header and body was already in the database). But if the corresponding
+			// snapshot layer is missing, forcibly rerun the execution to build it.
+			if bc.skipBlock(err, it) {
+				logger := log.Debug
+				if bc.chainConfig.Clique == nil {
+					logger = log.Warn
+				}
+				logger("Inserted known block", "number", block.Number(), "hash", block.Hash(),
+					"uncles", len(block.Uncles()), "txs", len(block.Transactions()), "gas", block.GasUsed(),
+					"root", block.Root())
+
+				// Special case. Commit the empty receipt slice if we meet the known
+				// block in the middle. It can only happen in the clique chain. Whenever
+				// we insert blocks via `insertSideChain`, we only commit `td`, `header`
+				// and `body` if it's non-existent. Since we don't have receipts without
+				// reexecution, so nothing to commit. But if the sidechain will be adopted
+				// as the canonical chain eventually, it needs to be reexecuted for missing
+				// state, but if it's this special case here(skip reexecution) we will lose
+				// the empty receipt entry.
+				if len(block.Transactions()) == 0 {
+					rawdb.WriteReceipts(bc.db, block.Hash(), block.NumberU64(), nil)
+				} else {
+					log.Error("Please file an issue, skip known block execution without receipt",
+						"hash", block.Hash(), "number", block.NumberU64())
+				}
+				if err := bc.writeKnownBlock(block); err != nil {
+					witness = nil
+					return
+				}
+				stats.processed++
+				if bc.logger != nil && bc.logger.OnSkippedBlock != nil {
+					bc.logger.OnSkippedBlock(tracing.BlockEvent{
+						Block:     block,
+						Finalized: bc.CurrentFinalBlock(),
+						Safe:      bc.CurrentSafeBlock(),
+					})
+				}
+				// We can assume that logs are empty here, since the only way for consecutive
+				// Clique blocks to have the same state is if there are no transactions.
+				lastCanon = block
+				continue
 			}
+			// Retrieve the parent block and it's state to execute on top
+			parent := it.previous()
+			if parent == nil {
+				parent = bc.GetHeader(block.ParentHash(), block.NumberU64()-1)
+			}
+			// The traced section of block import.
+			pstart := time.Now()
+			statedbForBlock := <-prestateCh
+			PrefetchChTime += time.Since(pstart)
+
+			res, err := bc.processBlockWithState(parent.Root, block, setHead, makeWitness && len(chain) == 1, statedbForBlock)
+			if err != nil {
+				witness = nil
+				log.Crit("Failed to processBlock", "error", err)
+				return
+			}
+
+			// collect logs and receipts for N-blocks
+			allReceipts = append(allReceipts, res.receipts...)
+			allLogs = append(allLogs, res.logs...)
+			totalGasUsed += res.usedGas
+			// Report the import stats before returning the various results
 			stats.processed++
-			if bc.logger != nil && bc.logger.OnSkippedBlock != nil {
-				bc.logger.OnSkippedBlock(tracing.BlockEvent{
-					Block:     block,
-					Finalized: bc.CurrentFinalBlock(),
-					Safe:      bc.CurrentSafeBlock(),
-				})
+			stats.usedGas += res.usedGas
+			witness = res.witness
+
+			var snapDiffItems, snapBufItems common.StorageSize
+			if bc.snaps != nil {
+				snapDiffItems, snapBufItems = bc.snaps.Size()
+			}
+			trieDiffNodes, trieBufNodes, _ := bc.triedb.Size()
+			stats.report(chain, it.index, snapDiffItems, snapBufItems, trieDiffNodes, trieBufNodes, setHead)
+
+			// Print confirmation that a future fork is scheduled, but not yet active.
+			bc.logForkReadiness(block)
+
+			if !setHead {
+				// After merge we expect few side chains. Simply count
+				// all blocks the CL gives us for GC processing time
+				bc.gcproc += res.procTime
+				witness = nil
+				return // Direct block insertion of a single block
 			}
-			// We can assume that logs are empty here, since the only way for consecutive
-			// Clique blocks to have the same state is if there are no transactions.
-			lastCanon = block
-			continue
 		}
-		// Retrieve the parent block and it's state to execute on top
-		parent := it.previous()
-		if parent == nil {
-			parent = bc.GetHeader(block.ParentHash(), block.NumberU64()-1)
-		}
-		// The traced section of block import.
-		start := time.Now()
-		res, err := bc.processBlock(parent.Root, block, setHead, makeWitness && len(chain) == 1)
-		if err != nil {
-			return nil, it.index, err
-		}
-		// Report the import stats before returning the various results
-		stats.processed++
-		stats.usedGas += res.usedGas
-		witness = res.witness
 
-		var snapDiffItems, snapBufItems common.StorageSize
-		if bc.snaps != nil {
-			snapDiffItems, snapBufItems = bc.snaps.Size()
-		}
-		trieDiffNodes, trieBufNodes, _ := bc.triedb.Size()
-		stats.report(chain, it.index, snapDiffItems, snapBufItems, trieDiffNodes, trieBufNodes, setHead)
+		stats.ignored += it.remaining()
+	}()
+	wg.Wait()
 
-		// Print confirmation that a future fork is scheduled, but not yet active.
-		bc.logForkReadiness(block)
+	// Validate stateRoot for N-blocks at once.
+	xvtime := time.Now()
+	header := endBlock.Header()
+	if root := statedb.IntermediateRoot(true); header.Root != root {
+		log.Crit("invalid merkle root (remote: %x local: %x) dberr: %w", header.Root, root, statedb.Error())
+	}
+	blockCrossValidationTimer.Update(time.Since(xvtime))
 
-		if !setHead {
-			// After merge we expect few side chains. Simply count
-			// all blocks the CL gives us for GC processing time
-			bc.gcproc += res.procTime
-			return witness, it.index, nil // Direct block insertion of a single block
-		}
-		switch res.status {
-		case CanonStatTy:
-			log.Debug("Inserted new block", "number", block.Number(), "hash", block.Hash(),
-				"uncles", len(block.Uncles()), "txs", len(block.Transactions()), "gas", block.GasUsed(),
-				"elapsed", common.PrettyDuration(time.Since(start)),
-				"root", block.Root())
-
-			lastCanon = block
-
-			// Only count canonical blocks for GC processing time
-			bc.gcproc += res.procTime
-
-		case SideStatTy:
-			log.Debug("Inserted forked block", "number", block.Number(), "hash", block.Hash(),
-				"diff", block.Difficulty(), "elapsed", common.PrettyDuration(time.Since(start)),
-				"txs", len(block.Transactions()), "gas", block.GasUsed(), "uncles", len(block.Uncles()),
-				"root", block.Root())
-
-		default:
-			// This in theory is impossible, but lets be nice to our future selves and leave
-			// a log, instead of trying to track down blocks imports that don't emit logs.
-			log.Warn("Inserted block with unknown status", "number", block.Number(), "hash", block.Hash(),
-				"diff", block.Difficulty(), "elapsed", common.PrettyDuration(time.Since(start)),
-				"txs", len(block.Transactions()), "gas", block.GasUsed(), "uncles", len(block.Uncles()),
-				"root", block.Root())
-		}
+	// Commit N-blocks together to the chain and get the status.
+	var (
+		wstart = time.Now()
+		status WriteStatus
+	)
+	if !setHead {
+		// Don't set the head, only insert the block
+		err = bc.writeNBlocksWithState(startBlock, endBlock, allReceipts, statedb)
+	} else {
+		status, err = bc.writeNBlocksAndSetHead(startBlock, endBlock, allReceipts, allLogs, statedb, false)
+	}
+	if err != nil {
+		return nil, it.index, err
 	}
 
-	stats.ignored += it.remaining()
+	// Update the metrics touched during N-blocks commit
+	accountCommitTimer.Update(statedb.AccountCommits)   // Account commits are complete, we can mark them
+	storageCommitTimer.Update(statedb.StorageCommits)   // Storage commits are complete, we can mark them
+	snapshotCommitTimer.Update(statedb.SnapshotCommits) // Snapshot commits are complete, we can mark them
+	triedbCommitTimer.Update(statedb.TrieDBCommits)     // Trie database commits are complete, we can mark them
+
+	blockWriteTimer.Update(time.Since(wstart) - max(statedb.AccountCommits, statedb.StorageCommits) /* concurrent */ - statedb.SnapshotCommits - statedb.TrieDBCommits + headerTime)
+
+	switch status {
+	case CanonStatTy:
+		log.Debug("Inserted new blocks", "numberStart", startBlock.Number(), "numberEnd", endBlock.Number(),
+			"elapsed", common.PrettyDuration(time.Since(prefetchStart)),
+			"root", endBlock.Root())
+
+		lastCanon = endBlock
+
+	default:
+		// This in theory is impossible, but lets be nice to our future selves and leave
+		// a log, instead of trying to track down blocks imports that don't emit logs.
+		log.Warn("Inserted block with unknown status", "number", endBlock.Number(), "hash", endBlock.Hash(),
+			"diff", endBlock.Difficulty(), "elapsed", common.PrettyDuration(time.Since(prefetchStart)),
+			"txs", len(endBlock.Transactions()), "gas", endBlock.GasUsed(), "uncles", len(endBlock.Uncles()),
+			"root", endBlock.Root())
+	}
+
+	elapsed := time.Since(prefetchStart) + 1 // prevent zero division
+	blockInsertTimer.Update(elapsed)
+
+	// TODO(rjl493456442) generalize the ResettingTimer
+	mgasps := float64(totalGasUsed) * 1000 / float64(elapsed)
+	chainMgaspsMeter.Update(time.Duration(mgasps))
+
 	return witness, it.index, err
 }
 
-func (bc *BlockChain) writeNBlocksWithState(block *types.Block, receipts []*types.Receipt, statedb *state.StateDB) error {
-	if !bc.HasHeader(block.ParentHash(), block.NumberU64()-1) {
+func (bc *BlockChain) writeNBlocksWithState(startBlock, endBlock *types.Block, receipts []*types.Receipt, statedb *state.StateDB) error {
+	if !bc.HasHeader(startBlock.ParentHash(), startBlock.NumberU64()-1) {
 		return consensus.ErrUnknownAncestor
 	}
 	// Irrelevant of the canonical status, write the block itself to the database.
@@ -259,14 +370,14 @@ func (bc *BlockChain) writeNBlocksWithState(block *types.Block, receipts []*type
 	// Note all the components of block(hash->number map, header, body, receipts)
 	// should be written atomically. BlockBatch is used for containing all components.
 	blockBatch := bc.db.NewBatch()
-	rawdb.WriteBlock(blockBatch, block)
-	rawdb.WriteReceipts(blockBatch, block.Hash(), block.NumberU64(), receipts)
+	rawdb.WriteBlock(blockBatch, endBlock)
+	rawdb.WriteReceipts(blockBatch, endBlock.Hash(), endBlock.NumberU64(), receipts)
 	rawdb.WritePreimages(blockBatch, statedb.Preimages())
 	if err := blockBatch.Write(); err != nil {
 		log.Crit("Failed to write block into disk", "err", err)
 	}
 	// Commit all cached state changes into underlying memory database.
-	root, err := statedb.Commit(block.NumberU64(), bc.chainConfig.IsEIP158(block.Number()), bc.chainConfig.IsCancun(block.Number(), block.Time()))
+	root, err := statedb.Commit(endBlock.NumberU64(), bc.chainConfig.IsEIP158(endBlock.Number()), bc.chainConfig.IsCancun(endBlock.Number(), endBlock.Time()))
 	if err != nil {
 		return err
 	}
@@ -281,10 +392,10 @@ func (bc *BlockChain) writeNBlocksWithState(block *types.Block, receipts []*type
 	}
 	// Full but not archive node, do proper garbage collection
 	bc.triedb.Reference(root, common.Hash{}) // metadata reference to keep trie alive
-	bc.triegc.Push(root, -int64(block.NumberU64()))
+	bc.triegc.Push(root, -int64(endBlock.NumberU64()))
 
 	// Flush limits are not considered for the first TriesInMemory blocks.
-	current := block.NumberU64()
+	current := endBlock.NumberU64()
 	if current <= state.TriesInMemory {
 		return nil
 	}
@@ -330,23 +441,34 @@ func (bc *BlockChain) writeNBlocksWithState(block *types.Block, receipts []*type
 	return nil
 }
 
-func (bc *BlockChain) writeNBlocksAndSetHead(block *types.Block, receipts []*types.Receipt, logs []*types.Log, state *state.StateDB, emitHeadEvent bool) (status WriteStatus, err error) {
-	if err := bc.writeBlockWithState(block, receipts, state); err != nil {
+// Due to blockHash opCode, blockHeaders must be written to DB before commit.
+func (bc *BlockChain) writeNBlockHeaders(chain types.Blocks) {
+	blockBatch := bc.db.NewBatch()
+	for _, block := range chain {
+		rawdb.WriteHeader(blockBatch, block.Header())
+	}
+	if err := blockBatch.Write(); err != nil {
+		log.Crit("Failed to write block into disk", "err", err)
+	}
+}
+
+func (bc *BlockChain) writeNBlocksAndSetHead(startBlock, endBlock *types.Block, receipts []*types.Receipt, logs []*types.Log, state *state.StateDB, emitHeadEvent bool) (status WriteStatus, err error) {
+	if err := bc.writeBlockWithState(endBlock, receipts, state); err != nil {
 		return NonStatTy, err
 	}
 	currentBlock := bc.CurrentBlock()
 
 	// Reorganise the chain if the parent is not the head block
-	if block.ParentHash() != currentBlock.Hash() {
-		if err := bc.reorg(currentBlock, block.Header()); err != nil {
+	if startBlock.ParentHash() != currentBlock.Hash() {
+		if err := bc.reorg(currentBlock, endBlock.Header()); err != nil {
 			return NonStatTy, err
 		}
 	}
 
 	// Set new head.
-	bc.writeHeadBlock(block)
+	bc.writeHeadBlock(endBlock)
 
-	bc.chainFeed.Send(ChainEvent{Header: block.Header()})
+	bc.chainFeed.Send(ChainEvent{Header: endBlock.Header()})
 	if len(logs) > 0 {
 		bc.logsFeed.Send(logs)
 	}
@@ -356,33 +478,24 @@ func (bc *BlockChain) writeNBlocksAndSetHead(block *types.Block, receipts []*typ
 	// we will fire an accumulated ChainHeadEvent and disable fire
 	// event here.
 	if emitHeadEvent {
-		bc.chainHeadFeed.Send(ChainHeadEvent{Header: block.Header()})
+		bc.chainHeadFeed.Send(ChainHeadEvent{Header: endBlock.Header()})
 	}
 	return CanonStatTy, nil
 }
 
-func (bc *BlockChain) processBlockWithState(parentRoot common.Hash, block *types.Block, setHead bool, makeWitness bool) (_ *blockProcessingResult, blockEndErr error) {
+func (bc *BlockChain) processBlockWithState(parentRoot common.Hash, block *types.Block, setHead bool, makeWitness bool, statedb *state.StateDB) (_ *blockProcessingResult, blockEndErr error) {
 	var (
 		err       error
 		startTime = time.Now()
-		statedb   *state.StateDB
 		interrupt atomic.Bool
 	)
 	defer interrupt.Store(true) // terminate the prefetch at the end
+	if !bc.cfg.NoPrefetch {
+		panic("Must enable --cache.noprefetch to perf N-blocks")
+	}
 
 	if bc.cfg.NoPrefetch {
-		statedb, err = state.New(parentRoot, bc.statedb)
-		if err != nil {
-			return nil, err
-		}
-		// reader, err := bc.statedb.ReaderWithCache(parentRoot)
-		// if err != nil {
-		// 	return nil, err
-		// }
-		// statedb, err = state.NewWithReader(parentRoot, bc.statedb, reader)
-		// if err != nil {
-		// 	return nil, err
-		// }
+		// Use stateDB snapshotted from statedb.MergePostBalStates
 	} else {
 		fmt.Println("prefetch enabled===========================")
 		// If prefetching is enabled, run that against the current state to pre-cache
@@ -443,8 +556,9 @@ func (bc *BlockChain) processBlockWithState(parentRoot common.Hash, block *types
 				return nil, err
 			}
 		}
-		statedb.StartPrefetcher("chain", witness)
-		defer statedb.StopPrefetcher()
+		// Don't start the prefetcher, as it can significantly degrade performance. We've already prefetched the trie in insertChainN.
+		// statedb.StartPrefetcher("chain", witness)
+		// defer statedb.StopPrefetcher()
 	}
 
 	if bc.logger != nil && bc.logger.OnBlockStart != nil {
@@ -525,38 +639,10 @@ func (bc *BlockChain) processBlockWithState(parentRoot common.Hash, block *types
 	blockValidationTimer.Update(vtime - (triehash + trieUpdate))                      // The time spent on block validation
 	blockCrossValidationTimer.Update(xvtime)                                          // The time spent on stateless cross validation
 
-	// Write the block to the chain and get the status.
-	var (
-		wstart = time.Now()
-		status WriteStatus
-	)
-	if !setHead {
-		// Don't set the head, only insert the block
-		err = bc.writeBlockWithState(block, res.Receipts, statedb)
-	} else {
-		status, err = bc.writeBlockAndSetHead(block, res.Receipts, res.Logs, statedb, false)
-	}
-	if err != nil {
-		return nil, err
-	}
-	// Update the metrics touched during block commit
-	accountCommitTimer.Update(statedb.AccountCommits)   // Account commits are complete, we can mark them
-	storageCommitTimer.Update(statedb.StorageCommits)   // Storage commits are complete, we can mark them
-	snapshotCommitTimer.Update(statedb.SnapshotCommits) // Snapshot commits are complete, we can mark them
-	triedbCommitTimer.Update(statedb.TrieDBCommits)     // Trie database commits are complete, we can mark them
-
-	blockWriteTimer.Update(time.Since(wstart) - max(statedb.AccountCommits, statedb.StorageCommits) /* concurrent */ - statedb.SnapshotCommits - statedb.TrieDBCommits)
-	elapsed := time.Since(startTime) + 1 // prevent zero division
-	blockInsertTimer.Update(elapsed)
-
-	// TODO(rjl493456442) generalize the ResettingTimer
-	mgasps := float64(res.GasUsed) * 1000 / float64(elapsed)
-	chainMgaspsMeter.Update(time.Duration(mgasps))
-
 	return &blockProcessingResult{
 		usedGas:  res.GasUsed,
 		procTime: proctime,
-		status:   status,
+		status:   CanonStatTy, // Assue write status is always correct
 		witness:  witness,
 	}, nil
 }
diff --git a/core/blockchain_insert.go b/core/blockchain_insert.go
index 7edd674f17..56e6ffc1fb 100644
--- a/core/blockchain_insert.go
+++ b/core/blockchain_insert.go
@@ -46,8 +46,6 @@ func (st *insertStats) report(chain []*types.Block, index int, snapDiffItems, sn
 		elapsed = now.Sub(st.startTime) + 1 // prevent zero division
 		mgasps  = float64(st.usedGas) * 1000 / float64(elapsed)
 	)
-	// Update the Mgas per second gauge
-	mgaspsHist.Update(int64(mgasps))
 
 	// If we're at the last block of the batch or report period reached, log
 	if index == len(chain)-1 || elapsed >= statsReportLimit {
diff --git a/core/parallel_state_processor.go b/core/parallel_state_processor.go
index 2b76ce468e..d8ea4c75da 100644
--- a/core/parallel_state_processor.go
+++ b/core/parallel_state_processor.go
@@ -4,7 +4,6 @@ import (
 	"fmt"
 	"math/big"
 	"runtime"
-	"sync"
 	"time"
 
 	"github.com/ethereum/go-ethereum/common"
@@ -17,11 +16,9 @@ import (
 )
 
 var (
-	PrefetchBALTime      = time.Duration(0)
-	PrefetchMergeBALTime = time.Duration(0)
 	ParallelExeTime      = time.Duration(0)
-	PostMergeTime        = time.Duration(0)
-	PrefetchTrieTimer    = time.Duration(0)
+	PrefetchMergeBALTime = time.Duration(0)
+	EmptyStatedb         = state.NewEmptyDB()
 )
 
 type ParallelStateProcessor struct {
@@ -43,10 +40,8 @@ func (p *ParallelStateProcessor) Process(block *types.Block, statedb *state.Stat
 		context   vm.BlockContext
 		gp        = new(GasPool).AddGas(block.GasLimit())
 		signer    = types.MakeSigner(p.config, header.Number, header.Time)
-		initialdb = statedb.Copy()
-		postState *state.StateDB
+		initialdb = EmptyStatedb
 
-		wg     sync.WaitGroup
 		result *ProcessResult
 		err    error
 	)
@@ -56,14 +51,9 @@ func (p *ParallelStateProcessor) Process(block *types.Block, statedb *state.Stat
 		misc.ApplyDAOHardFork(statedb)
 	}
 
-	if preStateType == BALPreState {
-		// copy initialdb before PrefetchStateBAL to avoid redundant copy
-		start := time.Now()
-		// Must prefetch bal before syscall to avoid overriding syscall's state, thus merkle root might mismatch
-		statedb.PrefetchStateBAL(block.NumberU64())
-		// statedb.SetBlocknumber(block.NumberU64())
-		PrefetchBALTime += time.Since(start)
-	}
+	preCalPostStart := time.Now()
+	statedb.PreComputePostState(block.NumberU64(), runtime.NumCPU()/2)
+	PrefetchMergeBALTime += time.Since(preCalPostStart)
 
 	// Apply pre-execution system calls.
 	var tracingStateDB = vm.StateDB(statedb)
@@ -80,43 +70,9 @@ func (p *ParallelStateProcessor) Process(block *types.Block, statedb *state.Stat
 		ProcessParentBlockHash(block.ParentHash(), evm)
 	}
 
-	wg.Add(2)
-	go func() {
-		defer wg.Done()
-
-		start := time.Now()
-		postState = statedb.CopyState()
-		// Stop prefetcher cause we'll directly fetch tries in parallel
-		postState.StopPrefetcher()
-
-		postState.MergePostBalStates()
-		PostMergeTime += time.Since(start)
-
-		// Prewarm the updating trie
-		start = time.Now()
-		postState.PrefetchTrie()
-		PrefetchTrieTimer += time.Since(start)
-	}()
-
-	go func() {
-		defer wg.Done()
-
-		preCalPostStart := time.Now()
-		statedb.PreComputePostState(block.NumberU64(), runtime.NumCPU()/2)
-		PrefetchMergeBALTime += time.Since(preCalPostStart)
-
-		exeStart := time.Now()
-		result, err = p.executeParallel(block, statedb, cfg, gp, signer, context, initialdb)
-		ParallelExeTime += time.Since(exeStart)
-
-	}()
-
-	wg.Wait()
-
-	// Last tx alreadly includes the state change in requests after all txs
-	if preStateType == BALPreState {
-		*statedb = *postState
-	}
+	exeStart := time.Now()
+	result, err = p.executeParallel(block, statedb, cfg, gp, signer, context, initialdb)
+	ParallelExeTime += time.Since(exeStart)
 	return result, err
 }
 
@@ -135,15 +91,16 @@ func (p *ParallelStateProcessor) executeParallel(block *types.Block, statedb *st
 	)
 
 	// leave some cpus for prefetching
-	workers.SetLimit(runtime.NumCPU() / 2)
 
 	switch preStateType {
 	case BALPreState:
 		{
+			workers.SetLimit(runtime.NumCPU() / 2)
 		}
 
 	case SeqPreState: // must set workers limit = 1
 		{
+			workers.SetLimit(1)
 			preStatedb := statedb.Copy()
 			gpcp := *gp
 			preStateProvider = &SequentialPrestateProvider{
@@ -297,7 +254,7 @@ func ApplyTransactionWithParallelEVM(msg *Message, gp *GasPool, statedb *state.S
 
 	// Merge the tx-local access event into the "block-local" one, in order to collect
 	// all values, so that the witness can be built.
-	if statedb.GetTrie().IsVerkle() {
+	if statedb.GetTrie() != nil && statedb.GetTrie().IsVerkle() {
 		statedb.AccessEvents().Merge(evm.AccessEvents)
 	}
 
diff --git a/core/state/statedb.go b/core/state/statedb.go
index b37a980a2c..8badaa3940 100644
--- a/core/state/statedb.go
+++ b/core/state/statedb.go
@@ -246,6 +246,23 @@ func New(root common.Hash, db Database) (*StateDB, error) {
 	return NewWithReader(root, db, reader)
 }
 
+// Empty statedb without tr
+func NewEmptyDB() *StateDB {
+	sdb := &StateDB{
+		stateObjects:         make(map[common.Address]*stateObject),
+		stateObjectsDestruct: make(map[common.Address]*stateObject),
+		mutations:            make(map[common.Address]*mutation),
+		logs:                 make(map[common.Hash][]*types.Log),
+		preimages:            make(map[common.Hash][]byte),
+		journal:              newJournal(),
+		updateJournal:        newJournal(),
+		accessList:           newAccessList(),
+		transientStorage:     newTransientStorage(),
+		postStates:           make(map[int]*StateDB),
+	}
+	return sdb
+}
+
 // NewWithReader creates a new state for the specified state root. Unlike New,
 // this function accepts an additional Reader which is bound to the given root.
 func NewWithReader(root common.Hash, db Database, reader Reader) (*StateDB, error) {
@@ -357,7 +374,7 @@ func (s *StateDB) PrefetchStateBAL(blockNumber uint64) {
 }
 
 func (s *StateDB) prefetchBalPreblockKeys() {
-	log.Info("PrefetchBalPreblockKeys...")
+	log.Info("PrefetchBalPreblockKeys for:", "block:", s.blockNumber)
 	type StorageKV struct {
 		addr *common.Address
 		key  *common.Hash
@@ -370,12 +387,17 @@ func (s *StateDB) prefetchBalPreblockKeys() {
 		workers     errgroup.Group
 	)
 
-	workers.SetLimit(runtime.NumCPU() - 2)
+	workers.SetLimit(runtime.NumCPU() / 2)
 	// Fetch pre-block acccount state
 	accounts := make(chan *stateObject, len(preBal))
+	lenAcctCh := 0
 	for _, acl := range preBal {
 		addr := acl.Address
 		lenMaxSlots += len(acl.StorageKeys)
+		if s.stateObjects[addr] != nil { // Skip already fetched pre-block bals or newObject will override existing storage!
+			continue
+		}
+		lenAcctCh++
 
 		workers.Go(func() error {
 			acctBal, err := s.reader.Account(addr)
@@ -393,9 +415,8 @@ func (s *StateDB) prefetchBalPreblockKeys() {
 			return nil
 		})
 	}
-	workers.Wait()
 
-	for range len(preBal) {
+	for range lenAcctCh {
 		obj := <-accounts
 		// must set it first to avoid accounts read later in storage fetching
 		if obj != nil {
@@ -404,7 +425,6 @@ func (s *StateDB) prefetchBalPreblockKeys() {
 	}
 
 	close(accounts)
-
 	// Fetch pre-block storage state
 	storages := make(chan *StorageKV, lenMaxSlots)
 	for _, acl := range preBal {
@@ -417,9 +437,13 @@ func (s *StateDB) prefetchBalPreblockKeys() {
 			continue
 		}
 
-		lenSlots += len(acl.StorageKeys)
-
 		for _, key := range keys {
+			// Skip already fetched slots to simulate merged N-blocks BALs
+			if _, cached := obj.originStorage[key]; cached {
+				continue
+			}
+			lenSlots++
+
 			workers.Go(func() error {
 				val, err := s.reader.Storage(addr, key)
 				kv := &StorageKV{&addr, &key, &val}
@@ -433,7 +457,6 @@ func (s *StateDB) prefetchBalPreblockKeys() {
 			})
 		}
 	}
-	workers.Wait()
 
 	for range lenSlots {
 		kv := <-storages
@@ -443,7 +466,6 @@ func (s *StateDB) prefetchBalPreblockKeys() {
 			s.setStateObject(account)
 		}
 	}
-
 	close(storages)
 }
 
@@ -683,16 +705,17 @@ func (s *StateDB) MergeState(entries []JournalEntry) {
 	}
 }
 
-func (s *StateDB) MergePostBalStates() {
+func (s *StateDB) MergePostBalStates(blockNumber uint64) {
 	if balType != BalPreblockKeysPostValues {
 		panic("MergePostBal is only supported with BalPreblockKeysPostValues")
 	}
 	var (
-		postBal = AllBlockTxPostValues[s.blockNumber]
+		postBal = AllBlockTxPostValues[blockNumber]
 	)
+	s.blockNumber = blockNumber
 
-	// -1 and len(postBal)-1 (not used) are syscalls pre and post all txs.
-	for txIndex := -1; txIndex < len(postBal)-1; txIndex++ {
+	// -1 and len(postBal)-1 (used for N-blocks) are syscalls pre and post all txs.
+	for txIndex := -1; txIndex < len(postBal); txIndex++ {
 		postVals := postBal[txIndex]
 
 		for addr, acct := range postVals {
@@ -1241,23 +1264,26 @@ func (s *StateDB) Copy() *StateDB {
 }
 
 func (s *StateDB) CopyState() *StateDB {
-	// Copy all the basic fields, initialize the memory ones
+	// Only copy all the basic fields for N-blocks state, initialize the memory ones
 	state := &StateDB{
-		db:                   s.db,
-		prefetcher:           s.prefetcher,
+		db: s.db,
+		// prefetcher:           s.prefetcher,  // don't need prefetcher for BAL
 		trie:                 s.trie,
 		reader:               s.reader,
 		originalRoot:         s.originalRoot,
 		stateObjects:         make(map[common.Address]*stateObject, len(s.stateObjects)),
 		stateObjectsDestruct: make(map[common.Address]*stateObject, len(s.stateObjectsDestruct)),
-		mutations:            make(map[common.Address]*mutation, len(s.mutations)),
-		dbErr:                s.dbErr,
-		refund:               s.refund,
-		thash:                s.thash,
-		txIndex:              s.txIndex,
-		logs:                 make(map[common.Hash][]*types.Log, len(s.logs)),
-		logSize:              s.logSize,
-		preimages:            maps.Clone(s.preimages),
+		// mutations:            make(map[common.Address]*mutation, len(s.mutations)),
+		mutations: make(map[common.Address]*mutation),
+		dbErr:     s.dbErr,
+		refund:    s.refund,
+		thash:     s.thash,
+		txIndex:   s.txIndex,
+		// logs:                 make(map[common.Hash][]*types.Log, len(s.logs)),
+		logs: make(map[common.Hash][]*types.Log),
+
+		logSize: s.logSize,
+		// preimages:            maps.Clone(s.preimages),
 
 		// Do we need to copy the access list and transient storage?
 		// In practice: No. At the start of a transaction, these two lists are empty.
@@ -1265,9 +1291,12 @@ func (s *StateDB) CopyState() *StateDB {
 		// in the middle of a transaction. However, it doesn't cost us much to copy
 		// empty lists, so we do it anyway to not blow up if we ever decide copy them
 		// in the middle of a transaction.
-		accessList:       s.accessList.Copy(),
-		transientStorage: s.transientStorage.Copy(),
-		journal:          s.journal.copy(),
+		// accessList:       s.accessList.Copy(),
+		// transientStorage: s.transientStorage.Copy(),
+		// journal:          s.journal.copy(),
+		accessList:       newAccessList(),
+		transientStorage: newTransientStorage(),
+		journal:          newJournal(),
 		// The update journal is not copies to avoid duplicated updates in later transactions.
 		blockNumber:   s.blockNumber,
 		updateJournal: newJournal(),
@@ -1277,9 +1306,9 @@ func (s *StateDB) CopyState() *StateDB {
 	if s.witness != nil {
 		state.witness = s.witness.Copy()
 	}
-	if s.accessEvents != nil {
-		state.accessEvents = s.accessEvents.Copy()
-	}
+	// if s.accessEvents != nil {
+	// 	state.accessEvents = s.accessEvents.Copy()
+	// }
 	// Deep copy cached state objects.
 	for addr, obj := range s.stateObjects {
 		state.stateObjects[addr] = obj.simpleCopy(state)
@@ -1289,18 +1318,19 @@ func (s *StateDB) CopyState() *StateDB {
 		state.stateObjectsDestruct[addr] = obj.simpleCopy(state)
 	}
 	// Deep copy the object state markers.
-	for addr, op := range s.mutations {
-		state.mutations[addr] = op.copy()
-	}
+	// for addr, op := range s.mutations {
+	// 	state.mutations[addr] = op.copy()
+	// }
 	// Deep copy the logs occurred in the scope of block
-	for hash, logs := range s.logs {
-		cpy := make([]*types.Log, len(logs))
-		for i, l := range logs {
-			cpy[i] = new(types.Log)
-			*cpy[i] = *l
-		}
-		state.logs[hash] = cpy
-	}
+	// for hash, logs := range s.logs {
+	// 	cpy := make([]*types.Log, len(logs))
+	// 	for i, l := range logs {
+	// 		cpy[i] = new(types.Log)
+	// 		*cpy[i] = *l
+	// 	}
+	// 	state.logs[hash] = cpy
+	// }
+
 	return state
 }
 
diff --git a/core/state/statedb_readerbal.go b/core/state/statedb_readerbal.go
index 37845e918f..263e37024a 100644
--- a/core/state/statedb_readerbal.go
+++ b/core/state/statedb_readerbal.go
@@ -28,6 +28,13 @@ func (s *StateDB) CodeSize(addr common.Address, codeHash common.Hash) (int, erro
 func (s *StateDB) Storage(addr common.Address, slot common.Hash) (common.Hash, error) {
 	stateObject := s.stateObjects[addr]
 	if stateObject != nil {
+		// Must use dirtyStorage and pendingStorage first, because statedb.MergePostBalStates will modify originStorage during N-blocks scenerio.
+		if value, cached := stateObject.dirtyStorage[slot]; cached {
+			return value, nil
+		}
+		if value, cached := stateObject.pendingStorage[slot]; cached {
+			return value, nil
+		}
 		if value, cached := stateObject.originStorage[slot]; cached {
 			return value, nil
 		}
diff --git a/core/state_processor.go b/core/state_processor.go
index ee98326467..a3ed60fe47 100644
--- a/core/state_processor.go
+++ b/core/state_processor.go
@@ -161,7 +161,7 @@ func ApplyTransactionWithEVM(msg *Message, gp *GasPool, statedb *state.StateDB,
 
 	// Merge the tx-local access event into the "block-local" one, in order to collect
 	// all values, so that the witness can be built.
-	if statedb.Database().TrieDB().IsVerkle() {
+	if statedb.Database() != nil && statedb.Database().TrieDB().IsVerkle() {
 		statedb.AccessEvents().Merge(evm.AccessEvents)
 	}
 	return MakeReceipt(evm, result, statedb, blockNumber, blockHash, blockTime, tx, *usedGas, root), nil