diff --git a/server/v2/stf/branch/bench_test.go b/server/v2/stf/branch/bench_test.go
index db46846c1677..d1aa56a2d5ab 100644
--- a/server/v2/stf/branch/bench_test.go
+++ b/server/v2/stf/branch/bench_test.go
@@ -1,6 +1,7 @@
 package branch
 
 import (
+	"encoding/binary"
 	"fmt"
 	"testing"
 
@@ -14,29 +15,61 @@ var (
 )
 
 func Benchmark_CacheStack_Set(b *testing.B) {
+	var sink any
 	for _, stackSize := range stackSizes {
 		b.Run(fmt.Sprintf("StackSize%d", stackSize), func(b *testing.B) {
 			bs := makeBranchStack(b, stackSize)
 			b.ResetTimer()
 			b.ReportAllocs()
 			for i := 0; i < b.N; i++ {
-				_ = bs.Set([]byte{0}, []byte{0})
+				sink = bs.Set([]byte{0}, []byte{0})
 			}
 		})
 	}
+	if sink != nil {
+		b.Fatal("prevent compiler optimization")
+	}
 }
 
 func Benchmark_Get(b *testing.B) {
+	var sink any
 	for _, stackSize := range stackSizes {
 		b.Run(fmt.Sprintf("StackSize%d", stackSize), func(b *testing.B) {
 			bs := makeBranchStack(b, stackSize)
 			b.ResetTimer()
 			b.ReportAllocs()
 			for i := 0; i < b.N; i++ {
-				_, _ = bs.Get([]byte{0})
+				sink, _ = bs.Get([]byte{0})
+			}
+		})
+	}
+	if sink == nil {
+		b.Fatal("prevent compiler optimization")
+	}
+}
+
+func Benchmark_GetSparse(b *testing.B) {
+	var sink any
+	for _, stackSize := range stackSizes {
+		b.Run(fmt.Sprintf("StackSize%d", stackSize), func(b *testing.B) {
+			bs := makeBranchStack(b, stackSize)
+			keys := func() [][]byte {
+				var keys [][]byte
+				for i := 0; i < b.N; i++ {
+					keys = append(keys, numToBytes(i))
+				}
+				return keys
+			}()
+			b.ResetTimer()
+			b.ReportAllocs()
+			for _, key := range keys {
+				sink, _ = bs.Get(key)
 			}
 		})
 	}
+	if sink == nil {
+		b.Fatal("prevent compiler optimization")
+	}
 }
 
 func Benchmark_Iterate(b *testing.B) {
@@ -71,7 +104,7 @@ func makeBranchStack(b *testing.B, stackSize int) Store[store.KVStore] {
 		branch = NewStore[store.KVStore](branch)
 		for j := 0; j < elemsInStack; j++ {
 			// create unique keys by including the branch index.
-			key := []byte{byte(i), byte(j)}
+			key := append(numToBytes(i), numToBytes(j)...)
 			value := []byte{byte(j)}
 			err := branch.Set(key, value)
 			if err != nil {
@@ -81,3 +114,7 @@ func makeBranchStack(b *testing.B, stackSize int) Store[store.KVStore] {
 	}
 	return branch
 }
+
+func numToBytes[T ~int](n T) []byte {
+	return binary.BigEndian.AppendUint64(nil, uint64(n))
+}
diff --git a/server/v2/stf/branch/store.go b/server/v2/stf/branch/store.go
index 19e32e8392a3..16358d22f30f 100644
--- a/server/v2/stf/branch/store.go
+++ b/server/v2/stf/branch/store.go
@@ -29,8 +29,7 @@ func (s Store[T]) Get(key []byte) (value []byte, err error) {
 	if found {
 		return
 	}
-	// after we get it from parent store, we cache it.
-	// if it is not found in parent store, we still cache it as nil.
+	// if not found in the changeset, then check the parent.
 	value, err = s.parent.Get(key)
 	if err != nil {
 		return nil, err
diff --git a/store/cachekv/branch_bench_test.go b/store/cachekv/branch_bench_test.go
new file mode 100644
index 000000000000..cb1064e7c5c5
--- /dev/null
+++ b/store/cachekv/branch_bench_test.go
@@ -0,0 +1,116 @@
+package cachekv_test
+
+import (
+	"encoding/binary"
+	"fmt"
+	"testing"
+
+	coretesting "cosmossdk.io/core/testing"
+	"cosmossdk.io/store/cachekv"
+	"cosmossdk.io/store/dbadapter"
+)
+
+var (
+	stackSizes   = []int{1, 10, 100}
+	elemsInStack = 10
+)
+
+func Benchmark_CacheStack_Set(b *testing.B) {
+	for _, stackSize := range stackSizes {
+		b.Run(fmt.Sprintf("StackSize%d", stackSize), func(b *testing.B) {
+			bs := makeBranchStack(b, stackSize)
+			b.ResetTimer()
+			b.ReportAllocs()
+			for i := 0; i < b.N; i++ {
+				bs.Set([]byte{0}, []byte{0})
+			}
+		})
+	}
+}
+
+// Gets the same key from the branch store.
+func Benchmark_Get(b *testing.B) {
+	var sink any
+	for _, stackSize := range stackSizes {
+		b.Run(fmt.Sprintf("StackSize%d", stackSize), func(b *testing.B) {
+			bs := makeBranchStack(b, stackSize)
+			b.ResetTimer()
+			b.ReportAllocs()
+			for i := 0; i < b.N; i++ {
+				sink = bs.Get([]byte{0})
+			}
+		})
+	}
+	if sink == nil {
+		b.Fatal("prevent compiler optimization")
+	}
+}
+
+// Gets always different keys.
+func Benchmark_GetSparse(b *testing.B) {
+	var sink any
+	for _, stackSize := range stackSizes {
+		b.Run(fmt.Sprintf("StackSize%d", stackSize), func(b *testing.B) {
+			bs := makeBranchStack(b, stackSize)
+			keys := func() [][]byte {
+				var keys [][]byte
+				for i := 0; i < b.N; i++ {
+					keys = append(keys, numToBytes(i))
+				}
+				return keys
+			}()
+			b.ResetTimer()
+			b.ReportAllocs()
+			for _, key := range keys {
+				sink = bs.Get(key)
+			}
+		})
+	}
+	if sink == nil {
+		b.Fatal("prevent compiler optimization")
+	}
+}
+
+func Benchmark_Iterate(b *testing.B) {
+	var keySink, valueSink any
+
+	for _, stackSize := range stackSizes {
+		b.Run(fmt.Sprintf("StackSize%d", stackSize), func(b *testing.B) {
+			bs := makeBranchStack(b, stackSize)
+			b.ResetTimer()
+			b.ReportAllocs()
+			for i := 0; i < b.N; i++ {
+				iter := bs.Iterator(nil, nil)
+				for iter.Valid() {
+					keySink = iter.Key()
+					valueSink = iter.Value()
+					iter.Next()
+				}
+				_ = iter.Close()
+			}
+		})
+	}
+
+	_ = keySink
+	_ = valueSink
+}
+
+// makeBranchStack creates a branch stack of the given size and initializes it with unique key-value pairs.
+func makeBranchStack(_ *testing.B, stackSize int) *cachekv.Store {
+	parent := dbadapter.Store{DB: coretesting.NewMemDB()}
+	branch := cachekv.NewStore(parent)
+	for i := 1; i < stackSize; i++ {
+		branch = cachekv.NewStore(branch)
+		for j := 0; j < elemsInStack; j++ {
+			// create unique keys by including the branch index.
+			key := append(numToBytes(i), numToBytes(j)...)
+			value := []byte{byte(j)}
+			branch.Set(key, value)
+		}
+	}
+	return branch
+}
+
+func numToBytes[T ~int](n T) []byte {
+	return binary.BigEndian.AppendUint64(nil, uint64(n))
+}