diff --git a/backends-velox/src/test/scala/org/apache/gluten/execution/VeloxTPCHSuite.scala b/backends-velox/src/test/scala/org/apache/gluten/execution/VeloxTPCHSuite.scala
index 22f96bbbc4c2..0e94c242c1db 100644
--- a/backends-velox/src/test/scala/org/apache/gluten/execution/VeloxTPCHSuite.scala
+++ b/backends-velox/src/test/scala/org/apache/gluten/execution/VeloxTPCHSuite.scala
@@ -255,7 +255,7 @@ class VeloxTPCHDistinctSpillSuite extends VeloxTPCHTableSupport {
     super.sparkConf
       .set("spark.memory.offHeap.size", "50m")
       .set("spark.gluten.memory.overAcquiredMemoryRatio", "0.9") // to trigger distinct spill early
-      .set(GlutenConfig.GLUTEN_COLUMNAR_TO_ROW_MEM_THRESHOLD_KEY, "8k")
+      .set(GlutenConfig.GLUTEN_COLUMNAR_TO_ROW_MEM_THRESHOLD.key, "8k")
   }
 
   test("distinct spill") {
diff --git a/cpp/core/config/GlutenConfig.h b/cpp/core/config/GlutenConfig.h
index e4f5a884b920..057d85930d2a 100644
--- a/cpp/core/config/GlutenConfig.h
+++ b/cpp/core/config/GlutenConfig.h
@@ -57,7 +57,6 @@ const std::string kGzipWindowSize4k = "4096";
 const std::string kParquetCompressionCodec = "spark.sql.parquet.compression.codec";
 
 const std::string kColumnarToRowMemoryThreshold = "spark.gluten.sql.columnarToRowMemoryThreshold";
-const std::string kColumnarToRowMemoryDefaultThreshold = "67108864"; // 64MB
 
 const std::string kUGIUserName = "spark.gluten.ugi.username";
 const std::string kUGITokens = "spark.gluten.ugi.tokens";
diff --git a/cpp/core/jni/JniWrapper.cc b/cpp/core/jni/JniWrapper.cc
index 5c2752f18ae7..4be5e9142818 100644
--- a/cpp/core/jni/JniWrapper.cc
+++ b/cpp/core/jni/JniWrapper.cc
@@ -534,19 +534,8 @@ Java_org_apache_gluten_vectorized_NativeColumnarToRowJniWrapper_nativeColumnarTo
   auto& conf = ctx->getConfMap();
   int64_t column2RowMemThreshold;
   auto it = conf.find(kColumnarToRowMemoryThreshold);
-  bool confIsLegal =
-      ((it == conf.end()) ? false : std::all_of(it->second.begin(), it->second.end(), [](unsigned char c) {
-        return std::isdigit(c);
-      }));
-  if (confIsLegal) {
-    column2RowMemThreshold = std::stoll(it->second);
-  } else {
-    LOG(INFO)
-        << "Because the spark.gluten.sql.columnarToRowMemoryThreshold configuration item is invalid, the kColumnarToRowMemoryDefaultThreshold default value is used, which is "
-        << kColumnarToRowMemoryDefaultThreshold << " byte";
-    column2RowMemThreshold = std::stoll(kColumnarToRowMemoryDefaultThreshold);
-  }
-
+  GLUTEN_CHECK(!(it == conf.end()), "Required key not found in runtime config: " + kColumnarToRowMemoryThreshold);
+  column2RowMemThreshold = std::stoll(it->second);
   // Convert the native batch to Spark unsafe row.
   return ctx->saveObject(ctx->createColumnar2RowConverter(column2RowMemThreshold));
   JNI_METHOD_END(kInvalidObjectHandle)
diff --git a/shims/common/src/main/scala/org/apache/gluten/GlutenConfig.scala b/shims/common/src/main/scala/org/apache/gluten/GlutenConfig.scala
index b1ef4be5cf63..0146c3604f25 100644
--- a/shims/common/src/main/scala/org/apache/gluten/GlutenConfig.scala
+++ b/shims/common/src/main/scala/org/apache/gluten/GlutenConfig.scala
@@ -587,9 +587,6 @@ object GlutenConfig {
 
   val GLUTEN_SHUFFLE_WRITER_MERGE_THRESHOLD = "spark.gluten.sql.columnar.shuffle.merge.threshold"
 
-  // Columnar to row memory threshold.
-  val GLUTEN_COLUMNAR_TO_ROW_MEM_THRESHOLD_KEY = "spark.gluten.sql.columnarToRowMemoryThreshold"
-
   // Controls whether to load DLL from jars. User can get dependent native libs packed into a jar
   // by executing dev/package.sh. Then, with that jar configured, Gluten can load the native libs
   // at runtime. This config is just for velox backend. And it is NOT applicable to the situation
@@ -654,7 +651,6 @@ object GlutenConfig {
       GLUTEN_SAVE_DIR,
       GLUTEN_TASK_OFFHEAP_SIZE_IN_BYTES_KEY,
       GLUTEN_MAX_BATCH_SIZE_KEY,
-      GLUTEN_COLUMNAR_TO_ROW_MEM_THRESHOLD_KEY,
       GLUTEN_SHUFFLE_WRITER_BUFFER_SIZE,
       SQLConf.SESSION_LOCAL_TIMEZONE.key,
       GLUTEN_DEFAULT_SESSION_TIMEZONE_KEY,
@@ -690,7 +686,10 @@ object GlutenConfig {
       (SQLConf.IGNORE_MISSING_FILES.key, SQLConf.IGNORE_MISSING_FILES.defaultValueString),
       (
         COLUMNAR_MEMORY_BACKTRACE_ALLOCATION.key,
-        COLUMNAR_MEMORY_BACKTRACE_ALLOCATION.defaultValueString)
+        COLUMNAR_MEMORY_BACKTRACE_ALLOCATION.defaultValueString),
+      (
+        GLUTEN_COLUMNAR_TO_ROW_MEM_THRESHOLD.key,
+        GLUTEN_COLUMNAR_TO_ROW_MEM_THRESHOLD.defaultValue.get.toString)
     )
     keyWithDefault.forEach(e => nativeConfMap.put(e._1, conf.getOrElse(e._1, e._2)))
 
@@ -1123,7 +1122,7 @@ object GlutenConfig {
       .createWithDefault(4096)
 
   val GLUTEN_COLUMNAR_TO_ROW_MEM_THRESHOLD =
-    buildConf(GLUTEN_COLUMNAR_TO_ROW_MEM_THRESHOLD_KEY)
+    buildConf("spark.gluten.sql.columnarToRowMemoryThreshold")
       .internal()
       .bytesConf(ByteUnit.BYTE)
       .createWithDefaultString("64MB")