diff --git a/opencl/test/unit_test/helpers/hw_helper_tests_pvc_and_later.cpp b/opencl/test/unit_test/helpers/hw_helper_tests_pvc_and_later.cpp
index a857181b2d..ca36e803e3 100644
--- a/opencl/test/unit_test/helpers/hw_helper_tests_pvc_and_later.cpp
+++ b/opencl/test/unit_test/helpers/hw_helper_tests_pvc_and_later.cpp
@@ -86,6 +86,12 @@ HWTEST2_F(HwHelperTestPvcAndLater, GivenVariousValuesWhenCallingGetBarriersCount
     EXPECT_EQ(32u, hwHelper.getBarriersCountFromHasBarriers(7u));
 }
 
+HWTEST2_F(HwHelperTestPvcAndLater, givenHwHelperWhenCheckIsUpdateTaskCountFromWaitSupportedThenReturnsTrue, IsAtLeastXeHpcCore) {
+    auto &hwHelper = HwHelper::get(hardwareInfo.platform.eRenderCoreFamily);
+
+    EXPECT_TRUE(hwHelper.isUpdateTaskCountFromWaitSupported());
+}
+
 HWTEST2_F(HwHelperTestPvcAndLater, givenCooperativeContextSupportedWhenGetEngineInstancesThenReturnCorrectAmountOfCooperativeCcs, IsAtLeastXeHpcCore) {
     HardwareInfo hwInfo = *defaultHwInfo;
     hwInfo.gtSystemInfo.CCSInfo.NumberOfCCSEnabled = 2;
diff --git a/opencl/test/unit_test/os_interface/linux/drm_command_stream_tests_1.cpp b/opencl/test/unit_test/os_interface/linux/drm_command_stream_tests_1.cpp
index f38a200aac..ecc208c9f9 100644
--- a/opencl/test/unit_test/os_interface/linux/drm_command_stream_tests_1.cpp
+++ b/opencl/test/unit_test/os_interface/linux/drm_command_stream_tests_1.cpp
@@ -931,6 +931,33 @@ HWTEST_TEMPLATED_F(DrmCommandStreamTest, givenLocalMemoryEnabledWhenCreatingDrmC
     {
         DebugManagerStateRestore restore;
         DebugManager.flags.EnableLocalMemory.set(1);
+        DebugManager.flags.UpdateTaskCountFromWait.set(3);
+
+        MockDrmCsr<FamilyType> csr1(executionEnvironment, 0, 1, gemCloseWorkerMode::gemCloseWorkerInactive);
+        EXPECT_EQ(DispatchMode::ImmediateDispatch, csr1.dispatchMode);
+
+        DebugManager.flags.CsrDispatchMode.set(static_cast<int32_t>(DispatchMode::BatchedDispatch));
+        MockDrmCsr<FamilyType> csr2(executionEnvironment, 0, 1, gemCloseWorkerMode::gemCloseWorkerInactive);
+        EXPECT_EQ(DispatchMode::BatchedDispatch, csr2.dispatchMode);
+    }
+
+    {
+        DebugManagerStateRestore restore;
+        DebugManager.flags.EnableLocalMemory.set(0);
+        DebugManager.flags.UpdateTaskCountFromWait.set(3);
+
+        MockDrmCsr<FamilyType> csr1(executionEnvironment, 0, 1, gemCloseWorkerMode::gemCloseWorkerInactive);
+        EXPECT_EQ(DispatchMode::ImmediateDispatch, csr1.dispatchMode);
+
+        DebugManager.flags.CsrDispatchMode.set(static_cast<int32_t>(DispatchMode::BatchedDispatch));
+        MockDrmCsr<FamilyType> csr2(executionEnvironment, 0, 1, gemCloseWorkerMode::gemCloseWorkerInactive);
+        EXPECT_EQ(DispatchMode::BatchedDispatch, csr2.dispatchMode);
+    }
+
+    {
+        DebugManagerStateRestore restore;
+        DebugManager.flags.EnableLocalMemory.set(1);
+        DebugManager.flags.UpdateTaskCountFromWait.set(0);
 
         MockDrmCsr<FamilyType> csr1(executionEnvironment, 0, 1, gemCloseWorkerMode::gemCloseWorkerInactive);
         EXPECT_EQ(DispatchMode::BatchedDispatch, csr1.dispatchMode);
@@ -943,6 +970,7 @@ HWTEST_TEMPLATED_F(DrmCommandStreamTest, givenLocalMemoryEnabledWhenCreatingDrmC
     {
         DebugManagerStateRestore restore;
         DebugManager.flags.EnableLocalMemory.set(0);
+        DebugManager.flags.UpdateTaskCountFromWait.set(0);
 
         MockDrmCsr<FamilyType> csr1(executionEnvironment, 0, 1, gemCloseWorkerMode::gemCloseWorkerInactive);
         EXPECT_EQ(DispatchMode::ImmediateDispatch, csr1.dispatchMode);
diff --git a/shared/source/command_stream/command_stream_receiver_hw_base.inl b/shared/source/command_stream/command_stream_receiver_hw_base.inl
index e1aa01899d..3e550bb184 100644
--- a/shared/source/command_stream/command_stream_receiver_hw_base.inl
+++ b/shared/source/command_stream/command_stream_receiver_hw_base.inl
@@ -187,12 +187,14 @@ CompletionStamp CommandStreamReceiverHw<GfxFamily>::flushTask(
 
     bool updateTag = false;
     if (dispatchFlags.blocking || dispatchFlags.dcFlush || dispatchFlags.guardCommandBufferWithPipeControl) {
+        updateTag = !isUpdateTagFromWaitEnabled();
+
         if (this->dispatchMode == DispatchMode::ImmediateDispatch) {
             //for ImmediateDispatch we will send this right away, therefore this pipe control will close the level
             //for BatchedSubmissions it will be nooped and only last ppc in batch will be emitted.
             levelClosed = true;
             //if we guard with ppc, flush dc as well to speed up completion latency
-            if (dispatchFlags.guardCommandBufferWithPipeControl) {
+            if (dispatchFlags.guardCommandBufferWithPipeControl && updateTag) {
                 dispatchFlags.dcFlush = true;
             }
         }
@@ -206,7 +208,6 @@ CompletionStamp CommandStreamReceiverHw<GfxFamily>::flushTask(
 
         auto address = getTagAllocation()->getGpuAddress();
 
-        updateTag = !isUpdateTagFromWaitEnabled();
         updateTag |= dispatchFlags.blocking;
         updateTag |= dispatchFlags.dcFlush;
 
@@ -1311,7 +1312,9 @@ inline void CommandStreamReceiverHw<GfxFamily>::flushHandler(BatchBuffer &batchB
 
 template <typename GfxFamily>
 inline bool CommandStreamReceiverHw<GfxFamily>::isUpdateTagFromWaitEnabled() {
-    bool enabled = false;
+    auto &hwHelper = HwHelper::get(peekHwInfo().platform.eRenderCoreFamily);
+    auto enabled = hwHelper.isUpdateTaskCountFromWaitSupported();
+    enabled &= this->isAnyDirectSubmissionEnabled();
 
     switch (DebugManager.flags.UpdateTaskCountFromWait.get()) {
     case 0:
diff --git a/shared/source/helpers/hw_helper.h b/shared/source/helpers/hw_helper.h
index 834e4926de..e6bea92da8 100644
--- a/shared/source/helpers/hw_helper.h
+++ b/shared/source/helpers/hw_helper.h
@@ -65,6 +65,7 @@ class HwHelper {
     static bool cacheFlushAfterWalkerSupported(const HardwareInfo &hwInfo);
     virtual bool timestampPacketWriteSupported() const = 0;
     virtual bool isTimestampWaitSupported() const = 0;
+    virtual bool isUpdateTaskCountFromWaitSupported() const = 0;
     virtual size_t getRenderSurfaceStateSize() const = 0;
     virtual void setRenderSurfaceStateForBuffer(const RootDeviceEnvironment &rootDeviceEnvironment,
                                                 void *surfaceStateBuffer,
@@ -242,6 +243,8 @@ class HwHelperHw : public HwHelper {
 
     bool isTimestampWaitSupported() const override;
 
+    bool isUpdateTaskCountFromWaitSupported() const override;
+
     bool is1MbAlignmentSupported(const HardwareInfo &hwInfo, bool isCompressionEnabled) const override;
 
     bool isFenceAllocationRequired(const HardwareInfo &hwInfo) const override;
diff --git a/shared/source/helpers/hw_helper_bdw_and_later.inl b/shared/source/helpers/hw_helper_bdw_and_later.inl
index 624933ef0e..cb01bc28b7 100644
--- a/shared/source/helpers/hw_helper_bdw_and_later.inl
+++ b/shared/source/helpers/hw_helper_bdw_and_later.inl
@@ -45,6 +45,11 @@ bool HwHelperHw<GfxFamily>::isTimestampWaitSupported() const {
     return false;
 }
 
+template <typename GfxFamily>
+bool HwHelperHw<GfxFamily>::isUpdateTaskCountFromWaitSupported() const {
+    return false;
+}
+
 template <typename GfxFamily>
 bool HwHelperHw<GfxFamily>::isAssignEngineRoundRobinSupported() const {
     return false;
diff --git a/shared/source/helpers/hw_helper_pvc_and_later.inl b/shared/source/helpers/hw_helper_pvc_and_later.inl
index f144834ab5..0ece986b7f 100644
--- a/shared/source/helpers/hw_helper_pvc_and_later.inl
+++ b/shared/source/helpers/hw_helper_pvc_and_later.inl
@@ -41,6 +41,11 @@ bool HwHelperHw<Family>::isTimestampWaitSupported() const {
     return true;
 }
 
+template <>
+bool HwHelperHw<Family>::isUpdateTaskCountFromWaitSupported() const {
+    return true;
+}
+
 template <>
 uint32_t HwHelperHw<Family>::adjustMaxWorkGroupCount(uint32_t maxWorkGroupCount, const EngineGroupType engineGroupType,
                                                      const HardwareInfo &hwInfo, bool isEngineInstanced) const {
diff --git a/shared/source/helpers/hw_helper_xehp_and_later.inl b/shared/source/helpers/hw_helper_xehp_and_later.inl
index acd43d20ab..80012b41c8 100644
--- a/shared/source/helpers/hw_helper_xehp_and_later.inl
+++ b/shared/source/helpers/hw_helper_xehp_and_later.inl
@@ -59,6 +59,11 @@ bool HwHelperHw<GfxFamily>::isTimestampWaitSupported() const {
     return false;
 }
 
+template <typename GfxFamily>
+bool HwHelperHw<GfxFamily>::isUpdateTaskCountFromWaitSupported() const {
+    return false;
+}
+
 template <typename GfxFamily>
 const EngineInstancesContainer HwHelperHw<GfxFamily>::getGpgpuEngineInstances(const HardwareInfo &hwInfo) const {
     auto defaultEngine = getChosenEngineType(hwInfo);
diff --git a/shared/source/os_interface/linux/drm_command_stream.inl b/shared/source/os_interface/linux/drm_command_stream.inl
index 1f65df931f..5f1dd6247c 100644
--- a/shared/source/os_interface/linux/drm_command_stream.inl
+++ b/shared/source/os_interface/linux/drm_command_stream.inl
@@ -55,7 +55,7 @@ DrmCommandStreamReceiver<GfxFamily>::DrmCommandStreamReceiver(ExecutionEnvironme
     auto hwInfo = rootDeviceEnvironment->getHardwareInfo();
     auto localMemoryEnabled = HwHelper::get(hwInfo->platform.eRenderCoreFamily).getEnableLocalMemory(*hwInfo);
 
-    this->dispatchMode = localMemoryEnabled ? DispatchMode::BatchedDispatch : DispatchMode::ImmediateDispatch;
+    this->dispatchMode = localMemoryEnabled && !this->isUpdateTagFromWaitEnabled() ? DispatchMode::BatchedDispatch : DispatchMode::ImmediateDispatch;
 
     if (DebugManager.flags.CsrDispatchMode.get()) {
         this->dispatchMode = static_cast<DispatchMode>(DebugManager.flags.CsrDispatchMode.get());
diff --git a/shared/test/unit_test/command_stream/command_stream_receiver_tests.cpp b/shared/test/unit_test/command_stream/command_stream_receiver_tests.cpp
index fa6526cf7b..92b4e6c712 100644
--- a/shared/test/unit_test/command_stream/command_stream_receiver_tests.cpp
+++ b/shared/test/unit_test/command_stream/command_stream_receiver_tests.cpp
@@ -388,6 +388,21 @@ HWTEST_F(CommandStreamReceiverTest, givenUpdateTaskCountFromWaitWhenCheckTaskCou
     }
 }
 
+HWTEST_F(CommandStreamReceiverTest, givenUpdateTaskCountFromWaitWhenCheckIfEnabledThenCanBeEnabledOnlyWithDirectSubmission) {
+    auto &csr = pDevice->getUltCommandStreamReceiver<FamilyType>();
+    auto &hwHelper = HwHelper::get(csr.peekHwInfo().platform.eRenderCoreFamily);
+
+    {
+        csr.directSubmissionAvailable = true;
+        EXPECT_EQ(csr.isUpdateTagFromWaitEnabled(), hwHelper.isUpdateTaskCountFromWaitSupported());
+    }
+
+    {
+        csr.directSubmissionAvailable = false;
+        EXPECT_FALSE(csr.isUpdateTagFromWaitEnabled());
+    }
+}
+
 struct InitDirectSubmissionFixture {
     void SetUp() {
         DebugManager.flags.EnableDirectSubmission.set(1);