diff --git a/runtime/command_stream/command_stream_receiver_hw_bdw_plus.inl b/runtime/command_stream/command_stream_receiver_hw_bdw_plus.inl
index 74c8c3978f..d64fe6d463 100644
--- a/runtime/command_stream/command_stream_receiver_hw_bdw_plus.inl
+++ b/runtime/command_stream/command_stream_receiver_hw_bdw_plus.inl
@@ -48,7 +48,7 @@ inline size_t CommandStreamReceiverHw<GfxFamily>::getCmdSizeForL3Config() const
 template <typename GfxFamily>
 void CommandStreamReceiverHw<GfxFamily>::programPipelineSelect(LinearStream &commandStream, DispatchFlags &dispatchFlags) {
     if (csrSizeRequestFlags.mediaSamplerConfigChanged || !isPreambleSent) {
-        PreambleHelper<GfxFamily>::programPipelineSelect(&commandStream, dispatchFlags);
+        PreambleHelper<GfxFamily>::programPipelineSelect(&commandStream, dispatchFlags, peekHwInfo());
         this->lastMediaSamplerConfig = dispatchFlags.mediaSamplerRequired;
     }
 }
diff --git a/runtime/gen11/preamble_gen11.cpp b/runtime/gen11/preamble_gen11.cpp
index b90e77b5c8..eae2dc5fdd 100644
--- a/runtime/gen11/preamble_gen11.cpp
+++ b/runtime/gen11/preamble_gen11.cpp
@@ -28,13 +28,19 @@ uint32_t PreambleHelper<ICLFamily>::getL3Config(const HardwareInfo &hwInfo, bool
 }
 
 template <>
-void PreambleHelper<ICLFamily>::programPipelineSelect(LinearStream *pCommandStream, const DispatchFlags &dispatchFlags) {
+void PreambleHelper<ICLFamily>::programPipelineSelect(LinearStream *pCommandStream,
+                                                      const DispatchFlags &dispatchFlags,
+                                                      const HardwareInfo &hwInfo) {
+
     typedef typename ICLFamily::PIPELINE_SELECT PIPELINE_SELECT;
 
     auto pCmd = (PIPELINE_SELECT *)pCommandStream->getSpace(sizeof(PIPELINE_SELECT));
     *pCmd = ICLFamily::cmdInitPipelineSelect;
 
-    auto mask = pipelineSelectEnablePipelineSelectMaskBits | pipelineSelectMediaSamplerDopClockGateMaskBits | pipelineSelectMediaSamplerPowerClockGateMaskBits;
+    auto mask = pipelineSelectEnablePipelineSelectMaskBits |
+                pipelineSelectMediaSamplerDopClockGateMaskBits |
+                pipelineSelectMediaSamplerPowerClockGateMaskBits;
+
     pCmd->setMaskBits(mask);
     pCmd->setPipelineSelection(PIPELINE_SELECT::PIPELINE_SELECTION_GPGPU);
     pCmd->setMediaSamplerDopClockGateEnable(!dispatchFlags.mediaSamplerRequired);
diff --git a/runtime/gen8/preamble_gen8.cpp b/runtime/gen8/preamble_gen8.cpp
index 28cac70bae..f4b84c3595 100644
--- a/runtime/gen8/preamble_gen8.cpp
+++ b/runtime/gen8/preamble_gen8.cpp
@@ -32,7 +32,10 @@ uint32_t PreambleHelper<BDWFamily>::getL3Config(const HardwareInfo &hwInfo, bool
 }
 
 template <>
-void PreambleHelper<BDWFamily>::programPipelineSelect(LinearStream *pCommandStream, const DispatchFlags &dispatchFlags) {
+void PreambleHelper<BDWFamily>::programPipelineSelect(LinearStream *pCommandStream,
+                                                      const DispatchFlags &dispatchFlags,
+                                                      const HardwareInfo &hwInfo) {
+
     typedef typename BDWFamily::PIPELINE_SELECT PIPELINE_SELECT;
     auto pCmd = (PIPELINE_SELECT *)pCommandStream->getSpace(sizeof(PIPELINE_SELECT));
     *pCmd = BDWFamily::cmdInitPipelineSelect;
diff --git a/runtime/gen9/preamble_gen9.cpp b/runtime/gen9/preamble_gen9.cpp
index 287441843d..9bf85e8807 100644
--- a/runtime/gen9/preamble_gen9.cpp
+++ b/runtime/gen9/preamble_gen9.cpp
@@ -28,7 +28,10 @@ uint32_t PreambleHelper<SKLFamily>::getL3Config(const HardwareInfo &hwInfo, bool
 }
 
 template <>
-void PreambleHelper<SKLFamily>::programPipelineSelect(LinearStream *pCommandStream, const DispatchFlags &dispatchFlags) {
+void PreambleHelper<SKLFamily>::programPipelineSelect(LinearStream *pCommandStream,
+                                                      const DispatchFlags &dispatchFlags,
+                                                      const HardwareInfo &hwInfo) {
+
     typedef typename SKLFamily::PIPELINE_SELECT PIPELINE_SELECT;
 
     auto pCmd = (PIPELINE_SELECT *)pCommandStream->getSpace(sizeof(PIPELINE_SELECT));
diff --git a/runtime/helpers/hardware_commands_helper.h b/runtime/helpers/hardware_commands_helper.h
index 329e14ac2d..6ec823c9da 100644
--- a/runtime/helpers/hardware_commands_helper.h
+++ b/runtime/helpers/hardware_commands_helper.h
@@ -143,6 +143,7 @@ struct HardwareCommandsHelper : public PerThreadDataHelper {
     static size_t getSizeRequiredCS(const Kernel *kernel);
     static size_t getSizeRequiredForCacheFlush(const CommandQueue &commandQueue, const Kernel *kernel, uint64_t postSyncAddress);
     static bool isPipeControlWArequired(const HardwareInfo &hwInfo);
+    static bool isPipeControlPriorToPipelineSelectWArequired(const HardwareInfo &hwInfo);
     static size_t getSizeRequiredDSH(
         const Kernel &kernel);
     static size_t getSizeRequiredIOH(
diff --git a/runtime/helpers/hardware_commands_helper.inl b/runtime/helpers/hardware_commands_helper.inl
index dbfcb941cd..969f29c0d0 100644
--- a/runtime/helpers/hardware_commands_helper.inl
+++ b/runtime/helpers/hardware_commands_helper.inl
@@ -25,6 +25,11 @@ namespace NEO {
 template <typename GfxFamily>
 bool HardwareCommandsHelper<GfxFamily>::isPipeControlWArequired(const HardwareInfo &hwInfo) { return false; }
 
+template <typename GfxFamily>
+bool HardwareCommandsHelper<GfxFamily>::isPipeControlPriorToPipelineSelectWArequired(const HardwareInfo &hwInfo) {
+    return false;
+}
+
 template <typename GfxFamily>
 uint32_t HardwareCommandsHelper<GfxFamily>::computeSlmValues(uint32_t valueIn) {
     auto value = std::max(valueIn, 1024u);
diff --git a/runtime/helpers/preamble.h b/runtime/helpers/preamble.h
index 359a6f6471..caeb028286 100644
--- a/runtime/helpers/preamble.h
+++ b/runtime/helpers/preamble.h
@@ -27,7 +27,9 @@ struct PreambleHelper {
     using PIPE_CONTROL = typename GfxFamily::PIPE_CONTROL;
 
     static void programL3(LinearStream *pCommandStream, uint32_t l3Config);
-    static void programPipelineSelect(LinearStream *pCommandStream, const DispatchFlags &dispatchFlags);
+    static void programPipelineSelect(LinearStream *pCommandStream,
+                                      const DispatchFlags &dispatchFlags,
+                                      const HardwareInfo &hwInfo);
     static uint32_t getDefaultThreadArbitrationPolicy();
     static void programThreadArbitration(LinearStream *pCommandStream, uint32_t requiredThreadArbitrationPolicy);
     static void programPreemption(LinearStream *pCommandStream, Device &device, GraphicsAllocation *preemptionCsr);
diff --git a/unit_tests/command_stream/command_stream_receiver_flush_task_2_tests.cpp b/unit_tests/command_stream/command_stream_receiver_flush_task_2_tests.cpp
index a5e7fe3685..5ff35e74b4 100644
--- a/unit_tests/command_stream/command_stream_receiver_flush_task_2_tests.cpp
+++ b/unit_tests/command_stream/command_stream_receiver_flush_task_2_tests.cpp
@@ -9,6 +9,7 @@
 #include "runtime/command_stream/csr_definitions.h"
 #include "runtime/command_stream/scratch_space_controller.h"
 #include "runtime/gmm_helper/gmm_helper.h"
+#include "runtime/helpers/hardware_commands_helper.h"
 #include "runtime/helpers/hw_helper.h"
 #include "runtime/helpers/state_base_address.h"
 #include "runtime/memory_manager/internal_allocation_storage.h"
@@ -983,8 +984,29 @@ HWTEST_F(CommandStreamReceiverFlushTaskTests, givenCsrWhenPreambleNotSentThenReq
     EXPECT_EQ(mediaSamplerConfigChangedSize, mediaSamplerConfigNotChangedSize);
 }
 
+HWTEST_F(CommandStreamReceiverFlushTaskTests, givenSpecialPipelineSelectModeChangedWhenGetCmdSizeForPielineSelectIsCalledThenCorrectSizeIsReturned) {
+    using PIPELINE_SELECT = typename FamilyType::PIPELINE_SELECT;
+    using PIPE_CONTROL = typename FamilyType::PIPE_CONTROL;
+
+    UltCommandStreamReceiver<FamilyType> &commandStreamReceiver = (UltCommandStreamReceiver<FamilyType> &)pDevice->getGpgpuCommandStreamReceiver();
+
+    CsrSizeRequestFlags csrSizeRequest = {};
+    DispatchFlags flags;
+    csrSizeRequest.specialPipelineSelectModeChanged = true;
+    commandStreamReceiver.overrideCsrSizeReqFlags(csrSizeRequest);
+    size_t size = commandStreamReceiver.getCmdSizeForPipelineSelect();
+
+    size_t expectedSize = sizeof(PIPELINE_SELECT);
+    if (HardwareCommandsHelper<FamilyType>::isPipeControlPriorToPipelineSelectWArequired(pDevice->getHardwareInfo())) {
+        expectedSize += sizeof(PIPE_CONTROL);
+    }
+    EXPECT_EQ(expectedSize, size);
+}
+
 HWTEST_F(CommandStreamReceiverFlushTaskTests, givenCsrWhenPreambleSentThenRequiredCsrSizeDependsOnmediaSamplerConfigChanged) {
-    typedef typename FamilyType::PIPELINE_SELECT PIPELINE_SELECT;
+    using PIPELINE_SELECT = typename FamilyType::PIPELINE_SELECT;
+    using PIPE_CONTROL = typename FamilyType::PIPE_CONTROL;
+
     UltCommandStreamReceiver<FamilyType> &commandStreamReceiver = (UltCommandStreamReceiver<FamilyType> &)pDevice->getGpgpuCommandStreamReceiver();
     CsrSizeRequestFlags csrSizeRequest = {};
     DispatchFlags flags;
@@ -1000,7 +1022,13 @@ HWTEST_F(CommandStreamReceiverFlushTaskTests, givenCsrWhenPreambleSentThenRequir
 
     EXPECT_NE(mediaSamplerConfigChangedSize, mediaSamplerConfigNotChangedSize);
     auto difference = mediaSamplerConfigChangedSize - mediaSamplerConfigNotChangedSize;
-    EXPECT_EQ(sizeof(PIPELINE_SELECT), difference);
+
+    size_t expectedDifference = sizeof(PIPELINE_SELECT);
+    if (HardwareCommandsHelper<FamilyType>::isPipeControlPriorToPipelineSelectWArequired(pDevice->getHardwareInfo())) {
+        expectedDifference += sizeof(PIPE_CONTROL);
+    }
+
+    EXPECT_EQ(expectedDifference, difference);
 }
 
 HWTEST_F(CommandStreamReceiverFlushTaskTests, givenCsrWhenSamplerCacheFlushSentThenRequiredCsrSizeContainsPipecontrolSize) {