diff --git a/level_zero/core/test/unit_tests/sources/cmdlist/test_cmdlist_xe2_and_later.cpp b/level_zero/core/test/unit_tests/sources/cmdlist/test_cmdlist_xe2_and_later.cpp
index 83aa8bd2bf..0d0bb6f488 100644
--- a/level_zero/core/test/unit_tests/sources/cmdlist/test_cmdlist_xe2_and_later.cpp
+++ b/level_zero/core/test/unit_tests/sources/cmdlist/test_cmdlist_xe2_and_later.cpp
@@ -15,6 +15,7 @@
 
 #include "level_zero/core/source/cmdlist/cmdlist_hw.h"
 #include "level_zero/core/test/unit_tests/fixtures/cmdlist_fixture.h"
+#include "level_zero/core/test/unit_tests/fixtures/in_order_cmd_list_fixture.h"
 #include "level_zero/core/test/unit_tests/mocks/mock_cmdlist.h"
 #include "level_zero/core/test/unit_tests/mocks/mock_event.h"
 
@@ -383,5 +384,32 @@ HWTEST2_F(CommandListXe2AndLaterPreemptionTest,
     auto result = commandListCore->obtainKernelPreemptionMode(kernel.get());
     EXPECT_EQ(NEO::PreemptionMode::MidThread, result);
 }
+
+using InOrderCmdListTests = InOrderCmdListFixture;
+
+HWTEST2_F(InOrderCmdListTests, givenDebugFlagWhenPostSyncWithInOrderExecInfoIsCreateThenL1IsNotFlushed, Platforms) {
+    DebugManagerStateRestore restorer;
+    NEO::debugManager.flags.ForcePostSyncL1Flush.set(0);
+    using DefaultWalkerType = typename FamilyType::DefaultWalkerType;
+    using POSTSYNC_DATA = typename FamilyType::POSTSYNC_DATA;
+
+    auto immCmdList = createImmCmdList<gfxCoreFamily>();
+    auto cmdStream = immCmdList->getCmdContainer().getCommandStream();
+
+    immCmdList->appendLaunchKernel(kernel->toHandle(), groupCount, nullptr, 0, nullptr, launchParams, false);
+
+    GenCmdList cmdList;
+    ASSERT_TRUE(FamilyType::Parse::parseCommandBuffer(cmdList, cmdStream->getCpuBase(), cmdStream->getUsed()));
+
+    auto walkerItor = find<DefaultWalkerType *>(cmdList.begin(), cmdList.end());
+    ASSERT_NE(cmdList.end(), walkerItor);
+
+    auto walkerCmd = genCmdCast<DefaultWalkerType *>(*walkerItor);
+    auto &postSync = walkerCmd->getPostSync();
+
+    EXPECT_FALSE(postSync.getDataportPipelineFlush());
+    EXPECT_FALSE(postSync.getDataportSubsliceCacheFlush());
+}
+
 } // namespace ult
 } // namespace L0
diff --git a/opencl/source/command_queue/gpgpu_walker_xehp_and_later.inl b/opencl/source/command_queue/gpgpu_walker_xehp_and_later.inl
index fa5b0f6b98..199d021831 100644
--- a/opencl/source/command_queue/gpgpu_walker_xehp_and_later.inl
+++ b/opencl/source/command_queue/gpgpu_walker_xehp_and_later.inl
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2021-2023 Intel Corporation
+ * Copyright (C) 2021-2024 Intel Corporation
  *
  * SPDX-License-Identifier: MIT
  *
@@ -108,6 +108,11 @@ void GpgpuWalkerHelper<GfxFamily>::setupTimestampPacket(LinearStream *cmdStream,
     postSyncData.setDataportPipelineFlush(true);
     postSyncData.setDataportSubsliceCacheFlush(true);
 
+    if (NEO::debugManager.flags.ForcePostSyncL1Flush.get() != -1) {
+        postSyncData.setDataportPipelineFlush(!!NEO::debugManager.flags.ForcePostSyncL1Flush.get());
+        postSyncData.setDataportSubsliceCacheFlush(!!NEO::debugManager.flags.ForcePostSyncL1Flush.get());
+    }
+
     EncodeDispatchKernel<GfxFamily>::template setupPostSyncMocs<WalkerType>(*walkerCmd, rootDeviceEnvironment,
                                                                             MemorySynchronizationCommands<GfxFamily>::getDcFlushEnable(true, rootDeviceEnvironment));
 
diff --git a/opencl/test/unit_test/command_queue/dispatch_walker_tests_dg2_and_later.cpp b/opencl/test/unit_test/command_queue/dispatch_walker_tests_dg2_and_later.cpp
index f4e0bf7193..974e712f72 100644
--- a/opencl/test/unit_test/command_queue/dispatch_walker_tests_dg2_and_later.cpp
+++ b/opencl/test/unit_test/command_queue/dispatch_walker_tests_dg2_and_later.cpp
@@ -145,6 +145,42 @@ HWTEST2_F(Dg2AndLaterDispatchWalkerBasicTest, givenTimestampPacketWhenDispatchin
     EXPECT_EQ(contextStartAddress, secondWalker->getPostSync().getDestinationAddress());
 }
 
+HWTEST2_F(Dg2AndLaterDispatchWalkerBasicTest, givenDebugFlagToDisableL1FlushInPostSyncWhenKernelIsProgrammedThenL1FlushIsNotEnabled, matcherDG2AndLater) {
+    DebugManagerStateRestore restore;
+    NEO::debugManager.flags.ForcePostSyncL1Flush.set(0);
+
+    using DefaultWalkerType = typename FamilyType::DefaultWalkerType;
+
+    MockKernelWithInternals kernel1(*device);
+
+    device->getUltCommandStreamReceiver<FamilyType>().timestampPacketWriteEnabled = true;
+
+    TimestampPacketContainer timestampPacketContainer;
+    timestampPacketContainer.add(device->getGpgpuCommandStreamReceiver().getTimestampPacketAllocator()->getTag());
+    timestampPacketContainer.add(device->getGpgpuCommandStreamReceiver().getTimestampPacketAllocator()->getTag());
+
+    MockMultiDispatchInfo multiDispatchInfo(device.get(), std::vector<Kernel *>({kernel1.mockKernel}));
+
+    MockCommandQueue cmdQ(context.get(), device.get(), nullptr, false);
+    auto &cmdStream = cmdQ.getCS(0);
+
+    HardwareInterfaceWalkerArgs walkerArgs = createHardwareInterfaceWalkerArgs(CL_COMMAND_NDRANGE_KERNEL);
+    walkerArgs.currentTimestampPacketNodes = &timestampPacketContainer;
+    HardwareInterface<FamilyType>::template dispatchWalker<typename FamilyType::DefaultWalkerType>(
+        cmdQ,
+        multiDispatchInfo,
+        CsrDependencies(),
+        walkerArgs);
+
+    HardwareParse hwParser;
+    hwParser.parseCommands<FamilyType>(cmdStream, 0);
+    hwParser.findHardwareCommands<FamilyType>();
+    auto walker = genCmdCast<DefaultWalkerType *>(*hwParser.itorWalker);
+
+    EXPECT_FALSE(walker->getPostSync().getDataportPipelineFlush());
+    EXPECT_FALSE(walker->getPostSync().getDataportSubsliceCacheFlush());
+}
+
 HWTEST2_F(Dg2AndLaterDispatchWalkerBasicTest, givenDebugVariableEnabledWhenEnqueueingThenWriteWalkerStamp, matcherDG2AndLater) {
     using DefaultWalkerType = typename FamilyType::DefaultWalkerType;
     DebugManagerStateRestore restore;
diff --git a/shared/source/command_container/command_encoder_xehp_and_later.inl b/shared/source/command_container/command_encoder_xehp_and_later.inl
index 377e22e804..a466cbc8cf 100644
--- a/shared/source/command_container/command_encoder_xehp_and_later.inl
+++ b/shared/source/command_container/command_encoder_xehp_and_later.inl
@@ -438,6 +438,11 @@ void EncodeDispatchKernel<Family>::setupPostSyncForRegularEvent(WalkerType &walk
     postSync.setDataportPipelineFlush(true);
     postSync.setDataportSubsliceCacheFlush(true);
 
+    if (NEO::debugManager.flags.ForcePostSyncL1Flush.get() != -1) {
+        postSync.setDataportPipelineFlush(!!NEO::debugManager.flags.ForcePostSyncL1Flush.get());
+        postSync.setDataportSubsliceCacheFlush(!!NEO::debugManager.flags.ForcePostSyncL1Flush.get());
+    }
+
     auto operationType = POSTSYNC_DATA::OPERATION_WRITE_IMMEDIATE_DATA;
     uint64_t gpuVa = args.eventAddress;
     uint64_t immData = args.postSyncImmValue;
@@ -468,6 +473,10 @@ void EncodeDispatchKernel<Family>::setupPostSyncForInOrderExec(WalkerType &walke
 
     postSync.setDataportPipelineFlush(true);
     postSync.setDataportSubsliceCacheFlush(true);
+    if (NEO::debugManager.flags.ForcePostSyncL1Flush.get() != -1) {
+        postSync.setDataportPipelineFlush(!!NEO::debugManager.flags.ForcePostSyncL1Flush.get());
+        postSync.setDataportSubsliceCacheFlush(!!NEO::debugManager.flags.ForcePostSyncL1Flush.get());
+    }
 
     uint64_t gpuVa = args.inOrderExecInfo->getBaseDeviceAddress() + args.inOrderExecInfo->getAllocationOffset();
 
diff --git a/shared/source/debug_settings/debug_variables_base.inl b/shared/source/debug_settings/debug_variables_base.inl
index f889dc8498..9149dd04d0 100644
--- a/shared/source/debug_settings/debug_variables_base.inl
+++ b/shared/source/debug_settings/debug_variables_base.inl
@@ -393,6 +393,7 @@ DECLARE_DEBUG_VARIABLE(int32_t, EnableHostUsmAllocationPool, -1, "-1: default (e
 DECLARE_DEBUG_VARIABLE(int32_t, UseLocalPreferredForCacheableBuffers, -1, "Use localPreferred for cacheable buffers")
 DECLARE_DEBUG_VARIABLE(int32_t, EnableCopyWithStagingBuffers, -1, "Enable copy with non-usm memory through staging buffers. -1: default, 0: disabled, 1: enabled")
 DECLARE_DEBUG_VARIABLE(int32_t, StagingBufferSize, -1, "Size of single staging buffer. -1: default (2MB), >0: size in KB")
+DECLARE_DEBUG_VARIABLE(int32_t, ForcePostSyncL1Flush, -1, "-1: default (do nothing), 0: L1 flush disabled in post sync, 1: L1 flush enabled in post sync")
 
 /*DIRECT SUBMISSION FLAGS*/
 DECLARE_DEBUG_VARIABLE(int32_t, EnableDirectSubmission, -1, "-1: default (disabled), 0: disable, 1:enable. Enables direct submission of command buffers bypassing KMD")
diff --git a/shared/test/common/test_files/igdrcl.config b/shared/test/common/test_files/igdrcl.config
index 70888f0228..56b70900fd 100644
--- a/shared/test/common/test_files/igdrcl.config
+++ b/shared/test/common/test_files/igdrcl.config
@@ -616,4 +616,5 @@ EnableCopyWithStagingBuffers = -1
 StagingBufferSize = -1
 OverrideNumHighPriorityContexts = -1
 ForceScratchAndMTPBufferSizeMode = -1
+ForcePostSyncL1Flush = -1
 # Please don't edit below this line
diff --git a/shared/test/unit_test/encoders/test_encode_dispatch_kernel_dg2_and_later.cpp b/shared/test/unit_test/encoders/test_encode_dispatch_kernel_dg2_and_later.cpp
index 201ea0ae3f..c2a9fcc2b4 100644
--- a/shared/test/unit_test/encoders/test_encode_dispatch_kernel_dg2_and_later.cpp
+++ b/shared/test/unit_test/encoders/test_encode_dispatch_kernel_dg2_and_later.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2021-2023 Intel Corporation
+ * Copyright (C) 2021-2024 Intel Corporation
  *
  * SPDX-License-Identifier: MIT
  *
@@ -49,6 +49,33 @@ HWTEST2_F(CommandEncodeStatesTestDg2AndLater, givenEventAddressWhenEncodeAndPVCA
     EXPECT_EQ(true, cmd->getPostSync().getDataportSubsliceCacheFlush());
 }
 
+HWTEST2_F(CommandEncodeStatesTestDg2AndLater, givenDebugVariableToForceL1FlushWhenWalkerIsProgramedThenCacheFlushIsDisabled, IsAtLeastXeHpgCore) {
+    DebugManagerStateRestore restore;
+    NEO::debugManager.flags.ForcePostSyncL1Flush.set(0);
+    using POSTSYNC_DATA = typename FamilyType::POSTSYNC_DATA;
+    using DefaultWalkerType = typename FamilyType::DefaultWalkerType;
+    uint32_t dims[] = {2, 1, 1};
+    std::unique_ptr<MockDispatchKernelEncoder> dispatchInterface(new MockDispatchKernelEncoder());
+    uint64_t eventAddress = MemoryConstants::cacheLineSize * 123;
+
+    bool requiresUncachedMocs = false;
+    EncodeDispatchKernelArgs dispatchArgs = createDefaultDispatchKernelArgs(pDevice, dispatchInterface.get(), dims, requiresUncachedMocs);
+    dispatchArgs.eventAddress = eventAddress;
+    dispatchArgs.isTimestampEvent = true;
+
+    EncodeDispatchKernel<FamilyType>::template encode<DefaultWalkerType>(*cmdContainer.get(), dispatchArgs);
+
+    GenCmdList commands;
+    CmdParse<FamilyType>::parseCommandBuffer(commands, ptrOffset(cmdContainer->getCommandStream()->getCpuBase(), 0), cmdContainer->getCommandStream()->getUsed());
+
+    using DefaultWalkerType = typename FamilyType::DefaultWalkerType;
+    auto itor = find<DefaultWalkerType *>(commands.begin(), commands.end());
+    ASSERT_NE(itor, commands.end());
+    auto cmd = genCmdCast<DefaultWalkerType *>(*itor);
+    EXPECT_EQ(false, cmd->getPostSync().getDataportPipelineFlush());
+    EXPECT_EQ(false, cmd->getPostSync().getDataportSubsliceCacheFlush());
+}
+
 HWTEST2_F(CommandEncodeStatesTestDg2AndLater, givenEventAddressWhenEncodeThenMocsIndex2IsSet, IsXeHpgCore) {
     using POSTSYNC_DATA = typename FamilyType::POSTSYNC_DATA;
     using DefaultWalkerType = typename FamilyType::DefaultWalkerType;
diff --git a/shared/test/unit_test/encoders/test_encode_dispatch_kernel_pvc_and_later.cpp b/shared/test/unit_test/encoders/test_encode_dispatch_kernel_pvc_and_later.cpp
index 240e0ddad7..493781d49a 100644
--- a/shared/test/unit_test/encoders/test_encode_dispatch_kernel_pvc_and_later.cpp
+++ b/shared/test/unit_test/encoders/test_encode_dispatch_kernel_pvc_and_later.cpp
@@ -1,17 +1,19 @@
 /*
- * Copyright (C) 2021-2023 Intel Corporation
+ * Copyright (C) 2021-2024 Intel Corporation
  *
  * SPDX-License-Identifier: MIT
  *
  */
 
 #include "shared/source/command_stream/stream_properties.h"
+#include "shared/source/helpers/in_order_cmd_helpers.h"
 #include "shared/source/kernel/grf_config.h"
 #include "shared/source/os_interface/product_helper.h"
 #include "shared/test/common/cmd_parse/gen_cmd_parse.h"
 #include "shared/test/common/helpers/debug_manager_state_restore.h"
 #include "shared/test/common/mocks/mock_device.h"
 #include "shared/test/common/mocks/mock_execution_environment.h"
+#include "shared/test/common/mocks/mock_timestamp_container.h"
 #include "shared/test/common/test_macros/hw_test.h"
 #include "shared/test/common/test_macros/test.h"
 #include "shared/test/unit_test/fixtures/command_container_fixture.h"
@@ -97,3 +99,28 @@ HWCMDTEST_F(IGFX_XE_HP_CORE, CommandEncodeStatesTestPvcAndLater, givenCommandCon
     auto cmd = genCmdCast<STATE_COMPUTE_MODE *>(*itorCmd);
     EXPECT_EQ(productHelper.isGrfNumReportedWithScm(), cmd->getLargeGrfMode());
 }
+
+HWTEST2_F(CommandEncodeStatesTestPvcAndLater, givenDebugVariableWhenPostSyncIsProgrammedThenL1IsNotFlushed, IsAtLeastXeHpcCore) {
+    using INTERFACE_DESCRIPTOR_DATA = typename FamilyType::INTERFACE_DESCRIPTOR_DATA;
+    using DefaultWalkerType = typename FamilyType::DefaultWalkerType;
+    DebugManagerStateRestore restorer;
+    NEO::debugManager.flags.ForcePostSyncL1Flush.set(0);
+    uint32_t dims[] = {2, 1, 1};
+    std::unique_ptr<MockDispatchKernelEncoder> dispatchInterface(new MockDispatchKernelEncoder());
+
+    EncodeDispatchKernelArgs dispatchArgs = createDefaultDispatchKernelArgs(pDevice, dispatchInterface.get(), dims, false);
+
+    DefaultWalkerType walkerCmd = FamilyType::template getInitGpuWalker<DefaultWalkerType>();
+
+    MockTagAllocator<DeviceAllocNodeType<true>> deviceTagAllocator(0, pDevice->getMemoryManager());
+
+    auto inOrderExecInfo = InOrderExecInfo::create(deviceTagAllocator.getTag(), nullptr, *pDevice, 1, false);
+
+    dispatchArgs.inOrderExecInfo = inOrderExecInfo.get();
+
+    EncodeDispatchKernel<FamilyType>::template setupPostSyncForInOrderExec<DefaultWalkerType>(walkerCmd, dispatchArgs);
+
+    auto &postSyncData = walkerCmd.getPostSync();
+    EXPECT_FALSE(postSyncData.getDataportPipelineFlush());
+    EXPECT_FALSE(postSyncData.getDataportSubsliceCacheFlush());
+}
\ No newline at end of file