diff --git a/shared/source/command_container/encode_alu_helper.h b/shared/source/command_container/encode_alu_helper.h
index 8414ef122f..8fa0d20bbb 100644
--- a/shared/source/command_container/encode_alu_helper.h
+++ b/shared/source/command_container/encode_alu_helper.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2022 Intel Corporation
+ * Copyright (C) 2022-2023 Intel Corporation
  *
  * SPDX-License-Identifier: MIT
  *
@@ -24,6 +24,12 @@ class EncodeAluHelper {
         aluOps.miMath.DW0.BitField.DwordLength = AluCount - 1;
     }
 
+    void setMocs([[maybe_unused]] uint32_t mocs) {
+        if constexpr (GfxFamily::isUsingMiMathMocs) {
+            aluOps.miMath.DW0.BitField.MemoryObjectControlState = mocs;
+        }
+    }
+
     void setNextAlu(AluRegisters opcode) {
         setNextAlu(opcode, AluRegisters::OPCODE_NONE, AluRegisters::OPCODE_NONE);
     }
diff --git a/shared/source/direct_submission/direct_submission_hw.inl b/shared/source/direct_submission/direct_submission_hw.inl
index 93c4bdde76..922207be13 100644
--- a/shared/source/direct_submission/direct_submission_hw.inl
+++ b/shared/source/direct_submission/direct_submission_hw.inl
@@ -14,6 +14,8 @@
 #include "shared/source/direct_submission/direct_submission_hw_diagnostic_mode.h"
 #include "shared/source/direct_submission/relaxed_ordering_helper.h"
 #include "shared/source/execution_environment/root_device_environment.h"
+#include "shared/source/gmm_helper/gmm_helper.h"
+#include "shared/source/gmm_helper/gmm_lib.h"
 #include "shared/source/helpers/aligned_memory.h"
 #include "shared/source/helpers/flush_stamp.h"
 #include "shared/source/helpers/gfx_core_helper.h"
@@ -102,6 +104,8 @@ void DirectSubmissionHw<GfxFamily, Dispatcher>::dispatchStaticRelaxedOrderingSch
 
     uint64_t loopSectionStartAddress = schedulerStartAddress + RelaxedOrderingHelper::StaticSchedulerSizeAndOffsetSection<GfxFamily>::loopStartSectionStart;
 
+    const uint32_t miMathMocs = this->rootDeviceEnvironment.getGmmHelper()->getMOCS(GMM_RESOURCE_USAGE_OCL_BUFFER);
+
     // 1. Init section
     {
         EncodeMiPredicate<GfxFamily>::encode(schedulerCmdStream, MiPredicateType::Disable);
@@ -133,6 +137,7 @@ void DirectSubmissionHw<GfxFamily, Dispatcher>::dispatchStaticRelaxedOrderingSch
         LriHelper<GfxFamily>::program(&schedulerCmdStream, CS_GPR_R8 + 4, static_cast<uint32_t>(deferredTasksListGpuVa >> 32), true);
 
         EncodeAluHelper<GfxFamily, 10> aluHelper;
+        aluHelper.setMocs(miMathMocs);
         aluHelper.setNextAlu(AluRegisters::OPCODE_LOAD, AluRegisters::R_SRCA, AluRegisters::R_2);
         aluHelper.setNextAlu(AluRegisters::OPCODE_LOAD, AluRegisters::R_SRCB, AluRegisters::R_6);
         aluHelper.setNextAlu(AluRegisters::OPCODE_SHL);
@@ -168,6 +173,7 @@ void DirectSubmissionHw<GfxFamily, Dispatcher>::dispatchStaticRelaxedOrderingSch
         LriHelper<GfxFamily>::program(&schedulerCmdStream, CS_GPR_R8 + 4, static_cast<uint32_t>(deferredTasksListGpuVa >> 32), true);
 
         EncodeAluHelper<GfxFamily, 14> aluHelper;
+        aluHelper.setMocs(miMathMocs);
         aluHelper.setNextAlu(AluRegisters::OPCODE_LOAD, AluRegisters::R_SRCA, AluRegisters::R_1);
         aluHelper.setNextAlu(AluRegisters::OPCODE_LOAD, AluRegisters::R_SRCB, AluRegisters::R_7);
         aluHelper.setNextAlu(AluRegisters::OPCODE_SHL);
@@ -238,6 +244,7 @@ void DirectSubmissionHw<GfxFamily, Dispatcher>::dispatchStaticRelaxedOrderingSch
     LriHelper<GfxFamily>::program(&schedulerCmdStream, CS_GPR_R10 + 4, 0, true);
 
     EncodeAluHelper<GfxFamily, 4> aluHelper;
+    aluHelper.setMocs(miMathMocs);
     aluHelper.setNextAlu(AluRegisters::OPCODE_LOAD, AluRegisters::R_SRCA, AluRegisters::R_9);
     aluHelper.setNextAlu(AluRegisters::OPCODE_LOAD, AluRegisters::R_SRCB, AluRegisters::R_10);
     aluHelper.setNextAlu(AluRegisters::OPCODE_ADD);
@@ -815,7 +822,10 @@ void DirectSubmissionHw<GfxFamily, Dispatcher>::preinitializeRelaxedOrderingSect
     LriHelper<GfxFamily>::program(&stream, CS_GPR_R8, 8, true);
     LriHelper<GfxFamily>::program(&stream, CS_GPR_R8 + 4, 0, true);
 
+    const uint32_t miMathMocs = this->rootDeviceEnvironment.getGmmHelper()->getMOCS(GMM_RESOURCE_USAGE_OCL_BUFFER);
+
     EncodeAluHelper<GfxFamily, 9> aluHelper;
+    aluHelper.setMocs(miMathMocs);
     aluHelper.setNextAlu(AluRegisters::OPCODE_LOAD, AluRegisters::R_SRCA, AluRegisters::R_1);
     aluHelper.setNextAlu(AluRegisters::OPCODE_LOAD, AluRegisters::R_SRCB, AluRegisters::R_8);
     aluHelper.setNextAlu(AluRegisters::OPCODE_SHL);
diff --git a/shared/source/gen11/hw_cmds_base.h b/shared/source/gen11/hw_cmds_base.h
index 15e3ec87b9..8d04785a0a 100644
--- a/shared/source/gen11/hw_cmds_base.h
+++ b/shared/source/gen11/hw_cmds_base.h
@@ -27,6 +27,7 @@ struct Gen11 {
     static constexpr bool isUsingGenericMediaStateClear = true;
     static constexpr bool isUsingMiMemFence = false;
     static constexpr bool isUsingMiSetPredicate = false;
+    static constexpr bool isUsingMiMathMocs = false;
 
     struct FrontEndStateSupport {
         static constexpr bool scratchSize = true;
diff --git a/shared/source/gen12lp/hw_cmds_base.h b/shared/source/gen12lp/hw_cmds_base.h
index e7059082d8..a34b217a4f 100644
--- a/shared/source/gen12lp/hw_cmds_base.h
+++ b/shared/source/gen12lp/hw_cmds_base.h
@@ -26,6 +26,7 @@ struct Gen12Lp {
     static constexpr uint32_t stateComputeModeForceNonCoherentMask = (0b11u << 3);
     static constexpr bool isUsingMiMemFence = false;
     static constexpr bool isUsingMiSetPredicate = false;
+    static constexpr bool isUsingMiMathMocs = false;
 
     struct FrontEndStateSupport {
         static constexpr bool scratchSize = true;
diff --git a/shared/source/gen8/hw_cmds_base.h b/shared/source/gen8/hw_cmds_base.h
index 2386fff095..fb9ea71d2f 100644
--- a/shared/source/gen8/hw_cmds_base.h
+++ b/shared/source/gen8/hw_cmds_base.h
@@ -28,6 +28,7 @@ struct Gen8 {
     static constexpr bool isUsingGenericMediaStateClear = true;
     static constexpr bool isUsingMiMemFence = false;
     static constexpr bool isUsingMiSetPredicate = false;
+    static constexpr bool isUsingMiMathMocs = false;
 
     struct FrontEndStateSupport {
         static constexpr bool scratchSize = true;
diff --git a/shared/source/gen9/hw_cmds_base.h b/shared/source/gen9/hw_cmds_base.h
index 2522912f94..7927ba5aa6 100644
--- a/shared/source/gen9/hw_cmds_base.h
+++ b/shared/source/gen9/hw_cmds_base.h
@@ -27,6 +27,7 @@ struct Gen9 {
     static constexpr bool isUsingGenericMediaStateClear = true;
     static constexpr bool isUsingMiMemFence = false;
     static constexpr bool isUsingMiSetPredicate = false;
+    static constexpr bool isUsingMiMathMocs = false;
 
     struct FrontEndStateSupport {
         static constexpr bool scratchSize = true;
diff --git a/shared/source/generated/xe_hpc_core/hw_cmds_generated_xe_hpc_core.inl b/shared/source/generated/xe_hpc_core/hw_cmds_generated_xe_hpc_core.inl
index 14996a5fcd..5de2a4b8d2 100644
--- a/shared/source/generated/xe_hpc_core/hw_cmds_generated_xe_hpc_core.inl
+++ b/shared/source/generated/xe_hpc_core/hw_cmds_generated_xe_hpc_core.inl
@@ -652,7 +652,8 @@ typedef struct tagMI_MATH {
     union _DW0 {
         struct _BitField {
             uint32_t DwordLength : BITFIELD_RANGE(0, 7);
-            uint32_t Reserved : BITFIELD_RANGE(8, 22);
+            uint32_t MemoryObjectControlState : BITFIELD_RANGE(8, 14);
+            uint32_t Reserved : BITFIELD_RANGE(15, 22);
             uint32_t InstructionOpcode : BITFIELD_RANGE(23, 28);
             uint32_t InstructionType : BITFIELD_RANGE(29, 31);
         } BitField;
diff --git a/shared/source/generated/xe_hpg_core/hw_cmds_generated_xe_hpg_core.inl b/shared/source/generated/xe_hpg_core/hw_cmds_generated_xe_hpg_core.inl
index 51e60940fb..69a8e3c789 100644
--- a/shared/source/generated/xe_hpg_core/hw_cmds_generated_xe_hpg_core.inl
+++ b/shared/source/generated/xe_hpg_core/hw_cmds_generated_xe_hpg_core.inl
@@ -401,7 +401,8 @@ typedef struct tagMI_MATH {
     union _DW0 {
         struct _BitField {
             uint32_t DwordLength : BITFIELD_RANGE(0, 7);
-            uint32_t Reserved : BITFIELD_RANGE(8, 22);
+            uint32_t MemoryObjectControlState : BITFIELD_RANGE(8, 14);
+            uint32_t Reserved : BITFIELD_RANGE(15, 22);
             uint32_t InstructionOpcode : BITFIELD_RANGE(23, 28);
             uint32_t InstructionType : BITFIELD_RANGE(29, 31);
         } BitField;
diff --git a/shared/source/xe_hp_core/hw_cmds_base.h b/shared/source/xe_hp_core/hw_cmds_base.h
index c570cfd5af..5dbeac1de0 100644
--- a/shared/source/xe_hp_core/hw_cmds_base.h
+++ b/shared/source/xe_hp_core/hw_cmds_base.h
@@ -32,6 +32,7 @@ struct XeHpCore {
     static constexpr bool isUsingGenericMediaStateClear = true;
     static constexpr bool isUsingMiMemFence = false;
     static constexpr bool isUsingMiSetPredicate = true;
+    static constexpr bool isUsingMiMathMocs = false;
 
     struct FrontEndStateSupport {
         static constexpr bool scratchSize = true;
diff --git a/shared/source/xe_hpc_core/hw_cmds_xe_hpc_core_base.h b/shared/source/xe_hpc_core/hw_cmds_xe_hpc_core_base.h
index 46fcc9ccb4..a86c3a08c3 100644
--- a/shared/source/xe_hpc_core/hw_cmds_xe_hpc_core_base.h
+++ b/shared/source/xe_hpc_core/hw_cmds_xe_hpc_core_base.h
@@ -34,6 +34,7 @@ struct XeHpcCore {
     static constexpr bool isUsingGenericMediaStateClear = true;
     static constexpr bool isUsingMiMemFence = true;
     static constexpr bool isUsingMiSetPredicate = true;
+    static constexpr bool isUsingMiMathMocs = true;
 
     struct StateBaseAddressStateSupport {
         static constexpr bool globalAtomics = false;
diff --git a/shared/source/xe_hpg_core/hw_cmds_xe_hpg_core_base.h b/shared/source/xe_hpg_core/hw_cmds_xe_hpg_core_base.h
index 0f9c4b7db5..4e512fc86f 100644
--- a/shared/source/xe_hpg_core/hw_cmds_xe_hpg_core_base.h
+++ b/shared/source/xe_hpg_core/hw_cmds_xe_hpg_core_base.h
@@ -34,6 +34,7 @@ struct XeHpgCore {
     static constexpr bool isUsingGenericMediaStateClear = true;
     static constexpr bool isUsingMiMemFence = false;
     static constexpr bool isUsingMiSetPredicate = true;
+    static constexpr bool isUsingMiMathMocs = true;
 
     struct FrontEndStateSupport {
         static constexpr bool scratchSize = true;
diff --git a/shared/test/unit_test/direct_submission/direct_submission_tests_2.cpp b/shared/test/unit_test/direct_submission/direct_submission_tests_2.cpp
index d93f9871bb..e89257802c 100644
--- a/shared/test/unit_test/direct_submission/direct_submission_tests_2.cpp
+++ b/shared/test/unit_test/direct_submission/direct_submission_tests_2.cpp
@@ -12,6 +12,8 @@
 #include "shared/source/direct_submission/direct_submission_hw.h"
 #include "shared/source/direct_submission/dispatchers/render_dispatcher.h"
 #include "shared/source/direct_submission/relaxed_ordering_helper.h"
+#include "shared/source/gmm_helper/gmm_helper.h"
+#include "shared/source/gmm_helper/gmm_lib.h"
 #include "shared/source/helpers/flush_stamp.h"
 #include "shared/source/helpers/gfx_core_helper.h"
 #include "shared/source/helpers/register_offsets.h"
@@ -1054,7 +1056,7 @@ struct DirectSubmissionRelaxedOrderingTests : public DirectSubmissionDispatchBuf
     bool verifyDynamicSchedulerProgramming(LinearStream &cs, uint64_t schedulerAllocationGpuVa, uint64_t semaphoreGpuVa, uint32_t semaphoreValue, size_t offset, size_t &endOffset);
 
     template <typename FamilyType>
-    bool verifyStaticSchedulerProgramming(GraphicsAllocation &schedulerAllocation, uint64_t deferredTaskListVa, uint32_t expectedQueueSizeLimit);
+    bool verifyStaticSchedulerProgramming(GraphicsAllocation &schedulerAllocation, uint64_t deferredTaskListVa, uint32_t expectedQueueSizeLimit, uint32_t miMathMocs);
 
     template <typename FamilyType>
     bool verifyMiPredicate(void *miPredicateCmd, MiPredicateType predicateType);
@@ -1329,7 +1331,7 @@ bool DirectSubmissionRelaxedOrderingTests::verifyConditionalDataRegBbStart(void
 }
 
 template <typename FamilyType>
-bool DirectSubmissionRelaxedOrderingTests::verifyStaticSchedulerProgramming(GraphicsAllocation &schedulerAllocation, uint64_t deferredTaskListVa, uint32_t expectedQueueSizeLimit) {
+bool DirectSubmissionRelaxedOrderingTests::verifyStaticSchedulerProgramming(GraphicsAllocation &schedulerAllocation, uint64_t deferredTaskListVa, uint32_t expectedQueueSizeLimit, uint32_t miMathMocs) {
     using MI_BATCH_BUFFER_START = typename FamilyType::MI_BATCH_BUFFER_START;
     using MI_LOAD_REGISTER_IMM = typename FamilyType::MI_LOAD_REGISTER_IMM;
     using MI_LOAD_REGISTER_REG = typename FamilyType::MI_LOAD_REGISTER_REG;
@@ -1412,6 +1414,12 @@ bool DirectSubmissionRelaxedOrderingTests::verifyStaticSchedulerProgramming(Grap
         return false;
     }
 
+    if constexpr (FamilyType::isUsingMiMathMocs) {
+        if (miMathCmd->DW0.BitField.MemoryObjectControlState != miMathMocs) {
+            return false;
+        }
+    }
+
     auto miAluCmd = reinterpret_cast<MI_MATH_ALU_INST_INLINE *>(++miMathCmd);
     if (!verifyAlu<FamilyType>(miAluCmd, AluRegisters::OPCODE_LOAD, AluRegisters::R_SRCA, AluRegisters::R_2)) {
         return false;
@@ -1504,6 +1512,12 @@ bool DirectSubmissionRelaxedOrderingTests::verifyStaticSchedulerProgramming(Grap
         return false;
     }
 
+    if constexpr (FamilyType::isUsingMiMathMocs) {
+        if (miMathCmd->DW0.BitField.MemoryObjectControlState != miMathMocs) {
+            return false;
+        }
+    }
+
     miAluCmd = reinterpret_cast<MI_MATH_ALU_INST_INLINE *>(++miMathCmd);
     if (!verifyAlu<FamilyType>(miAluCmd, AluRegisters::OPCODE_LOAD, AluRegisters::R_SRCA, AluRegisters::R_1)) {
         return false;
@@ -1646,6 +1660,12 @@ bool DirectSubmissionRelaxedOrderingTests::verifyStaticSchedulerProgramming(Grap
         return false;
     }
 
+    if constexpr (FamilyType::isUsingMiMathMocs) {
+        if (miMathCmd->DW0.BitField.MemoryObjectControlState != miMathMocs) {
+            return false;
+        }
+    }
+
     miAluCmd = reinterpret_cast<MI_MATH_ALU_INST_INLINE *>(++miMathCmd);
     if (!verifyAlu<FamilyType>(miAluCmd, AluRegisters::OPCODE_LOAD, AluRegisters::R_SRCA, AluRegisters::R_9)) {
         return false;
@@ -1793,7 +1813,8 @@ HWTEST2_F(DirectSubmissionRelaxedOrderingTests, givenDebugFlagSetWhenDispatching
 
     EXPECT_EQ(1u, directSubmission.dispatchStaticRelaxedOrderingSchedulerCalled);
     EXPECT_TRUE(verifyStaticSchedulerProgramming<FamilyType>(*directSubmission.relaxedOrderingSchedulerAllocation,
-                                                             directSubmission.deferredTasksListAllocation->getGpuAddress(), 123));
+                                                             directSubmission.deferredTasksListAllocation->getGpuAddress(), 123,
+                                                             pDevice->getRootDeviceEnvironment().getGmmHelper()->getMOCS(GMM_RESOURCE_USAGE_OCL_BUFFER)));
 }
 
 HWTEST2_F(DirectSubmissionRelaxedOrderingTests, givenNewNumberOfClientsWhenDispatchingWorkThenIncraseQueueSize, IsAtLeastXeHpcCore) {
@@ -1806,7 +1827,8 @@ HWTEST2_F(DirectSubmissionRelaxedOrderingTests, givenNewNumberOfClientsWhenDispa
     EXPECT_EQ(1u, directSubmission.dispatchStaticRelaxedOrderingSchedulerCalled);
     EXPECT_EQ(RelaxedOrderingHelper::queueSizeMultiplier, directSubmission.currentRelaxedOrderingQueueSize);
     EXPECT_TRUE(verifyStaticSchedulerProgramming<FamilyType>(*directSubmission.relaxedOrderingSchedulerAllocation,
-                                                             directSubmission.deferredTasksListAllocation->getGpuAddress(), RelaxedOrderingHelper::queueSizeMultiplier));
+                                                             directSubmission.deferredTasksListAllocation->getGpuAddress(), RelaxedOrderingHelper::queueSizeMultiplier,
+                                                             pDevice->getRootDeviceEnvironment().getGmmHelper()->getMOCS(GMM_RESOURCE_USAGE_OCL_BUFFER)));
 
     const uint64_t expectedQueueSizeValueVa = directSubmission.relaxedOrderingSchedulerAllocation->getGpuAddress() +
                                               RelaxedOrderingHelper::StaticSchedulerSizeAndOffsetSection<FamilyType>::drainRequestSectionStart +
@@ -1876,7 +1898,8 @@ HWTEST2_F(DirectSubmissionRelaxedOrderingTests, whenInitializingThenDispatchStat
 
         EXPECT_EQ(1u, directSubmission.dispatchStaticRelaxedOrderingSchedulerCalled);
         EXPECT_TRUE(verifyStaticSchedulerProgramming<FamilyType>(*directSubmission.relaxedOrderingSchedulerAllocation,
-                                                                 directSubmission.deferredTasksListAllocation->getGpuAddress(), RelaxedOrderingHelper::queueSizeMultiplier));
+                                                                 directSubmission.deferredTasksListAllocation->getGpuAddress(), RelaxedOrderingHelper::queueSizeMultiplier,
+                                                                 pDevice->getRootDeviceEnvironment().getGmmHelper()->getMOCS(GMM_RESOURCE_USAGE_OCL_BUFFER)));
     }
 
     {
@@ -2031,6 +2054,10 @@ HWTEST_F(DirectSubmissionRelaxedOrderingTests, whenDispatchingWorkThenDispatchTa
     auto miMathCmd = reinterpret_cast<MI_MATH *>(++lriCmd);
     EXPECT_EQ(8u, miMathCmd->DW0.BitField.DwordLength);
 
+    if constexpr (FamilyType::isUsingMiMathMocs) {
+        EXPECT_EQ(miMathCmd->DW0.BitField.MemoryObjectControlState, pDevice->getRootDeviceEnvironment().getGmmHelper()->getMOCS(GMM_RESOURCE_USAGE_OCL_BUFFER));
+    }
+
     auto miAluCmd = reinterpret_cast<MI_MATH_ALU_INST_INLINE *>(++miMathCmd);
     EXPECT_TRUE(verifyAlu<FamilyType>(miAluCmd, AluRegisters::OPCODE_LOAD, AluRegisters::R_SRCA, AluRegisters::R_1));