diff --git a/opencl/source/command_queue/command_queue.cpp b/opencl/source/command_queue/command_queue.cpp
index 91eea51729..161c2bf446 100644
--- a/opencl/source/command_queue/command_queue.cpp
+++ b/opencl/source/command_queue/command_queue.cpp
@@ -1517,13 +1517,21 @@ cl_int CommandQueue::enqueueStagingBufferMemcpy(cl_bool blockingCopy, void *dstP
         profilingEvent.setQueueTimeStamp();
     }
 
+    // If there was only one chunk copy, no barrier for OOQ is needed
+    bool isSingleTransfer = false;
     auto chunkCopy = [&](void *chunkDst, void *stagingBuffer, const void *chunkSrc, size_t chunkSize) -> int32_t {
         auto isFirstTransfer = (chunkDst == dstPtr);
         auto isLastTransfer = ptrOffset(chunkDst, chunkSize) == ptrOffset(dstPtr, size);
+        isSingleTransfer = isFirstTransfer && isLastTransfer;
+
         if (isFirstTransfer && isProfilingEnabled()) {
             profilingEvent.setSubmitTimeStamp();
         }
         memcpy(stagingBuffer, chunkSrc, chunkSize);
+        if (isSingleTransfer) {
+            return this->enqueueSVMMemcpy(false, chunkDst, stagingBuffer, chunkSize, 0, nullptr, event);
+        }
+
         if (isFirstTransfer && isProfilingEnabled()) {
             profilingEvent.setStartTimeStamp();
         }
@@ -1543,12 +1551,12 @@ cl_int CommandQueue::enqueueStagingBufferMemcpy(cl_bool blockingCopy, void *dstP
     }
 
     if (event != nullptr) {
-        if (this->isOOQEnabled()) {
+        if (!isSingleTransfer && this->isOOQEnabled()) {
             ret = this->enqueueBarrierWithWaitList(0, nullptr, event);
         }
         if (isProfilingEnabled()) {
             auto pEvent = castToObjectOrAbort<Event>(*event);
-            pEvent->copyTimestamps(profilingEvent);
+            pEvent->copyTimestamps(profilingEvent, !isSingleTransfer);
             pEvent->setCPUProfilingPath(false);
         }
     }
diff --git a/opencl/source/event/event.cpp b/opencl/source/event/event.cpp
index 236a3150a0..bb32a9c62b 100644
--- a/opencl/source/event/event.cpp
+++ b/opencl/source/event/event.cpp
@@ -397,7 +397,7 @@ void Event::calculateProfilingDataInternal(uint64_t contextStartTS, uint64_t con
     auto &device = this->cmdQueue->getDevice();
     auto &gfxCoreHelper = device.getGfxCoreHelper();
     auto resolution = device.getDeviceInfo().profilingTimerResolution;
-    if (timestampsCopied) {
+    if (isAdjustmentNeeded) {
         // Adjust startTS since we calculate profiling based on other event timestamps
         contextStartTS = startTimeStamp.gpuTimeStamp;
     }
diff --git a/opencl/source/event/event.h b/opencl/source/event/event.h
index 2731f7a727..53f9927252 100644
--- a/opencl/source/event/event.h
+++ b/opencl/source/event/event.h
@@ -312,12 +312,12 @@ class Event : public BaseObject<_cl_event>, public IDNode<Event> {
 
     static void getBoundaryTimestampValues(TimestampPacketContainer *timestampContainer, uint64_t &globalStartTS, uint64_t &globalEndTS);
 
-    void copyTimestamps(const Event &srcEvent) {
+    void copyTimestamps(const Event &srcEvent, bool isAdjustmentNeeded) {
         this->queueTimeStamp = srcEvent.queueTimeStamp;
         this->submitTimeStamp = srcEvent.submitTimeStamp;
         this->startTimeStamp = srcEvent.startTimeStamp;
         this->endTimeStamp = srcEvent.endTimeStamp;
-        timestampsCopied = true;
+        this->isAdjustmentNeeded = isAdjustmentNeeded;
     }
 
   protected:
@@ -391,7 +391,7 @@ class Event : public BaseObject<_cl_event>, public IDNode<Event> {
     bool profilingEnabled = false;
     bool profilingCpuPath = false;
     bool dataCalculated = false;
-    bool timestampsCopied = false;
+    bool isAdjustmentNeeded = false;
 
     ProfilingInfo queueTimeStamp{};
     ProfilingInfo submitTimeStamp{};
diff --git a/opencl/test/unit_test/command_queue/enqueue_svm_tests.cpp b/opencl/test/unit_test/command_queue/enqueue_svm_tests.cpp
index a4c554e6b8..224ec46989 100644
--- a/opencl/test/unit_test/command_queue/enqueue_svm_tests.cpp
+++ b/opencl/test/unit_test/command_queue/enqueue_svm_tests.cpp
@@ -2456,6 +2456,34 @@ HWTEST_F(StagingBufferTest, givenOutOfOrderCmdQueueWhenEnqueueStagingBufferMemcp
     clReleaseEvent(event);
 }
 
+HWTEST_F(StagingBufferTest, givenOutOfOrderCmdQueueWhenEnqueueStagingBufferMemcpyNonBlockingWithSingleTransferThenNoBarrierEnqueued) {
+    constexpr cl_command_type expectedLastCmd = CL_COMMAND_SVM_MEMCPY;
+
+    cl_event event;
+    MockCommandQueueHw<FamilyType> myCmdQ(context, pClDevice, 0);
+    myCmdQ.setOoqEnabled();
+
+    auto initialUsmAllocs = svmManager->getNumAllocs();
+    retVal = myCmdQ.enqueueStagingBufferMemcpy(
+        false,             // cl_bool blocking_copy
+        dstPtr,            // void *dst_ptr
+        srcPtr,            // const void *src_ptr
+        stagingBufferSize, // size_t size
+        &event             // cl_event *event
+    );
+
+    auto pEvent = (Event *)event;
+    auto numOfStagingBuffers = svmManager->getNumAllocs() - initialUsmAllocs;
+    EXPECT_EQ(CL_SUCCESS, retVal);
+    EXPECT_EQ(1u, numOfStagingBuffers);
+    EXPECT_EQ(1u, myCmdQ.enqueueSVMMemcpyCalledCount);
+    EXPECT_EQ(0u, myCmdQ.finishCalledCount);
+    EXPECT_EQ(expectedLastCmd, myCmdQ.lastCommandType);
+    EXPECT_EQ(expectedLastCmd, pEvent->getCommandType());
+
+    clReleaseEvent(event);
+}
+
 HWTEST_F(StagingBufferTest, givenEnqueueStagingBufferMemcpyWhenTaskCountNotReadyThenCopySucessfullAndBuffersNotReused) {
     MockCommandQueueHw<FamilyType> myCmdQ(context, pClDevice, 0);
     auto initialUsmAllocs = svmManager->getNumAllocs();