diff --git a/shared/source/helpers/gfx_core_helper_base.inl b/shared/source/helpers/gfx_core_helper_base.inl
index bef2724616..49525c1ef9 100644
--- a/shared/source/helpers/gfx_core_helper_base.inl
+++ b/shared/source/helpers/gfx_core_helper_base.inl
@@ -802,6 +802,8 @@ uint32_t GfxCoreHelperHw<GfxFamily>::calculateAvailableThreadCount(const Hardwar
 
 template <typename GfxFamily>
 void GfxCoreHelperHw<GfxFamily>::alignThreadGroupCountToDssSize(uint32_t &threadCount, uint32_t dssCount, uint32_t threadsPerDss, uint32_t threadGroupSize) const {
+    uint32_t availableTreadCount = (threadsPerDss / threadGroupSize) * dssCount;
+    threadCount = std::min(threadCount, availableTreadCount);
 }
 
 template <typename GfxFamily>
diff --git a/shared/test/unit_test/helpers/gfx_core_helper_tests.cpp b/shared/test/unit_test/helpers/gfx_core_helper_tests.cpp
index add34c2b59..777df449ec 100644
--- a/shared/test/unit_test/helpers/gfx_core_helper_tests.cpp
+++ b/shared/test/unit_test/helpers/gfx_core_helper_tests.cpp
@@ -1874,10 +1874,32 @@ HWTEST_F(GfxCoreHelperTest, givenGetDeviceTimestampWidthCalledThenReturnCorrectV
     EXPECT_EQ(64u, helper.getDeviceTimestampWidth());
 }
 
-HWTEST2_F(GfxCoreHelperTest, givenHwHelperWhenAligningThreadGroupCountToDssSizeThenThreadGroupCountDoesNotChange, IsAtMostXe2HpgCore) {
+HWTEST_F(GfxCoreHelperTest, givenHwHelperWhenAligningThreadGroupCountToDssSizeThenThreadGroupCountChanged) {
     auto &helper = getHelper<GfxCoreHelper>();
     uint32_t threadGroupCountBefore = 4096;
     uint32_t threadCount = threadGroupCountBefore;
     helper.alignThreadGroupCountToDssSize(threadCount, 1, 1, 1);
-    EXPECT_EQ(threadGroupCountBefore, threadCount);
+    EXPECT_NE(threadGroupCountBefore, threadCount);
+}
+
+HWTEST_F(GfxCoreHelperTest, givenHwHelperWhenThreadGroupCountIsAlignedToDssThenThreadCountNotChanged) {
+    auto &helper = getHelper<GfxCoreHelper>();
+    uint32_t dssCount = 16;
+    uint32_t threadGroupSize = 32;
+    uint32_t threadsPerDss = 2 * threadGroupSize;
+    uint32_t maxThreadCount = (dssCount * threadsPerDss) / threadGroupSize;
+    uint32_t threadCount = maxThreadCount;
+    helper.alignThreadGroupCountToDssSize(threadCount, dssCount, threadsPerDss, threadGroupSize);
+    EXPECT_EQ(2 * dssCount, threadCount);
+}
+
+HWTEST_F(GfxCoreHelperTest, givenHwHelperWhenThreadGroupCountIsAlignedToDssThenThreadCountChanged) {
+    auto &helper = getHelper<GfxCoreHelper>();
+    uint32_t dssCount = 16;
+    uint32_t threadGroupSize = 32;
+    uint32_t threadsPerDss = 2 * threadGroupSize - 1;
+    uint32_t maxThreadCount = (dssCount * threadsPerDss) / threadGroupSize;
+    uint32_t threadCount = maxThreadCount;
+    helper.alignThreadGroupCountToDssSize(threadCount, dssCount, threadsPerDss, threadGroupSize);
+    EXPECT_EQ(dssCount, threadCount);
 }
\ No newline at end of file