diff --git a/src/plugins/intel_gpu/include/intel_gpu/runtime/execution_config.hpp b/src/plugins/intel_gpu/include/intel_gpu/runtime/execution_config.hpp
index 3e854e4c9c5ada..5e059b17da0e97 100644
--- a/src/plugins/intel_gpu/include/intel_gpu/runtime/execution_config.hpp
+++ b/src/plugins/intel_gpu/include/intel_gpu/runtime/execution_config.hpp
@@ -140,7 +140,7 @@ class ExecutionConfig {
 
     // Note that RT info property value has lower priority than values set by user via core.set_property or passed to compile_model call
     // So this method should be called after setting all user properties, but before apply_user_properties() call.
-    void apply_rt_info(const ov::RTMap& rt_info);
+    void apply_rt_info(const cldnn::device_info& info, const ov::RTMap& rt_info);
 
     std::string to_string() const;
 
diff --git a/src/plugins/intel_gpu/src/plugin/plugin.cpp b/src/plugins/intel_gpu/src/plugin/plugin.cpp
index c8839472a6d962..f139279bf5a9a8 100644
--- a/src/plugins/intel_gpu/src/plugin/plugin.cpp
+++ b/src/plugins/intel_gpu/src/plugin/plugin.cpp
@@ -190,7 +190,7 @@ std::shared_ptr<ov::ICompiledModel> Plugin::compile_model(const std::shared_ptr<
     ExecutionConfig config = m_configs_map.at(device_id);
     config.set_user_property(orig_config);
     if (model->has_rt_info("runtime_options"))
-        config.apply_rt_info(model->get_rt_info<ov::AnyMap>("runtime_options"));
+        config.apply_rt_info(context->get_engine().get_device_info(), model->get_rt_info<ov::AnyMap>("runtime_options"));
     config.apply_user_properties(context->get_engine().get_device_info());
 
     set_cache_info(model, config);
@@ -281,7 +281,7 @@ ov::SupportedOpsMap Plugin::query_model(const std::shared_ptr<const ov::Model>&
     ExecutionConfig config = m_configs_map.at(device_id);
     config.set_user_property(orig_config);
     if (model->has_rt_info("runtime_options"))
-        config.apply_rt_info(model->get_rt_info<ov::AnyMap>("runtime_options"));
+        config.apply_rt_info(ctx->get_engine().get_device_info(), model->get_rt_info<ov::AnyMap>("runtime_options"));
     config.apply_user_properties(ctx->get_engine().get_device_info());
 
     ProgramBuilder prog(ctx->get_engine(), config);
diff --git a/src/plugins/intel_gpu/src/runtime/execution_config.cpp b/src/plugins/intel_gpu/src/runtime/execution_config.cpp
index 804ad81f2d3735..a698ec7eb6c5a0 100644
--- a/src/plugins/intel_gpu/src/runtime/execution_config.cpp
+++ b/src/plugins/intel_gpu/src/runtime/execution_config.cpp
@@ -262,10 +262,12 @@ void ExecutionConfig::apply_user_properties(const cldnn::device_info& info) {
     user_properties.clear();
 }
 
-void ExecutionConfig::apply_rt_info(const ov::RTMap& rt_info) {
-    apply_rt_info_property(ov::hint::kv_cache_precision, rt_info);
+void ExecutionConfig::apply_rt_info(const cldnn::device_info& info, const ov::RTMap& rt_info) {
+    if (!info.supports_immad) {
+        apply_rt_info_property(ov::hint::kv_cache_precision, rt_info);
+        apply_rt_info_property(ov::hint::activations_scale_factor, rt_info);
+    }
     apply_rt_info_property(ov::hint::dynamic_quantization_group_size, rt_info);
-    apply_rt_info_property(ov::hint::activations_scale_factor, rt_info);
 }
 
 std::string ExecutionConfig::to_string() const {
diff --git a/src/plugins/intel_gpu/tests/functional/behavior/properties.cpp b/src/plugins/intel_gpu/tests/functional/behavior/properties.cpp
index 93a00262db35c2..4af05f30c6b4b0 100644
--- a/src/plugins/intel_gpu/tests/functional/behavior/properties.cpp
+++ b/src/plugins/intel_gpu/tests/functional/behavior/properties.cpp
@@ -2,7 +2,9 @@
 // SPDX-License-Identifier: Apache-2.0
 //
 
+#include <algorithm>
 #include "openvino/runtime/properties.hpp"
+#include "openvino/runtime/intel_gpu/properties.hpp"
 #include "base/ov_behavior_test_utils.hpp"
 #include "openvino/runtime/core.hpp"
 #include "common_test_utils/subgraph_builders/conv_pool_relu.hpp"
@@ -43,11 +45,17 @@ TEST_F(TestPropertiesGPU, RTInfoPropertiesWithDefault) {
     model->set_rt_info("8.0", "runtime_options", ov::hint::activations_scale_factor.name());
 
     OV_ASSERT_NO_THROW(compiled_model = core.compile_model(model, ov::test::utils::DEVICE_GPU));
-    OV_ASSERT_NO_THROW(type = compiled_model.get_property(ov::hint::kv_cache_precision));
     OV_ASSERT_NO_THROW(size = compiled_model.get_property(ov::hint::dynamic_quantization_group_size));
+    ASSERT_EQ(size.as<uint64_t>(), 0);
+
+    // GPU with systolic does not support some of rt_info
+    auto capabilities = core.get_property(ov::test::utils::DEVICE_GPU, ov::device::capabilities);
+    if (find(capabilities.cbegin(), capabilities.cend(), ov::intel_gpu::capability::HW_MATMUL) != capabilities.cend())
+        return;
+
+    OV_ASSERT_NO_THROW(type = compiled_model.get_property(ov::hint::kv_cache_precision));
     OV_ASSERT_NO_THROW(scale = compiled_model.get_property(ov::hint::activations_scale_factor));
     ASSERT_EQ(type.as<ov::element::Type>(), ov::element::f16);
-    ASSERT_EQ(size.as<uint64_t>(), 0);
     ASSERT_EQ(scale.as<float>(), 8.0f);
 }