diff --git a/.github/workflows/precommit.yml b/.github/workflows/precommit.yml
index 58f7d888d07..f4a2e478488 100644
--- a/.github/workflows/precommit.yml
+++ b/.github/workflows/precommit.yml
@@ -90,46 +90,44 @@ jobs:
   #         name: coverage_openvino
   #         flags: OPENVINO
 
-  pytorch:
+
+  pytorch-cpu:
+    defaults:
+      run:
+        shell: bash
+    runs-on: ubuntu-20.04
+      - uses: actions/checkout@v3
+        with:
+          lfs: true
+      - uses: actions/setup-python@v3
+        with:
+          python-version: 3.8.10
+      - name: Install NNCF and test requirements
+        run: make install-torch-test
+      - name: Run PyTorch precommit test scope
+        run: make test-torch-cpu
+
+  pytorch-cuda:
     defaults:
       run:
         shell: bash
     runs-on: aks-linux-4-cores-28gb-gpu-testla-t4
 
     steps:
-      - name: uname -a
-        continue-on-error: true
-        run: |
-            uname -a
-            ls -l /usr/local/
-            ls -l /dev/*
-
-      - name: nvidia-smi
-        continue-on-error: true
-        run: |
-            nvidia-smi
-      - name: nvidia-smi
-        continue-on-error: true
-        run: |
-            nvcc --version
-
       - name: Install dependencies
         run : |
           sudo apt-get update
           sudo apt-get --assume-yes install gcc ninja-build libgl1-mesa-dev libglib2.0-0
-
       - name: Install CUDA
         run: |
           wget -q https://developer.download.nvidia.com/compute/cuda/12.1.1/local_installers/cuda_12.1.1_530.30.02_linux.run
           sudo sh cuda_12.1.1_530.30.02_linux.run --toolkit --silent
-
       - name: Runner info
         continue-on-error: true
         run: |
           nvidia-smi
           cat /proc/cpuinfo
           nvcc --version
-
       - uses: actions/checkout@v3
         with:
           lfs: true
@@ -141,5 +139,5 @@ jobs:
       - name: Check CUDA
         run: |
           python -c "import torch; print(torch.cuda.is_available())"
-      - name: Run Pytorch precommit test scope
-        run: pytest tests/torch -m "not weekly and not nightly and not models_hub" -rA
+      - name: Run PyTorch precommit test scope
+        run: make test-torch-cuda
diff --git a/Makefile b/Makefile
index 545d35ec4f2..c72da3ebeb1 100644
--- a/Makefile
+++ b/Makefile
@@ -145,6 +145,12 @@ install-models-hub-torch:
 test-torch:
 	pytest ${COVERAGE_ARGS} tests/torch -m "not weekly and not nightly and not models_hub" --junitxml ${JUNITXML_PATH} $(DATA_ARG)
 
+test-torch-cpu:
+	pytest ${COVERAGE_ARGS} tests/torch -rS -m "not cuda and not weekly and not nightly and not models_hub" --junitxml ${JUNITXML_PATH} $(DATA_ARG)
+
+test-torch-cuda:
+	pytest ${COVERAGE_ARGS} tests/torch -rS -m "cuda and not weekly and not nightly and not models_hub" --junitxml ${JUNITXML_PATH} $(DATA_ARG)
+
 test-torch-nightly:
 	pytest ${COVERAGE_ARGS} tests/torch -m nightly --junitxml ${JUNITXML_PATH} $(DATA_ARG)
 
diff --git a/tests/torch/conftest.py b/tests/torch/conftest.py
index 6546c1085f0..c4a8910e826 100644
--- a/tests/torch/conftest.py
+++ b/tests/torch/conftest.py
@@ -213,7 +213,7 @@ def is_weights(request: FixtureRequest):
     return request.param
 
 
-@pytest.fixture(params=[True, False], ids=["cuda", "cpu"])
+@pytest.fixture(params=[pytest.param(True, marks=pytest.mark.cuda), False], ids=["cuda", "cpu"])
 def use_cuda(request: FixtureRequest):
     return request.param
 
diff --git a/tests/torch/pytest.ini b/tests/torch/pytest.ini
index 26ef8aedef9..0741cf4b300 100644
--- a/tests/torch/pytest.ini
+++ b/tests/torch/pytest.ini
@@ -8,5 +8,6 @@ markers =
     nightly
     weekly
     models_hub
+    cuda
 python_files = test_*
 xfail_strict = true
diff --git a/tests/torch/test_distributed_data_parallel_mode.py b/tests/torch/test_distributed_data_parallel_mode.py
index c186ea97318..c9e12cc5297 100644
--- a/tests/torch/test_distributed_data_parallel_mode.py
+++ b/tests/torch/test_distributed_data_parallel_mode.py
@@ -82,6 +82,7 @@ def worker(rank: int, world_size: int) -> None:
     _ = torch.nn.parallel.DistributedDataParallel(compressed_model, device_ids=[rank])
 
 
+@pytest.mark.cuda
 @pytest.mark.parametrize("waiting_time", [20.0])
 def test_is_ddp_freezing(waiting_time: float) -> None:
     # Number of processes the same as GPU count
diff --git a/tests/torch/test_tensor.py b/tests/torch/test_tensor.py
index ee4d1d19209..1755b3badd8 100644
--- a/tests/torch/test_tensor.py
+++ b/tests/torch/test_tensor.py
@@ -35,6 +35,7 @@ def cast_to(x: torch.Tensor, dtype: TensorDataType) -> torch.Tensor:
         return cast_to(x, dtype)
 
 
+@pytest.mark.cuda
 @pytest.mark.skipif(not torch.cuda.is_available(), reason="Skipping for CPU-only setups")
 class TestCudaPTNNCFTensorOperators(TemplateTestNNCFTensorOperators):
     @staticmethod