Bug in latest Intel Graphics Driver on Windows 11 home

PatGelsinger — Wed, 03 Dec 2025 14:18:41 GMT

I do have installed all latest drivers & packages on Windows 11.

Setup "normal" = Leads to a bug, XPU initialisation fails!

Intel® Graphics Driver 32.0.101.8250
date 11/12/2025

Checking what this installs on windows .. Win + X => Device Manager => Display Adaptors:

Intel Arc A730M Graphics
Driver date: 07.11.2025
Driver version: 32.0.101.8250

Intel Iris Xe Graphics
Driver date: 29.04.2022
Driver version: 30.0.101.1960

I want to use my A730M to run AI models locally in PyTorch, thus installed:

Intel oneAPI Base Toolkit 2025.3.0
date October 21, 2025

IPEX 2.8.10
python -m pip install torch==2.8.0 torchvision==0.23.0 torchaudio==2.8.0 --index-url https://download.pytorch.org/whl/xpu
python -m pip install intel-extension-for-pytorch==2.8.10+xpu --index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/

Performing the sanity check
python -c "import torch; import intel_extension_for_pytorch as ipex; print(torch.__version__); print(ipex.__version__); [print(f'[{i}]: {torch.xpu.get_device_properties(i)}') for i in range(torch.xpu.device_count())];"

[W1201 10:19:58.000000000 OperatorEntry.cpp:225] Warning: Warning only once for all operators, other operators may also be overridden.
Overriding a previously registered kernel for the same operator and the same dispatch key
operator: aten::geometric_(Tensor(a!) self, float p, *, Generator? generator=None) -> Tensor(a!)
registered at C:\actions-runner\_work\pytorch\pytorch\pytorch\build\aten\src\ATen\RegisterSchema.cpp:6
dispatch key: XPU
previous kernel: registered at C:\actions-runner\_work\pytorch\pytorch\pytorch\aten\src\ATen\VmapModeRegistrations.cpp:37
new kernel: registered at G:\frameworks.ai.pytorch.ipex-gpu\build\Release\csrc\gpu\csrc\gpu\xpu\ATen\RegisterXPU_0.cpp:172 (function operator ())
2.8.0+xpu
2.8.10+xpu
C:\Program Files\Python310\lib\site-packages\torch\xpu\__init__.py:60: UserWarning: Failed to initialize XPU devices. The driver may not be installed, installed incorrectly, or incompatible with the current setup. Please refer to the guideline (https://github.com/pytorch/pytorch?tab=readme-ov-file#intel-gpu-support) for proper installation and configuration. (Triggered internally at C:\actions-runner\_work\pytorch\pytorch\pytorch\c10\xpu\XPUFunctions.cpp:111.)
return torch._C._xpu_getDeviceCount()

Setup "not normal" = XPU initialisation succeeds!

"disable device" for Intel Iris Xe Graphics: Works!!
This loads an old default windows backup driver
MS Basic Display Adapter
Driver date: 21.06.2006
Driver version: 10.0.26100.1

Performing sanity
python -c "import torch; import intel_extension_for_pytorch as ipex; print(torch.__version__); print(ipex.__version__); [print(f'[{i}]: {torch.xpu.get_device_properties(i)}') for i in range(torch.xpu.device_count())];"
Let's find the XPU and can be used in IPEX
[0]: _XpuDeviceProperties(name='Intel(R) Arc(TM) A730M Graphics', platform_name='Intel(R) oneAPI Unified Runtime over Level-Zero', type='gpu', driver_version='1.13.35227', total_memory=11932MB, max_compute_units=384, gpu_eu_count=384, gpu_subslice_count=48, max_work_group_size=1024, max_num_sub_groups=128, sub_group_sizes=[8 16 32], has_fp16=1, has_fp64=0, has_atomic64=1)

Problem with that setup: The GUI and the whole Windows OS perform really slow (like scrolling over a webpage takes way longer, each mouse move or window/tab change takes more time etc.)

Question: Why is the common setup with Intel Iris Xe Graphics driver for the non-GPU part failing, whereas an oudated old windows basic driver lets the XPU intialize successfully?

topic Bug in latest Intel Graphics Driver on Windows 11 home in Intel® oneAPI Math Kernel Library

Bug in latest Intel Graphics Driver on Windows 11 home