| Summary: | No Hardware Acceleration with Nvidia RTX Pro 1000 Blackwell GPU? | ||
|---|---|---|---|
| Product: | [Applications] digikam | Reporter: | CM <carlo.mj.m> |
| Component: | Faces-Engine | Assignee: | Digikam Developers <digikam-bugs-null> |
| Status: | REPORTED --- | ||
| Severity: | normal | CC: | caulier.gilles, hummelthomas, kimdyre |
| Priority: | NOR | ||
| Version First Reported In: | 8.8.0 | ||
| Target Milestone: | --- | ||
| Platform: | Microsoft Windows | ||
| OS: | Microsoft Windows | ||
| Latest Commit: | Version Fixed/Implemented In: | ||
| Sentry Crash Report: | |||
| Attachments: | Output from clinfo.exe in my Windows 11 context | ||
|
Description
CM
2025-11-01 21:57:19 UTC
During Christmas time, i will rebuild the whole digiKam installer from scratch with VCPKG packages updated (opencv, qt, ffmpeg). We will see if the problem remain. Gilles Caulier (In reply to caulier.gilles from comment #1) > During Christmas time, i will rebuild the whole digiKam installer from > scratch > with VCPKG packages updated (opencv, qt, ffmpeg). We will see if the problem > remain. > > Gilles Caulier I wonder if this is why: https://www.tomshardware.com/pc-components/gpus/nvidias-poor-rtx-50-compute-test-results-due-to-missing-32-bit-opencl-support-says-passmark Created attachment 188188 [details]
Output from clinfo.exe in my Windows 11 context
Just to ease debugging, I want to provide some more info: I am experiencing same issue with digiKam 8.8.0 on Win11 with Intel Core i7 and Intel Iris Xe GPU (OS and drivers all updated to current stable versions). Running clinfo from terminal confirms that OpenCL is installed, but I still get "OpenCL availability: No" in the help menu and "Error: OpenCL is not available on this system" when testing in digiKam's "Settings - Miscellaneous". I have also tried adding the environment variable OPENCV_OPENCL_DEVICE with various settings: INTEL:GPU:0 :GPU:0 INTEL:GPU:1 All at no avail. In the "Shared libraries and components information" in digiKam I have following, that may be relevant: digikam version 8.8.0 OpenCV Configuration: Threads features: Parallel framework: openmp Number of Threads: 12 OpenCL availability: No .. .. LibOpenCV: 4.11.0 .. Build target architecture: AMD64 Build target: RelWithDebInfo Build host architecture: AMD64 Build date: 17/10/2025 13.48 I attach the output from running clinfo.exe. This is straightforward to understand. Since OpenCV acts as an entry-level API for processing computations with OpenCL (and indirectly with the GPU), if the 'OpenCL availability: No' status is reported in the Help/Component Info dialog, the OpenCL test on the Setup/Miscs/System page is expected to fail as a result. This behavior is typical if the NVIDIA drivers (and thus OpenCL support) are not correctly installed or configured on Windows systems. For more details, see the OpenCV documentation on OpenCL : https://docs.opencv.org/4.x/d7/d1b/group__gpu.html Gilles Caulier (In reply to caulier.gilles from comment #5) >... the OpenCL test on the Setup/Miscs/System page is expected to > fail as a result. ...and I wasn't wondering - just wanted to provide as much and complete info as possible > This behavior is typical if the NVIDIA drivers (and thus OpenCL support) are > not correctly installed or configured on Windows systems. I am not using a nVidia GPU as the initial author does. Again, just wanted to supplement with similar experience from another GPU > For more details, see the OpenCV documentation on OpenCL : > https://docs.opencv.org/4.x/d7/d1b/group__gpu.html Getting a "404 Not Found" when trying the provided link Kim Dyre HEUREKA Using Python from the c:\Program Files\digiKam\ folder I verified OpenCV and access to OpenCL: >>> import cv2; >>> print(cv2.__version__) 4.12.0 >>> cv2.ocl.setUseOpenCL(True) >>> print(cv2.ocl.haveOpenCL()) True With that info I looked in another direction: Added environment variable OPENCV_OPENCL_VERBOSE with value 1 Then changed environment variable OPENCV_OPENCL_DEVICE from INTEL:GPU:0 to :GPU:0 And now it is working: digikam version 8.8.0 OpenCV Configuration: Threads features: Parallel framework: openmp Number of Threads: 12 OpenCL platform: Intel(R) OpenCL Graphics: iGPU: Intel(R) Iris(R) Xe Graphics (OpenCL 3.0 NEO ) OpenCL Device extensions: cl_khr_throttle_hints: Yes ... The GPU test for AI compatibility also passes: "All tests completed successfully! Test completed successfully! Your system can use GPU acceleration for the AI models." In my situation, cuda installation was broken as previous posters mentioned. For me the Microsoft Redistributable was missing, but probably, you could check on all of the four points below. --- Cross-posting my answer from https://www.reddit.com/r/digiKam/comments/1prvsl0/comment/nylpuaz/ I had a similar issue with my NVIDIA GeForce RTX 3060 Ti and now it finally works. I am not fully sure, what the solution in the end was, but I feel that installing the latest Microsoft Redistributable C++ solved it for me. Probably, you could also install Microsoft Visual Studio as explained in cuda system requirements: https://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/index.html#system-requirements So, if I remember correctly: 1. Install cuda toolkit https://developer.nvidia.com/cuda/toolkit (I have 13.1, the latest) 2. Install cuda DNN https://developer.nvidia.com/cudnn 3. Install Microsoft Redistributable C++ https://learn.microsoft.com/en-us/cpp/windows/latest-supported-vc-redist?view=msvc-170#latest-supported-redistributable-version 4. Set OPENCV_OPENCL_DEVICE to NVIDIA:GPU:0 as system environment variable (not sure whether this is needed, though). |