Skip to main content

⚠️ Please note that this topic or post has been archived. The information contained here may no longer be accurate or up-to-date. ⚠️

13.1 Final - Hardware acceleration not working ?

Knorp

Saw this in the betas, but apparently it is still in the final 13.1 release.

0

Comments

7 comments

Date Votes
  • Luka
    • Edited

    Hi Bart de Vries,

    this was already the same in the released version 13.04.

    0
  • Olaf Pokorny

    Same problem here, but 13.0.4 was ok. May be a bug in OpenCL compilation during initialisation. Excerpt from „CaptureOneICOCL.log“: Compilation of CloneBlend3 failed: CL_BUILD_PROGRAM_FAILURE.

    I filed a bug report to PhaseOne.

    0
  • I P

    Importing files to catalog, drag files in an album,  adding keywords ... is so slow that working with C13.1 makes no sense for me (MBPro 15“ 2017, MacOS 10.15.4, 16GB,1TBHDRadeon Pro 5404GB). %CPU shows 350%; %GPU 0%.

    C1 13.04 was fine.

    I opened a support case, too.

    0
  • SFA

    Try this to see what happens.

    Open the Activities Window and leave it open.

    Close C1 and wait a minute to ensure all closing processes are finished.

    Now open C1 again. The Activities Window should open as it starts.

    Check the Activities Window to see if there is any information presented about the GPU assessment process and kernel building.

    My Windows system has a problem on first start after an update when the hardware check discovers 2 OpenCL capable devices - the Intel CPU with built in but low power for OpenCL and an NVidia GPU which, being old and a notebook, only just has enough capability to make it worth using..

    The Intel will always be rejected for lack of power. However, for about the past year some change in, I suspect, Windows means that the attempt to build a kernel fails for the Intel option at about 93% complete (so about 48% on the Activity graphic).

    That crashes the whole process. A the Intel option is identified as the first device to have a kernel built the NVidia GPU is not processed before the crash occurs. So Hardware acceleration shows up as not working.

    When I restart C1 the check is run again. C1 needs to confirm that the configuration available to is is the same or to reassess any changes that may have been made.

    C1 sees that the Intel option failed last time and skips it for the new assessment. (This is fine because if it completed it would be rejected anyway due to lack of processing power.)

    It then moves on to the NVidia GPU and completes its kernel build successfully. Hardware Acceleration then shows as in use.

    For all subsequent starts (until something changes that forces re-assessment) the system will know that the Intel option should be ignored and will find the NVidia kernel ready for use.

    It is possible that something similar may be happening with Macs so it may be worth checking.

    There should be a log file for the process that, ultimately, indicates what has been identified as available and whether the process worked. It takes a little care and understanding to interpret  - especially if it fails in part or in full.

     

    It is also possible that Mac's run this process in an entirely different way to Windows machines and so this would be of no help. But it should not take long to run the checks so little time would be lost.

    There are documents in the in the Support FAQ system that relate to dealing with GPU/OpenCL problems and how to reset things when a failure means that the reassessment is never repeated  - thus forcing re-assessment on the next start up.

    To find the latest documents I would suggest using the search facility. With OS and GPU driver changes appearing with ever greater frequency the whole assessment process could be prone to frequent changes that influence the need to re-assess the capability and of course any change introduces the potential for different components to become out of step and processes to fail but with changes to the steps required to correct the problem of the moment.

    I would still report this to C1 but perhaps the immediate local problem can be resolved (or a way of working around it can be found) by running these checks.

     

    HTH.

     

    Grant

     

    0
  • Olaf Pokorny

    My MacMini has no second video card to try out, only the builtin Intel Iris GPU. It's not the fastest in the world for sure, but it was powerful enough to get activated for hardware acceleration until CO 13.0.4. In CO 13.1 the GPU is detected and initialised correctly but the kenel build process breaks after nr. 934/1298. „Compilation of CloneBlend3 failed“ states the CaptureOneICOCL.log. The error is explained as:

    <program source>:6:8: warning: no previous prototype for function 'z1b3e97e0d4'
    float4 z1b3e97e0d4(int x,int y,__global float*z42064e4f2e,__constant uint*

    …

    <program source>:159:16: warning: unused variable 'z135749130b'
    .s3);const int z135749130b=z377b7c40c9.s4;float z746ff6d23f;float z5fa3549b57=
    ^
    Error: parse error.

    For me this looks like a bug in the kernel compilation source code. As said before, I filed a bug report to CapureOne including these logfiles. I hope they are able to fix the problem at Phase One.

    0
  • SFA
    • Edited

    OK, that seems like a very similar situation to mine but as I recall the kernel build problem arose with V12 as well. It seemed to coincide with one of the last Win7 updates  - I am trying to avoid the seemingly endless Win 10 update sagas by staying with Win 7 for as long as possible.

    The power of the built in Intel on my system has never been viable for C1 compared to the CPU route so that fact that it fails to initialise is not a problem for me, just a very slight inconvenience from time to time.

    I have assumed it to be a local problem and Microsoft related. Maybe not if you have something similar happening but on Mac.

    Of course it could be Intel I suppose? Bad parameters or something.

    0
  • FirstName LastName

    Dear, I'm using CaptureOne 13.1.3.13 and NVIDIA GTX 1660Ti but during editing or when I save and process the image, CaptureOne uses only the integrated Intel GPU.

    In the NVIDIA control panel I've forced the use of the NVIDIA GPU for the APP CaptureOne, but still the system uses only the Intel GPU. Currently I'm using Windows 10 ver 2004 build 19041.508

    I've already tried to rebuild the kernel delating the ImageCore but without any sucess. However in the log file of captureone I've found this error:

    2020-10-20 01:11:03.094> (ERROR) bin file failed parse [C:\Users\Archimede\AppData\Local\CaptureOne\ImageCore\13.1.3.13\ICOCL.bin] (verificationCode=2)
    2020-10-20 01:11:03.094> (ERROR) bin file failed parse [C:\Users\Archimede\AppData\Local\CaptureOne\ImageCore\13.1.3.13\ICOCL1.bin] (verificationCode=2)

    And for some reason the Intel GPU is the device zero, instead of NVIDIA:

    Found 1 OpenCL devices on platform 0 (Intel(R) OpenCL)

    Found 1 OpenCL devices on platform 1 (NVIDIA CUDA)

    OpenCL Device 0 : Intel(R) UHD Graphics 630

    OpenCL Device 1 : GeForce GTX 1660 Ti

    Please help me to have a good performance when using captureone.

    Thanks.

    1

Post is closed for comments.

Didn't find what you were looking for?

New post