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⚠️ Please note that this topic or post has been archived.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. ⚠️

Option to switch to another GPU accelerator?

Tak Fujiwara
  • Edited

Having the export corruption problem, I wander how to switch to another GPU.

My PC carries two GPUs, one on mother board native, and an NVIDIA video card plugged in.  I use the video card currently as it has much better performance.   C1 seems to use this one.

Now, is there an option to use mother board GPU as the C1 accelerator?  I tweaked around but could not find.

Tak

0

Comments

9 comments

Date Votes
  • Pawel Szczupak

    Hello Fujiwara-san,

    As far as I remember from previous discussions and notes, the Software uses all of the GPUs it can find in your PC in parallel. Therefore there is no need to switch between them.

    Best regards,
    Pawel

    0
  • SFA
    • Edited

    Pawel is correct.

    If the Intel (?) software-based driver does not match basic processing requirements after assessment it will be rejected and not used.

    There is a log file that can provide information about the assessment process results.

    You may also have access to settings in the NVidia control panel that will force use of the Nvidia card for specific applications but that may not be necessary. 

     

    As you are running Windows you could check some reliable sources for NVidia device drivers to see whether a newer driver exists and might solve the problem.

     

    If not, try an older driver.

    0
  • Tak Fujiwara

    Thanks, Pawel and SFA.

    >there is no need to switch

    I need.     That is to avoid the malfunctioning GPU.

    >There is a log file

    Good info. I searched and found about ten log files in c:\users\(username)\Appdata\Local\CaptureOne\Logs\. Which is the one you mentioned?  Want to look into it.

    > ,,,  after assessment it will be rejected and not used.

    Does that mean the corruption will not happen, as it is tested beforehand?   The problem is actually happening, though.  I think the test is not almighty because the problem is intermittent, say, 5 pictures out of one hundred processes.

    >As you are running,,,

    Yes.

    I updated C1 sometime in Spring this year, and that was the beginning of the failure.

    At the first time when I had the problem, the driver was what Dell (My PC) provided as the latest.  Then, I introduced a driver from NVIDIA, which was the latest at that time, newer than Dell's.  It failed.  About one month later, I found further new driver at NVIDIA, that failed again.   It repeated once gain in September.  To be sure, the Intel driver is also the latest, updated twice.

    Trying older driver is not preferred as the problem started by new C1.  I think older driver will not resolve.

    0
  • SFA

    @Tak Fujiwara

    My Dell supplied driver update application seems to keep My Intel and Nvidia GPU facilities working OK.

    My previous Dell (Win 7) rejected the Intel option because it offered no benefit. The NVidia GPU on that system was just about OK to have some purpose. But Dell offered few updates.

    I installed NVidia updates but one needs to be careful as there are 2 separate development paths for NVidia drivers.

    Also, not all new drivers work better than the old ones. Somtimes one may need to reverse an update.

     

    The Log file will be the one with OCL in the name although the "Application" log file may also have information.

    Log file identification and content tends to change from time to time as new versions and releases appear.

    0
  • SFA

    @Tak Fujiwara

    I forgot to mention that you can controls some application access to the NVidia GPU using NVidia Control Panel. That might help if you are sure that NVidia is the failure point.

    0
  • Tak Fujiwara
    • Edited

    Thanks SFA,

    >using NVidia Control Panel.

    I tried it to your suggestion, in vain.

    I also looked into ImgCoreOCL.log.  I seems that C1 tests capabilities of both Intel and NVIDIA and comes out they are satisfactory.    However, it seems do not diagnose the function itself.  I also looked into Application.log.  It records the process that C1 does, with no error in it.

    There seems no solution of this problem sadly.  I will use C1 cutting GPU for a while.   Thanks anyway.

    As a side note, I want to leave some investigation result.  I hope C1 developers read the forum.  It may help them.

    1. The corruption most probably occur under following condition:

          [Clone layer is used], AND, [Crop is used with its boundary crossing the cloned area]

    Am not sure re catalog/session dependency, but I use session.

    2. Exceptions exist.  The corruption may not happen with above condition, low percentage though.  And, corruption does happen with other condition such as Heal layer is used, but occurrence percentage is low in this, too.

    I want to close this thread, as "No switching option" is the answer.

    Tak

    0
  • SFA

    Hi Tak Fujiwara,

    Based on your analysis and the description of when the problem occurs I think it might be useful to report you findings fully to Capture One using the "Submit a request" option to create a Support Case.

    When you do that and provide the information requested it will create a bug report and give the developers some information to work with. 

    Most of the Forum is a User to User resource. Your analysis seems to suggest there may a code problem in some sort of situation that may be rarely used and probably not something that a developer or tester would think of for a test plan.

    It may not even occur unless other specific configurations exist.

    There are so many possible variables that knowing the complete information about software, hardware and files involved is likely to be important to identify the reason for the problem and defining a fix.

     

    My Notebook is a late 2020 Dell Precision. It successfully uses both the built-in Intel software GPU and the NVidia T2000 that it came with. It looks like the load is usually balanced across both but there may be times when GPU usage practicality is not high and would be more efficiently delivered by the Intel device alone. Less data transfer perhaps?

    My old system (Win 7) has too little power in the Intel facility to make it useful (it is rejected) and the NVidia card is only just offering more potential capability (compared to the CPU alone) to make it useful sometimes.

    0
  • Tak Fujiwara

    >I think it might be useful to report you findings

    Yup, will do.  I got to gather much data before that.

    >There are so many possible variables

    I know.  Am a retiree from a large IT company, where I used to be a large processor HW management.  Am good at PC HW, too, down to component level, also, write/wrote a lot of softwares, so, I know the complexity.

    Thanks for your help.  Tak

    0
  • SFA

    Ah! That explains the thorough and well-described analysis.

    I picked an image with a similar sort of image are and ran some experiments with Cloning. A single clone layer seems to be OK but in some situations, 2 clone layers may be compromised when both are applied to the same area of the image. And of course things are similar if a clone action is made from 2 different selections to the same "fix" area on the same layer.

    0

Post is closed for comments.