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⚠️ Please note that this topic or post has been archived.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

GPU cores

Comments

14 comments

  • Wesley

    24 core GPU seems to be a special order so no one has had their hands on it yet.

    0
  • Tom Ang

    Sure. I wondered if any GPU gurus had any theoretical thoughts.

    0
  • BeO
    Top Commenter

    Theoretically, 32 cores should perform better than 24, for the steps in the processing pipeline which utilize the GPU.

    You don't have to be a guru to state this theory... :-)

    0
  • Tom Ang

    OK. You're right. Even I could come to that conclusion, theoretically. Ask a question badly ...

    Perhaps a better set is:

    (a) does C1 make heavy use of the GPU cores in its processing? (b) If so, which processes? And (c) would there be a practical, noticeable improvement in processing speeds with the bigger core for those processes?

    0
  • BeO
    Top Commenter

    C1 makes heavy use of GPU (for some steps in the processes).

    But every system has a bottleneck (or many bottlenecks, depending on different tasks and conditions (e.g. below list)). A bottleneck is either the slowest step or component in a sequential process, or the step or component which prevents other steps or components from working to their full capabilities, if parallel. In my system, sometimes the GPU works 100%, sometimes it does not, presumably slowed down by the CPU or data feed. (My system is Windows but this general principle applies to all operating systems.).

    My point is, it always depends on many variables. Here's a non-exhaustive list:

    Preview generation (default or incl. your own auto-adjustments), exporting/processing, to which file format and settings, tool sliders in actual use (which tools), speed when browsing/viewing thru your catalog, driving a hi-res or normal-res monitor (preview size, preview used, zoom), all this depends on CPU and GPU, RAM, storage and system architecture to variable degrees; and even depending on which camera the files come from the usage of the system components can differ (I think Canon and Fuji are/were candidates for heavier CPU useage?). Even the image content has impact on performance e.g. when you export to jpg.

    Also which tools you use/don't use (e.g. keystone correction, lens tools, and maybe the upcoming tools like panorama?), same applies for layers and masks.

    Some tools could be CPU for some camera vendor's files, thus could be a bottleneck which leaves some portion of your GPU idle.


    Copy and apply adjustments?, incl. layers? including Luma Range? (attention: RAM eater in my case)

    WIthout actually testing 32 cores vs. 24 cores on a predefined test set and repeatable test cases and procedures, you'll never get any better answer than our trivial theory which we began with...

    Do you have the possibility to borrow from a friend or even to rent a new MBP and test your typcial images, and typical image edits, incl. your preferred external monitor? And what processing steps you care most, and what user interaction smoothness you actually expect.

     

    0
  • Tom Ang

    Thank you! That's a lot better than trivial; indeed, it's the kind of 'it depends' answer I like.

    Short of testing two machines side by side with my typical work (55MP files, burn/dodge, Color Editor), I think that's as far we can go for the time being. But what I've learned is that saving a few seconds here and there every day for years add up to good ROI on the cost of higher spec machines.

    Thanks again.

    0
  • RK

    I have a support ticket open since 26-10 regarding C1 21 GPU use on my set up. 

    16" MBP 2019 -2.6 i7 -32Gb RAM - AMD radeon 5300M 4Gb, with and without external 4K monitor.

    C1 uses very little, if any, GPU capacity, especially during preview display when changing between images, sometimes 3-4 secs. Frustrating to compare images when on a shoot.

    I get there are many parameters and have done extensive testing. Some changes slightly improve the lag but don't seem to increase GPU use. 

    This issue has been raised elsewhere on the forum with no real conclusions.

    I think the suggestion of testing a machine with your own images/workflow is a good one. 

    When / If I get any info from support I will update.

    take care

    1
  • Wesley

    If I recall correctly, preview generation is mainly done by CPU & SSD than kept in RAM and exporting is where GPU is used the most.

    1
  • RK

    During my investigation and testing I went through the C1 documentation. Please see below for an excerpt from the FAQs.

    it states 

    Hardware acceleration will dedicate memory to several actions in Capture One:
    • Preview Update with different settings, styles, etc.: RAM
    • Sorting/Rating: CPU cores and SSD speed
    • Fit Image to Screen: GPU cores
    • Process time: GPU processing units and CPU's and RAM

    https://support.captureone.com/hc/en-us/articles/360002412798-What-does-Hardware-Acceleration-do-and-how-do-I-use-it-in-Capture-One-

    My expectations aren't outside C1s parameters.

    1
  • Tom Ang

    Thank you, gurus one and all, for being so free with your knowledge and experience. Glad I asked!

    Have a great day!

    0
  • BeO
    Top Commenter

    You too. In case you decide to buy one let us know your experiences...

    Cheers,
    BeO

    0
  • Tom Ang

    Oh, I certainly intend to get one. My MBP 17" from 2009 is ready to go out to pasture as a music server.

    I was trying to decide between the 24-core and the 36-core. For the NZ$350 more, I think I'll bear with the pain for small gains over the years!

    Watch this space!

    0
  • Tom Ang

    Bless these geeks who have the time and inclination to run these tests and save us the bother! Thanks for sharing!

     

    0

Post is closed for comments.