Performances using 3 stages left/mid/right third.

  • Hi,

    I have these performances on a pc running win7
    I have a performance issue that is a bit strange.
    When using a triplehead2go, it's great to setup three stages in prefs and to use the left/mid/right third function.
    But there is a huge difference between performances like this and performances using only one wide stage.
    I have a simple test patch:
    movie player ( movie 1080@25fps )
    3 pixel choppers
    3 projectors
    The performance is 48 fps if I have only one wide stage
    The performance is 20 fps if I have 3 stages
    This avoided me to use the left/mid/third function and to use only one wide stage, covering 3 beamers.
    And this makes patching a bit more complicated on some points ( mostly stage/projector settings on tour )
    linked are 2 screenshots
    bug report posted

    0fced3-performance-1-stage.jpg 918087-performance-3-stages.jpg

  • I am preparing for a December performance using three projectors and the th2go. I have used three movie players running different movie files linked to three projectors in the left/mid/third pref setting. The performance is 30fps. The Chop Pixels actors must be using a lot of cpu capacity.

  • Beta Tester

    How does performance compare with three movie players and three projectors that are either going to separate stages or separate thirds of one wide stage?

  • Hi,

    I think the pixel chopper is not a problem. I use it a lot with high resolutions ( 3072x768 ) and I have no problems.
    I did a second test:
    setup one:
    1 stage, 1 movie, 3 pixel choppers, 3 projectors

    result: 49 fps
    setup 2:
    3 stages, 3 projectors / no pixel chopper

    result: 23 fps
    I think there is really a performance problem under windows with this ( left/mid/third ) function.

    beb0e1-performance-1-stage-pixchop.jpg dc785e-performance-3-stages-no-pixchop.jpg

  • Dear Mehdi,

    What's happening under the hood is that I am changing the OpenGL "viewport" -- once for left, once for mid, once for right. I am surprised if there is a performance penalty for this, but I'm sure what you are reporting is accurate. I will check into it once I solve more urgent problems (i.e., a complete freeze up) in 1.3.0f25.
    Best WIshes,