Skip navigation

Lag, is it the cache?

Apr 4, 2012 7:29 AM

  Latest reply: TimShelbourne, Apr 6, 2012 3:26 AM
Replies
  • Currently Being Moderated
    Apr 5, 2012 10:46 AM   in reply to Vahakn01

    For those of you experiencing lag, are you using a pen and tablet or a mouse?

     

    Pattie

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 5, 2012 10:49 AM   in reply to Pattie F

    Pattie,

    This action uses neither... it simply strokes the path.

    T

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 5, 2012 10:55 AM   in reply to Vahakn01

    I am using a Cintiq and a Wacom tablet at the same time.  Problems happens in both.  Also while zooming in or moving parts of a file.

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 5, 2012 11:35 AM   in reply to Pattie F

    I am using a supported video card with the latest drivers, a Wacom Intuos 4 with the latest drivers.

    It occurs using the grip pen, the art pen, and my mouse (not an intuos mouse).

    My system information is posted above in another post in this thread.

    It occurs using any brush presets with any mixer brush, including those presets used in CS5 that successfully reduced or stopped the lag.

    It occurs at all zoom levels.

     
    |
    Mark as:
  • Noel Carboni
    23,526 posts
    Dec 23, 2006
    Currently Being Moderated
    Apr 5, 2012 11:49 AM   in reply to TimShelbourne

    TimShelbourne wrote:

     

    Ooops Noel! I think you'll find that when your action selects the Mixer Brush, it selects it with SAL not active! This means that even if you add the layer and activate SAL before running the action, the action cancels the SAL option. That will be why you get identical timings!

     

    I took that into account and inserted a stop so as to turn on SAL.  I really did test it properly.

     

    -Noel

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 5, 2012 11:54 AM   in reply to Noel Carboni

    Wow... that's incredible! No difference between SAL and no SAL? Now I'm getting system envy! I hate it when that happens!

    Just thought it was worth pointing out.

    T

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 5, 2012 11:59 AM   in reply to TimShelbourne

    And you were painting on the added layer,... not still the background?

     
    |
    Mark as:
  • Noel Carboni
    23,526 posts
    Dec 23, 2006
    Currently Being Moderated
    Apr 5, 2012 12:00 PM   in reply to TimShelbourne

    Well, to be fair my times aren't all that good either way, but I'm running PS CS6 in a Windows 8 virtual machine (VMware), so the GPU support is different than it would be if I put the software on my host system.  I've been resisting doing that, but perhaps I will, since some things I'd like to be able to test I really can't this way.

     

    I'm working on updating the actions and image so as to test brushing more thoroughly, including Sample All Layers and possibly using the new capability for recording of brushstrokes (which you or someone suggested a few days ago).  I think it's important we have a way we can objectively measure brush interactivity.

     

    -Noel

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 5, 2012 12:06 PM   in reply to Noel Carboni

    Absolutely Noel, we can do that... on a 4000 x 4000 pixel canvas, draw around the outer edge with the Round Fan, at a specific pixel size, counting seconds between corners, and see how long it takes for the stroke to catch up with what you're doing.

    I'm really interested in comparing results from this simple approach between nVidia and ATI/AMD cards (with a few very specific CS6 performance settings)

    T

     
    |
    Mark as:
  • Noel Carboni
    23,526 posts
    Dec 23, 2006
    Currently Being Moderated
    Apr 5, 2012 1:31 PM   in reply to TimShelbourne

    While reworking the action and controlling everything carefully, I found that I can reproduce the much slower operation with the Sample All Layers setting checked.  It seemed to take using something other than white as a base layer to get it to happen.  Or something else changed.  I could have made a mistake before.

     

    The Sample All Layers setting makes the thing run about 5 times slower.  This action will show that clearly.  Run it the first time with the Sample All Layers off (the default), then run it a second time and at the Stop set it do Sample All Layers.

     

    http://Noel.ProDigitalSoftware.com/temp/SpeedTest2.zip

     

    I don't think I sense a difference between Photoshop CS5 and CS6 though.  The above action can't run on CS5.

     

    -Noel

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 5, 2012 1:44 PM   in reply to TimShelbourne

    Right now I am seeing the problem in spades.

     

    Tim I did as you suggest, but my drawing was too erratic to figure the time. So I selected timing (at the bottom left drop down - usually used for dimensions) and simply dragged a 100px brush diagonally. The time it took to quickly swipe the stylus from corner to corner was negligible, but on 25 consecutive swipes the drawing time averaged 1.6 seconds with the shortest being .5 seconds and the longest being 2.8.

     

    That is a wild variation. .5 seconds is not a problem. 2.8 seconds is.

     

     

    So I tried it on a gray layer over a pattern and used the dodge tool. Sometimes it took 0.5 seconds to dodge at 100% with a 200px brush (typical for me), but 2 out of three passes were excruciatingly long - the worst was 5.2 seconds. It seems like the first pass would lag, the second would be fast, the next two would lag, and so on.

     

    I doubt this is a sufficiently quantitative test, but it shows the problem.

     

    Memory stats:

    In Use: 6837 MB

    Modified: 97 MB

    Standby: 8780 MB

    Free: 686 MB

     

    Memory allocated to Ps 8703 MB

     

    Win 7 ver. 6.1.7601.17514 - SP1

    i7-2600K @ 3.4 GHz

    Video: nvidi GTX 550Ti w/ 1GB

    nvidia driver 8.17.12.9610

    Physical Memory 16GB

     

     

    Ted

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 5, 2012 2:06 PM   in reply to PixbyTed

    Loaded the same 4K x 4K image into CS5 and repeated the drag the dodge tool test.

     

    The best time on CS5 (0.6 sec) was slower than the best on CS6 (0.5 sec). BUT the slowest on CS5 was 0 7 sec versus the slowest on CS6 at 5.6 sec.

     

    Ted

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 5, 2012 2:18 PM   in reply to Noel Carboni

    Noel,

    Thanks for your effort.

    I may have missed a few posts here and there, my apologies if that's so. Are you running an nVidia card or an ATI/AMD/Radeon?

    Thank you,

    T

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 5, 2012 2:17 PM   in reply to PixbyTed

    Thanks Ted... all being logged!

     
    |
    Mark as:
  • Noel Carboni
    23,526 posts
    Dec 23, 2006
    Currently Being Moderated
    Apr 5, 2012 2:39 PM   in reply to TimShelbourne

    I have a VisionTek ATI Radeon HD 5670 1 GB DDR5.

     

    Up to just now I've been running Photoshop CS6 in a VMware VM.  I've just gotten it installed on my host system, so I'm about to run through some of the same speed tests.  It starts a bit faster, and more menu entries are available (e.g., OpenCL).

     

    -Noel

     
    |
    Mark as:
  • Noel Carboni
    23,526 posts
    Dec 23, 2006
    Currently Being Moderated
    Apr 5, 2012 2:51 PM   in reply to Noel Carboni

    My measured times with the action I posted just above are as follows:

     

    12.9 seconds with Sample All Layers OFF.

    50.0 seconds with Sample All Layers ON.

     

    -Noel

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 6, 2012 3:26 AM   in reply to Noel Carboni

    After some more experimentation, this is interesting stuff.

    As you know, for a pretty unscientific test, I’ve been painting in a single stroke around the edges of a 4000 x 4000 pixel document with a bristle tip, using the Mixer Brush Tool on a separate layer above the white background. Settings for the MB were:

    Wet: 80, Load 75, Mix 39 and SAL checked.

    Bristle settings for the bristle tip were:

    Bristles 16, Length 137, Thickness 1, Stiffness 56 with no other brush dynamics active. This is using the Round Blunt shape for the brush. Brush Size: 121 pixels.

    For the first set of stroke timing tests in CS6 and then CS5 the brush was set to its default Spacing value (2%), to really give Photoshop something to chew on. Doing my pretty unscientific test of drawing around the outer edges of the document, at the various tile sizes gave these results:

     

    CS6:

    1028k: 22.8s

    1024k: 22.4s

    132k: 15.9s

    128k: 15.1s

    One interesting point here is that in CS6, completely turning off ‘Use Graphics Processor’ actually speeds the brush up quite significantly at the lower tile size settings, making the 128k: 15.1s above 128k: 12.5s! This effect is not apparent in CS5.

     

    CS5:

    1028k: 17.6s

    1024k: 17.4s

    132k: 11.5s

    128k: 11.5s

    So, I’s clear to see that there’s quite a timing difference between CS5 and CS6, across all tile sizes, with CS5 winning hands down every time.

    Now… here’s the interesting bit. Of course, we know that increasing the spacing a little on the bristle tips speeds brush response up no end, so for the second run of tests I increased it to 5%, with these results:

     

    CS6:

    1028k: 8.4s

    1024k: 8.4s

    132k: 5.8s

    128k: 5.9s

     

    CS5:

    1028k: 7.3s

    1024k: 7.3s

    132k: 5.9s

    128k: 5.7s

    So you can clearly see by this that increasing the Spacing to 5%, not only speeds up the performance of the brush (as you would expect), but oddly results in there being virtually no difference at all between the brush speed in CS6 and CS5. For me, this seems to indicate that something in CS6 is much more sensitive to low brush spacing in terms of brush speed than CS5 is? In reality, with a higher level of Spacing there is virtually no speed difference between the two.

    I’m almost tempted to say that one solution to this could be to increase the default spacing for the standard Bristle Tip Presets in the Brush Presets to 5%, rather than 2%, but that doesn’t really address the issue of exactly why CS6 is more susceptible to the slow-down effects of low brush spacing!

    In reality, increasing the spacing this much doesn’t introduce any really noticeable tyre-tracking, especially if you’re using a texture within the stroke (although I wasn’t here).

    I was convinced that the MB slow-down in CS6 was purely a tile size/re-draw problem, but whilst smaller tile sizes make a difference to overall brush speed, that doesn’t explain why the speed differences between CS6 and CS5 apparently disappear at a higher brush spacing setting. Even though you’d expect the brush to be faster in both, you’d still expect to see the marked timing differences between the two.

    Mike and Pattie, I’ll also email this to you privately.

     
    |
    Mark as:

More Like This

  • Retrieving data ...

Bookmarked By (0)

Answers + Points = Status

  • 10 points awarded for Correct Answers
  • 5 points awarded for Helpful Answers
  • 10,000+ points
  • 1,001-10,000 points
  • 501-1,000 points
  • 5-500 points