Skip navigation
Noel Carboni 23,526 posts
Dec 23, 2006
Currently Being Moderated

Photoshop CS5 Users - Performance Checks Wanted From You

Jan 28, 2012 1:34 PM

A few days ago, while discussing various system configurations in answer to posts asking about what's best in new computers, it occurred to me that while we all know our own computers pretty well, it's practically impossible, without a whole bunch of different systems to work with on hand, to judge what system works best with Photoshop.  Is more cores better?  A superduper video card?  Huge RAM?

 

Got a few minutes?

 

I propose we try a few "real world" things with Photoshop CS5 on our systems and report how quickly/interactively it does them, as well as our system configurations, in an attempt to understand what works best.  This isn't a "one upsmanship" contest, but just a poll to see what systems really blaze through Photoshop work.

 

Let's try to keep the list of "some things" fairly small, so we can all easily try them.  And the timing isn't too critical... Use a stopwatch if you can, or just watch the second hand on your watch or a clock and guestimate the timing to the nearest second.

 

List these things about your system:

 

A. Your architecture (PC / Mac) and what OS you're running.

B.  Number of processors / processor models / GHz speed.

C.  RAM size, speed.

D.  Hard drives (OS, Photo storage, Photoshop scratch), their sizes and speeds.

E.  Video card (model, graphics memory on board).

F.  Your Photoshop Preferences settings - Amount of RAM you've allotted Photoshop, whether you have OpenGL enabled, scratch disk allocations.

 

Here are the things I think might be useful for multiple people to try and let us know the time it takes:

 

1.  Cold start Photoshop, from where it is not running.  How long does this take from start to everything on the screen painted and ready to work?  Do it several times.  If you use both the 32 bit version and the 64 bit version, see if there's a difference in startup time for each.  This test represents how long you'll have to wait when inspiration hits you and before you can start working.

 

NewTestImage.jpg

 

 

2. Create a big, new RGB image, 10000 x 8000 pixels in size, 16 bits/channel.  Time how long it takes to File - Save As a .PSD in the location you often save files.  Make sure you're set for Maximum Compatibility for saving PSDs.  This gives your disk subsystem a bit of a workout.

 

SaveAsTiming.jpg

 

 

3. Choose a round brush, set the Hardness to 0% (i.e., as fuzzy as possible), Spacing down to 10%, disable all the dynamics, and set the foreground color to something other than the background color of your image so you can see your strokes.  How do brush strokes on your screen keep up with you interactively under each of the following conditions?  If not, how long does it seem to take to catch up to where you just painted?  Does it follow everywhere you moved your mouse, or at some point start to skip places where your mouse went if you paint quickly?

 

3a. Brush Size 100px.

3b. Brush Size 200px.

3c. Brush Size 300px.

3d. Brush Size 400px.

3e. Brush Size 600px.

3f. Brush Size 1000px.

3g. How big does the Brush Size have to be at 10% spacing and 0% hardness before it gets too laggy to be practically useful without slowing your strokes down hugely.

 

BrushStrokes.jpg

 

 

4.  Filter - Blur - Lens Blur.  Change all the controls to 0 then the Radius to 20 pixels and time how long it takes to fill the preview with the image (all checkerboard gone).  You can cancel the Blur filter after the display is filled.  This is a single-threaded plug-in.

 

LensBlurTiming.jpg

 

 

5.  How long does it take to complete an Unsharp Mask on the image you've painted, using a setting of 250% amount, 250.0 pixel radius, 0 levels?  Time how long it takes after hitting the OK button to Photoshop being ready for you to do something else.  This is a multi-threaded plug-in.

 

UnsharpMaskTiming.jpg

You can go ahead and close the above image after the Unsharp Mask operation.

 

 

6. How long does it take to convert a Canon EOS 7D and a Nikon D7000 image with default settings, between hitting the [Open] button in Camera Raw and the image appearing in Photoshop to be ready to edit?  You can find raw files online to try with.  This is where I got the ones I tried:

 

http://www.imaging-resource.com/PRODS/E7D/E7DFARI00100.CR2.HTM


http://www.imaging-resource.com/PRODS/D7000/D7000FARI00100.NEF.HTM

 

 

Please don't be shy about posting your results. 

 

Your computer is what it is and there's NOTHING to be ashamed of.  Some are going to be faster than others at different things, and I think it will be really enlightening to see how the different system architectures handle these operations.  It can be a more or less objective way to see what architectures really do give great Photoshop performance.  Then when people ask "what's the best architecture" to get to run Photoshop, we can respond with more confidence and point them to this thread.

 

I'll post my own results in a bit.

 

Thanks!

 

-Noel

 
Replies 1 2 Previous Next
  • Currently Being Moderated
    Jan 28, 2012 5:10 PM   in reply to Noel Carboni

    So this is how PC users spend their time? Good luck with that.

     
    |
    Mark as:
  • Trevor Dennis
    5,931 posts
    May 24, 2010
    Currently Being Moderated
    Jan 28, 2012 11:16 PM   in reply to Noel Carboni

      It does look a wee bit of a hassle Noel.  I did wonder if there was too much chance of variables and user error making the results unsafe as well.    How did you measure the times BTW?

     

    I had a go at the brush thing, but I was able to use a max sized 2500 pixel brush with vertually no lag.   I'll try that lens blur thing seeing as you have gone to so much trouble.

     

    Sorry I lied about the brush size.  I had used a fully hard brush, (but that's par for the course with Windows users )

     

    After reading the instructions properly and using a (Mac like) fully soft 1000 pixel brush,  I could write the 1000 on the image with just a tiny bit of lag, but if I did a (very) vigorous scribble the brush lagged by about 0.5 seconds by the end of it.

     

    3930K @ 4Ghz and 32Mb DDR3 1600  GTX570

     
    |
    Mark as:
  • Currently Being Moderated
    Jan 28, 2012 11:34 PM   in reply to Noel Carboni

    A. I have a custim built PC running windows 7 64bit and featuring an i7 950 running at the stock speed of 3.04 Ghz. 4 cores 8 threads.

    B. 6 GB DDR3 RAM running at 2,000 Mhz.

    C. One hetachi 500GB 7200 RPM 3gb/s deskstar drive.

    D. One hetachi 1TB 7200 RPM 6gb/s deskstar drive.

    E. One 60GB SSD OCZ Agility 3 6gb/s (photoshop and windows are installed on this drive)

    F. Nvidia 9800 GTX+ 512 MB

    G. OpenGL disabled Photoshop allowed to use 3GB of RAM of 6GB

     

    1. Photoshop took 3 seconds to come up from a cold start.

    2. Saving the 10000x8000 file as PSD took 2 seconds.

    3. Brush sizes start to get laggy at 1,100px but at 1,000 and under it was fine.

    4. lense blur took a good 8 seconds to come up in the window and to save took close to a minute.

    5. The unsharp mask took 8 seconds to load after clicking ok.

    6. cannon vs nikon they seemed to be the same as far as I could tell closer to 2 seconds.

     
    |
    Mark as:
  • Pierre Courtejoie
    7,043 posts
    Jan 11, 2006
    Currently Being Moderated
    Jan 30, 2012 8:28 AM   in reply to Noel Carboni

    Hi, Noel!

    There might be a need for further information: is disable flate compression plug-in installed or not? http://kb2.adobe.com/cps/902/cpsid_90248.html

     

    Also, the preferences>Performance History and Cache should be stated (tall and thin, or large and flat, default) as well as the number of history states.

     

    The size, and presence or not of layer thumbnails, of the info or navigator panel could be a parameter that can influence the results.

     

    (This could also become a how to optimize Photoshop for certain tasks thread)

     
    |
    Mark as:
  • Currently Being Moderated
    Jan 30, 2012 11:15 AM   in reply to Noel Carboni

    Naw, dmcrescent's just too busy admiring himself in the reflection of that shiny Macbook screen...

    :-) ;-P :-> =|->

     

    <<there, added smilies for you>>

     
    |
    Mark as:
  • Pierre Courtejoie
    7,043 posts
    Jan 11, 2006
    Currently Being Moderated
    Jan 30, 2012 9:18 AM   in reply to Noel Carboni

    Please stay on topic, and no plaform wars post will be allowed, as anywhere else on the forum.

     
    |
    Mark as:
  • Currently Being Moderated
    Jan 30, 2012 10:39 PM   in reply to Noel Carboni

    First, I would suggest some standard benchmark tool so that we know where the system stands with respect to a common group of tests not formulated on the spot. I don't recall ever running a validation on a system without such standardization. Even then, we had to run the tests at least 5 times and average the result, or even compare mean against average, with a larger suite of tests.

     

    With a common mobo, cpu, video card and ram, the range between individual setups is ascertained and  plugged into the calculations.

     

    I realize we won't go quite that far, but seriously, what the best individual system does here will only be  a starting point. For instance, I can overclock. I can mess with memory timing and latencies. For some PS operations, I'll see a difference, others, not.

     

    A true test will hold everything constant then change one parameter. Tedious and boring, but oh so informative!

     

    Lawrence

     
    |
    Mark as:
  • Trevor Dennis
    5,931 posts
    May 24, 2010
    Currently Being Moderated
    Jan 31, 2012 12:52 AM   in reply to Noel Carboni

    Noel we used to have an OK Photoshop benchmark tool over on the flickr Photoshop Support Group but results table has been broken for several weeks now.  I have just PM'd the guy who started that thread and asked if he might get round to fixing it, but I have a feeling I am not the first person to ask, so I don't have high hopes.   The nice thing about this benchmark was that everyone used the same downloaded image, and it ran from a script, so there was limited user input, and the times were recorded by the computer - not that I have anything against your egg timer  stop watch. 

     
    |
    Mark as:
  • Trevor Dennis
    5,931 posts
    May 24, 2010
    Currently Being Moderated
    Jan 31, 2012 5:26 AM   in reply to Noel Carboni

    It just seemed to me that a one click, no nonsense, script driven test is more likely to get people taking the trouble, and there were a fair number of systems on the results table before it broke.  I have already heard back from the guy behind it, and he is making moves to get it fixed.

     
    |
    Mark as:
  • Currently Being Moderated
    Jan 31, 2012 6:02 AM   in reply to Trevor Dennis

    These kinds of tests are run at Anand Tech and Toms Hardware. Sufficient time to search these sites is an excellent way to get some good answers. They have the advantage of rigor which, if I were to sum up what I posted above, it would be one word: rigor.

     

    The one click script run setup to which you linked is one I would include in a checkout after building/buying a system and one that should give good feedback when tweaking.

     

    Passmark is another test I like to use. They also provide a rating system if you care to upload your results. You can also privately do your own comparisons by downloading test results for selected systems.

     
    |
    Mark as:
  • Currently Being Moderated
    Jan 31, 2012 7:20 AM   in reply to Noel Carboni

    I'm not sure there is a simple linear relationship here.

     

    I just painted extensively with a 2500 px brush with no lag whatsoever. That is on a very modest system with an ATI 5750 video card and an i5 750 CPU. I'm sure the benchmarks are nothing to write home about.

     

    In fact I always steer away from the fastest bleeding edge. I want stability. I don't really care about speed, I care about the thing not crashing. I deliberately picked the modest i5 750 over the (then) much fancier i7's, because it was by then tried and tested as compatible with almost anything, and was known as a solid workhorse.

     

    The tortoise and the hare, you know?

     
    |
    Mark as:
  • Pierre Courtejoie
    7,043 posts
    Jan 11, 2006
    Currently Being Moderated
    Jan 31, 2012 7:42 AM   in reply to twenty_one

    Dag, did you have hardness at 0% and spacing set to 10%?

     

    Noel, that thing called real life got into my benchmarking attempts on my laptop.

    But let me second the request for changing a parameter at a time. Like we do with 64bits/32bits when using more than 4Gb of RAM. If somebody wants to send me an SSD, I'll be able to compare too

    RetouchArtists is also a very widely used Benchmarking script.

     
    |
    Mark as:
  • Currently Being Moderated
    Jan 31, 2012 7:43 AM   in reply to twenty_one

    Funny, but the i5's and the i3's are derivatives from the i7 technology. You may have a processor that cannot pass i7 requirements and have only 2 cores activated.

     
    |
    Mark as:
  • Currently Being Moderated
    Jan 31, 2012 7:58 AM   in reply to Pierre Courtejoie

    PECourtejoie wrote:

     

    Dag, did you have hardness at 0% and spacing set to 10%?

     

    OK, hardness at zero introduces a slight lag, perhaps 1/4 second. But nothing unworkable. Hardness at 100 is instantaneous.

     

    Spacing all the way down to 1% brings the delay up to perhaps a full second.

     

    But this is, as I said, 2500 pixels.

     

    I'm sure these numbers could turn ugly in an instant with a different video card driver, for instance. The thing is, I'm not sure it has all that much to do with specifications taken at face value. I believe it is more finding components and drivers that work well together.

     
    |
    Mark as:
  • Currently Being Moderated
    Jan 31, 2012 8:01 AM   in reply to Hudechrome

    Hudechrome wrote:

     

    You may have a processor that cannot pass i7 requirements and have only 2 cores activated.

    All four cores are alive and well, thank you (the 750 is the only quad i5).

     
    |
    Mark as:
  • Currently Being Moderated
    Jan 31, 2012 8:03 AM   in reply to Hudechrome

    I haven't gotten too far into the tests, as I don't use the brush for drawing only as an adjustment like Dodge, Burn etc.

     

    Nonetheless, here is some results.

     

    Saving the big picture: 2 sec

    Opening the 3 big picture 2 sec.

    Painting with a 400 pix brush at specified hardness and spacing at 25% 1.4 sec to complete a swift stroke across a diagonal. At 1000pix, 10%, 4+ sec...useless.

    Opening PS 4.5 sec Average

     

    My system runs on an AMD Athlon IIx4 @2.8G stock speed

    12G RAM 1333, default timing

    ATI5670

    HD: all spinners

    Prefs set at 80% of 11,000 available

    PC running Win7 64 bit.

     

    Getting this far is much more than a few minutes! It is now 8 AM. I commenced this test about 6AM, with approx 1/2 hr for coffee, checking e-mails, arranging  dual screens to be useful to test and record, validating setup data etc.

     
    |
    Mark as:
  • Currently Being Moderated
    Jan 31, 2012 8:51 AM   in reply to twenty_one

    Thanks. I was thinking Sandy Bridge i5, not Lynnfield.

     

    The comment concerning binning also needs to be taken with a grain of salt at this point.

     
    |
    Mark as:
  • Currently Being Moderated
    Jan 31, 2012 9:08 AM   in reply to Noel Carboni

    With mine, at hardness 100%, Spacing 10, 2500pix, 3 sec, hardness, 0, a miserable 10+

     
    |
    Mark as:
  • Currently Being Moderated
    Jan 31, 2012 9:27 AM   in reply to Noel Carboni

    What does it do at 100% hardness, all else unchanged?

     
    |
    Mark as:
  • Currently Being Moderated
    Jan 31, 2012 9:39 AM   in reply to Noel Carboni

    Noel,

    On the specified task : 10000 x 8000 pixel 16 bits/channel image, 2500px diameter round brush, hardness 0%, Spacing 10  I made a straight line path (pen tool) from corner to corner. I then commenced the timing by clicking on the stroke path icon and read the result when the stroking stopped. I used the second hand on the Windows 64 Vista  clock gadget for the measurement. 

     

    Time taken 6 secs.

    Asus geforce 9600GT 512 MB

     

    Capture.PNG

     

    Paulo

     
    |
    Mark as:
  • Currently Being Moderated
    Jan 31, 2012 9:42 AM   in reply to Noel Carboni

    Whoops, it was a smaller canvas! Will retest tomorrow, sorry for not reading the fine print...Back at my home system now and looking at delays in the 3-5 second range.

     

    Anyway, I re-read my post #19 and realize it perhaps came across as...gloating (is that the word)? I apologize for that, didn't mean to. I just honestly believe this has more to do with compatibility than anything else (the brush strokes at least).

     

    The thing with the i5 750 is that at that time it had been on the market for about 18 months. All the motherboards were ready for it, all the BIOS'es knew it inside out. The top-of-the-line iMacs had used it for a long time. It was old stuff.

     

    And I'm sure modern GPU's are more than fast enough to deal with 2500 pixel brushes. But we all know video card drivers can be less than optimal, or optimal in one configuration and not in others.

     
    |
    Mark as:
  • Currently Being Moderated
    Jan 31, 2012 10:01 AM   in reply to twenty_one

    The i5 Lynnfield was near the point of release as I was finishing up a contract with Intel, at which point I was working with the Clarksdale.

     

    Confusion rains when generic terms like i5,  i7 etc is bandied around. And yes, I meant to use the word "rains"!

     

    Some folks I know refuse to budge from Core 2.

     
    |
    Mark as:
  • Currently Being Moderated
    Jan 31, 2012 11:15 AM   in reply to Noel Carboni

    Noel Carboni wrote:

     

    Hudechrome wrote:

     


    Confusion rains when generic terms like i5,  i7 etc is bandied around. And yes, I meant to use the word "rains"!

     

    Good, because I hate it when I have to pore over such phrases as confusion reigns 

     

    So...  Can it then be said that when it reigns it pores?

     

    -Noel

    Only if you are willing to ponder over it thoroughly while standing near a poodle!

     
    |
    Mark as:
  • Currently Being Moderated
    Feb 1, 2012 3:55 AM   in reply to Noel Carboni

    OK, this time I did it entirely by the book, by everything that is holy, cross my heart and hope to die.

     

    Fast sweep from corner to corner on a 10000 x 8000 16 bit canvas. Here goes:

     

    2500 pixel brush, hardness 0 (max fuzzy), spacing 10%. - 5 seconds flat.

    1000 pixel brush, hardness 0 (max fuzzy), spacing 10%. - 1,5 to 2 seconds.

     

    2500 pixel brush, hardness 100, spacing 10 % - 2,5 seconds

    1000 pixel brush, hardness 100, spacing 10 % - instantaneous.

     

    ----

     

    So it seems you guys can go get your sticks and baseball bats and punch me over the head for speaking up too quickly. I'll just be over here in the corner...

     

    H o w e v e r... those results seem very similar to yours, Noel, and certainly not any worse. But on paper your system should be much faster than mine. So I do think I'm coming out of this with my head high after all. At the moment I only have 8GB RAM and 512 MB VRAM, and my quad-core CPU is slower than each of your two. I have DDR3, though, if that matters.

     
    |
    Mark as:
  • Currently Being Moderated
    Feb 1, 2012 4:56 AM   in reply to twenty_one

    OK, now I'm really confused, and I'm almost afraid to say this:

     

    I've tried to read this carefully, but I must still be missing something here. I still don't understand why it should take 2 seconds to save a 10000 x 8000, 16-bit flat file (psd). Here it takes a half second, and that's with an image on it (not pure white). No RAID, straight 7200 SATA2. 458 MB on disk.

     

    What's going on with these numbers?

     
    |
    Mark as:
  • Pierre Courtejoie
    7,043 posts
    Jan 11, 2006
    Currently Being Moderated
    Feb 1, 2012 5:08 AM   in reply to twenty_one

    You are counting too slowly?

    D Fosse, are you by any chance running the Disable Flate Compression plug-in, as I linked earlier in the thread? From the file size, it does not seem so.

    Compatibility on or off? (should not matter on a flat file)

     
    |
    Mark as:
  • Pierre Courtejoie
    7,043 posts
    Jan 11, 2006
    Currently Being Moderated
    Feb 1, 2012 5:09 AM   in reply to Pierre Courtejoie

    A possible solution: you are saving on a disk that is not used as the scratch disk, or windows pagefile?

     
    |
    Mark as:
1 2 Previous Next
Actions

More Like This

  • Retrieving data ...

Bookmarked By (1)