Lightroom 'Test Mule' on the nMP

Discussion in 'Mac Pro' started by Alchemist, Jan 19, 2014.

  1. Alchemist, Jan 19, 2014
    Last edited: Jan 26, 2014

    Alchemist macrumors regular

    Alchemist

    Joined:
    Dec 22, 2004
    Location:
    UK
    #1
    UPDATE // I thought it'd be useful if I stick a link to the Spreadsheet here at the top of the thread:

    Last updated - 26th Jan 2014

    Latest Test Mule Spreadsheet

    Some of you may have seen my recent article about Lightroom on the New Mac Pro.

    ......

    If you're interested in running the exact same tests, using the exact same images, I've now made the 'Test Mule' folder available along with a Procedures document explaining exactly how everything was done. All the details here:

    Test Mule - Benchmark Your Machine

    If you run the tests, I'd be grateful if you'd drop me a line with the results. It'd be good to get some sort of comparable database going from a Lightroom users POV.

    There's so much out there at present about the nMP but the focus (unsurprisingly) has been on video work. Hopefully this will be of some use to photographers.
     
  2. FredT2, Jan 19, 2014
    Last edited: Jan 19, 2014

    FredT2 macrumors 6502

    Joined:
    Mar 18, 2009
    #2
    Brilliant Tony! Finally a benchmark I can run to see exactly what I could expect in improved performance in Lightroom.

    I ran the tests on my 2012 2.6 i7 Mac Mini with 4GB and fusion drive, with these results:

    Duplicate: 32s
    Import: 1m 50s
    Export1: 5m 41s
    Export2: 5m 35s
    Render: 5m 33s
     
  3. ozbimmer macrumors member

    Joined:
    Jun 15, 2012
    #3
    Hi Alchemist,

    I am going to give the "Test Mule" a try on my RMBP. Results coming soon :)
     
  4. Alchemist thread starter macrumors regular

    Alchemist

    Joined:
    Dec 22, 2004
    Location:
    UK
    #4
    Great, looking forward to it. Many thanks to Fred for running the tests and posting them. I'll be adding them to to the database. I'm thinking of transferring the results into a Google Docs setup or something similar possibly so I can make the details available to all.
     
  5. ozbimmer macrumors member

    Joined:
    Jun 15, 2012
    #5
    Test 1: 14.6s
    Test 2: 2m 1s
    Test 3: 5m 7s
    Test 4: 5m 11s
    Test 5: 5m 16s

    OS: Mac OS X 10.9.1
    LR: 5.3

    Mac Pro 5,1
    CPU: Intel Xeon W3680, 6 cores, 3.33GHz
    Storage: OWC Mercury Accelsior 512GB PCIe SSD
    RAM: 48GB 1333MHz DDR3
    GPU: ATI Radeon HD 5870 1024MB

    I am going to do the test on my Dell next.
     
  6. Alchemist thread starter macrumors regular

    Alchemist

    Joined:
    Dec 22, 2004
    Location:
    UK
    #6
    Thanks for sharing these results. What sort of machine was this? Mac Pro 5,1?
     
  7. ozbimmer macrumors member

    Joined:
    Jun 15, 2012
    #7
    Yes

    ----------

    One more thing: I observed that the CPU usage was around 80-90% during the test, or 5-5.5 cores equivalent.
     
  8. FredT2 macrumors 6502

    Joined:
    Mar 18, 2009
    #8
    So it was actually using all six cores? There's a report floating about that Lightroom only uses four.
     
  9. Alchemist thread starter macrumors regular

    Alchemist

    Joined:
    Dec 22, 2004
    Location:
    UK
    #9
    If you take a look at my report, you'll see an image, which I think I shot during the export tests that shows LR using 6 cores but with no hyper threading.
     
  10. FredT2 macrumors 6502

    Joined:
    Mar 18, 2009
    #10
    Thanks Tony, I missed that.
     
  11. ozbimmer macrumors member

    Joined:
    Jun 15, 2012
    #11
    My observation is the same as yours - cores are used but no hyper-threading.
     
  12. ozbimmer, Jan 20, 2014
    Last edited: Jan 20, 2014

    ozbimmer macrumors member

    Joined:
    Jun 15, 2012
    #12
    Test 1: 11.23s
    Test 2: 1m 49s
    Test 3: 6m 25s
    Test 4: 6m 52s
    Test 5: 9m 39s

    OS: Windows 8.1
    LR: 5.3

    Dell T7600
    CPU: 2x Intel Xeon E5-2680, 16 cores, 2.7GHz
    Storage: OWC Mercury Accelsior 1TB PCIe SSD
    RAM: 256GB 1600Mhz ECC/Reg DDR3
    GPU: Nvidia Quadro K5000 4GB


    I would love to see a Haswell 4c and the nMP 4c running these tests :)
     
  13. jbg232 macrumors 65816

    jbg232

    Joined:
    Oct 15, 2007
    #13
    Wow, this is a great thread! I'm an aperture user on a 2008 Aluminum Macbook so won't try this out but am looking at machines to upgrade to in the near future and my two contenders are the mac pro vs the fully loaded iMac.

    My only criticism of the tests you performed in the store were that the "fully loaded" iMac at the apple stores has a fusion drive which will not be nearly as fast as a 1tb pci ssd (available as an add-on for the iMac) in the finder duplicate test if the system has over 128gb of data on it (which if I remember correctly, they do.)

    Also, I don't think the fusion drive ssd's are as fast as the dedicated ssd's. Someone feel free to correct me if I'm incorrect about that. From my readings on the matter I think the iMac SSD tops out at about 720mb/s while the mac pro is about 990mb/s on blackmagic disk tests. The difference between 9s and 32s is too large to be accounted for by that ssd difference which is why I believe the HDD is being put into play in the "fully loaded" iMac. Curious to see the tests with a pure 2013 pure SSD iMac.
     
  14. FredT2 macrumors 6502

    Joined:
    Mar 18, 2009
    #14
    No doubt that the iMac was using disk drive. However I'm quite sure that the only test that would be significantly different on a pure SSD iMac would be that finder duplicate.
     
  15. jbg232, Jan 20, 2014
    Last edited: Jan 20, 2014

    jbg232 macrumors 65816

    jbg232

    Joined:
    Oct 15, 2007
    #15
    Well, it's hard to know because all the functions that the op tests do rely on storage speed and the mac pro ssd will be much faster than the iMac hdd which is why an ssd to ssd comparison is needed.

    Also, to the op: I would recommend putting the original blog entry in its entirety into the first post. I know that would be a lot of work but would be much more likely to generate views and other people doing tests. At least the main findings table should be in the original post. When others are performing the tests they should have your numbers to compare to in the thread itself. Just a suggestion.

    Edit: Barefeats did testing of the pure SSD vs the fusion SSD here. The pure SSD is much faster for both small and large sequential writes which is what these tests include.
     
  16. ha1o2surfer, Jan 20, 2014
    Last edited: Jan 20, 2014

    ha1o2surfer macrumors 6502

    Joined:
    Sep 24, 2013
    #16
    A W110ER: http://www.avadirect.com/gaming-laptop-configurator.asp?PRID=24271

    3840QM (3.6Ghz 4c Turbo overclock, TDP raised to 65w)
    GT 650m 2GB GDDR3 (disabled, used IntelHD 4000)
    16GB Ram
    512 PLextor SSD (no ram cache, I disabled it since it's technically cheating)
    LR 5.3 x64
    Windows 8.1
    I also did a fresh restart of Windows and then opened lightroom (since I was already in the program)

    Test 1: 22.3s
    Test 2: 1m 27s
    Test 3: 4m 24s
    Test 4: 4m 13s
    Test 5: 4m 18s

    EDIT: I might have had better scores but I had a lot of things open including a couple of windows 7 VM's. not very taxing but oh well, still happy with my scores lol
     
  17. Alchemist thread starter macrumors regular

    Alchemist

    Joined:
    Dec 22, 2004
    Location:
    UK
    #17
    Thanks to all those who have submitted results thus far. Let me address some points:

    - I recognise that the iMac and the Mac Pro store models aren't entirely comparable (the iMac was short on RAM also) but it's what was available. While it'd be nice to be able to isolate system components entirely, hopefully a large database of scores will enable a user to get an idea as to what elements in a system are integral to the overall LR performance. I'd also point out that the iMac with Fusion drive bested the Mac Pro in the import test. I've previously stated that I feel like this test needs repeating, but I'm confident of the time I got out of the iMac.

    - In terms of how the Fusion drive operates, Ars has some insights here: http://arstechnica.com/apple/2012/10/more-on-fusion-drive-how-it-works-and-how-to-roll-your-own/

    - Will perhaps make a consolidated page on my blog for all things Lightroom testing. For now, it is first and foremost about showing work and this is somewhat off-topic.
     
  18. theharpster macrumors newbie

    Joined:
    Jan 20, 2014
    #18
    Ran first two

    Test 1 <9 sec
    Test 2 ≈12 sec

    I'll try to get to the other ones later.

    rMBP 11,3
    1TB PCIe
    16GB RAM
    2.6 GHz i7
     
  19. Alchemist thread starter macrumors regular

    Alchemist

    Joined:
    Dec 22, 2004
    Location:
    UK
    #19
    Thanks for running the tests. I'll tell you now though that you've run Test 2 incorrectly. There's no way a time such as that is achievable. My guess is that you've used the 'Add' command, or possibly the 'Move' command rather than 'Copy' during the import test.
     
  20. wonderspark, Jan 20, 2014
    Last edited: Jan 20, 2014

    wonderspark macrumors 68030

    wonderspark

    Joined:
    Feb 4, 2010
    Location:
    Oregon
    #20
    This is using Lightroom 4, just to be clear. Thought you might find it interesting.

    Test 1: 7.05s
    Test 2: 57s 1 min 30 seconds (I must have screwed something up the first time, but this after three retests.)
    Test 3: 5m 14s
    Test 4: 5m 16s
    Test 5: 5m 21s

    OS: Mac OS X 10.8.5
    LR: 4.4

    Mac Pro 5,1
    CPU: Intel Xeon W3680, 6 cores, 3.33GHz
    Storage: 8-disk RAID 6 via Areca 1880ix-12 RAID card w/1GB cache
    RAM: 32GB 1333MHz DDR3
    GPU: ATI Radeon HD 5870 1024MB
     
  21. FredT2 macrumors 6502

    Joined:
    Mar 18, 2009
    #21
    I ran all of the tests on both Lightroom 4.4 and 5.3, and times were almost identical across the board.
     
  22. ozbimmer macrumors member

    Joined:
    Jun 15, 2012
    #22
    Test 1: 14.24s
    Test 2: 1m 37s
    Test 3: 4m 40s
    Test 4: 4m 41s
    Test 5: 4m 50s

    OS: Mac OS X 10.9.1
    LR: 5.3

    MacBook Pro Late 2013 (11.3)
    CPU: Intel Core i7 4960HQ, 4 cores, 2.6GHz (Haswell :))
    Storage: Apple 1TB flash
    RAM: 16GB 1600Mhz DDR3
    GPU: Intel Iris Pro 5200 1024MB, NVIDIA GeForce GT 750M

    I think storage speed plays a minor role in export.


    Hi Tony, in your article you mentioned that "3600px on long side, 90% quality, sharpen for matte paper (low)". Do you think you could update the procedure document (TEST 3 point 7) to reflect this? Thanks for the benchmark :)
     
  23. wonderspark macrumors 68030

    wonderspark

    Joined:
    Feb 4, 2010
    Location:
    Oregon
    #23
    Test 1: 16.2 s
    Test 2: 1 m 59s
    Test 3: 6m 01s
    Test 4: 6m 26s
    Test 5: 6m 02s

    Mac OS X 10.9.1
    LR 4.4

    2012 MacBook Pro (9,1) 15" anti-glare 2.3GHz i7 quad core
    16GB RAM @ 1600MHz
    GPU: Intel HD 4000 1024MB, NVIDIA GeForce GT 650M 512MB
    Working disk: Samsung 840 EVO 1TB
     
  24. Alchemist thread starter macrumors regular

    Alchemist

    Joined:
    Dec 22, 2004
    Location:
    UK
    #24
    Thanks for posting all the results guys. Just uploading an updated procedures document with some less ambiguous wording.

    I'm adding these to the database now and will hopefully get things online for people to have a look at tonight.
     
  25. Alchemist, Jan 21, 2014
    Last edited: Jan 22, 2014

    Alchemist thread starter macrumors regular

    Alchemist

    Joined:
    Dec 22, 2004
    Location:
    UK
    #25
    Spreadsheet

    Here we go guys. Still got some updates to do, and only the top table is relatively up-to-date but it's a start. While I'm quite tech savvy I'm not a tech super ninja so if I have anything wrong, let me know. There are some blanks in a few places while I wait for results. I also didn't realise that Number for iCloud only allows sharing with editing turned on so I've made a backup of the spreadsheet but please don't mess with anything!

    Thanks!

    Test Mule - Benchmarking Spreadsheet
     

Share This Page