NewMacPro is about 4-5 times slower then my OldMacPro for H264 compression

Discussion in 'Mac Pro' started by rueyloon, Aug 14, 2014.

  1. rueyloon macrumors regular

    Joined:
    Sep 24, 2013
    #1
    Hello

    just wanted to share this, I just got my nMP in and for h264 compression, the nMP is about 5 times slower.

    Sometimes I use toast for conversion, toast has this thing called videoboost, which is supported by some video cards by Nvidia, the GTX670 on my old Mac Pro is supported, as is my GTX285 which is in my cupboard (actually faster then the GTX670 in this operation).

    the D500 on the nMP isn't supported hence when doing conversion to H264, the speed difference is like comparing a Mac Pro to a Macbook Air, while the OMP has reached 20%, the nMP was still in the low single digits....

    Maybe it might be different for Compressor, but I'll have to test that out later...
     
  2. simsaladimbamba

    Joined:
    Nov 28, 2010
    Location:
    located
  3. mcnallym macrumors 6502a

    Joined:
    Oct 28, 2008
    #3
    http://corel.force.com/roxio/articles/en_US/Master_Article/Using-VideoBoost-in-Toast-11

    ONLY Nvidia cards are supported using VideoBoost.

    As the nMP is using an AMD GPU then hardly surprising that the Older Mac Pro is quicker, as it can use VideoBoost on the Nvidia GPU and the nMP cannot. I would suspect is using CUDA for this which is why Nvidia only.

    Nvidia have been dialling back the compute support in the later cards, the 580 was better then the 680 series for this as well.

    http://macperformanceguide.com/blog/2014/20140729_2045-H264-video-compression.html

    A 13" MBPro can do H.264 compression quicker then the nMP as well if can use the QuickSync feature found in the CPU.

    As said do a Handbrake conversion and see what the difference is. Handbrake will purely use your CPU rather then any specific feature that may be in the CPU or a GPU boosting capability.
     
  4. foodog macrumors 6502a

    foodog

    Joined:
    Sep 6, 2006
    Location:
    Atlanta, GA
    #4
    much, much, much faster than my old MP 2010
     
  5. kennyman macrumors regular

    Joined:
    May 4, 2011
    #5
  6. brentsg macrumors 68040

    Joined:
    Oct 15, 2008
    #6
    The moral of the story is that you have to make sure your hardware supports your software.
     
  7. brand macrumors 601

    brand

    Joined:
    Oct 3, 2006
    Location:
    127.0.0.1
    #7
    Did you check the hardware compatibility list of the software that you were planning on using prior to purchasing the computer?
     
  8. AidenShaw macrumors P6

    AidenShaw

    Joined:
    Feb 8, 2003
    Location:
    The Peninsula
    #8
    It is "hardware compatible" - it correctly renders the videos.
     
  9. brand macrumors 601

    brand

    Joined:
    Oct 3, 2006
    Location:
    127.0.0.1
    #9
    Its not compatible she it comes to VideoBoost as was stated earlier in the thread. Just because I can use a CPU to encode video doesn't mean that a system is compatible if I am wanting to use the GPU for the the encoding.

    I see what you are saying and at the most basic feature of just basic encoding yes the system renders the video so you are correct. Its the advanced features that also have to be compatible.
     
  10. goMac macrumors 603

    Joined:
    Apr 15, 2004
    #10
    Unless there is something special about VideoBoost's output quality, it might be time to switch compressors.

    There are tons of good, modern compressors out there which are extremely fast. The ones that use Intel QuickSync don't even need a GPU to be fast. There aren't many CUDA only compressors left.

    I think Compressor might use QuickSync at least these days, so it should be pretty fast. I don't know if Apple has added OpenCL, but it wouldn't surprise me.
     
  11. chrfr macrumors 603

    Joined:
    Jul 11, 2009
    #11
    Compressor 4.1 did gain QuickSync support, but the 2013 Mac Pro does not support QuickSync.
    http://www.larryjordan.biz/mac-pro-video-compression/
     
  12. goMac macrumors 603

    Joined:
    Apr 15, 2004
    #12
  13. chrfr macrumors 603

    Joined:
    Jul 11, 2009
    #13
    Apple must be doing Airplay differently with those machines as the Intel specifications do not include QuickSync:
    http://ark.intel.com/search/advanced?FamilyText=Intel® Xeon® Processor E5 v2 Family
     
  14. purplecu6e macrumors newbie

    Joined:
    Aug 4, 2014
    #14
    AirPlay requires a CPU feature introduced with Sandy Bridge, I don't think Apple has come out and said what exactly it is. If it is Quick Sync, then Apple made an exception for the nMP.
     
  15. joema2 macrumors 65816

    joema2

    Joined:
    Sep 3, 2013
    #15
    As already discussed, even though Compressor, Handbrake and FCP X support QuickSync, Xeon CPUs do not have QuickSync transcoding instructions. This is because QuickSync depends on Intel's on-chip GPU: https://software.intel.com/en-us/articles/harnessing-handbrake

    This Intel web site implies Quick Sync is (currently) inextricably tied to their on-chip GPU:
    http://www.intel.com/content/www/us...uick-sync-video/quick-sync-video-general.html

    I've never seen an explanation for the chain of decisions which led Intel to this situation. They may have felt Quick Sync was a narrow-purpose consumer technology, or Xeon just couldn't afford the transistor budget at the current state of fabrication technology. Even the low-end on-chip HD Graphics 2500 GPU takes about 400 million transistors (out of 1.4 billion available in Haswell and lower-end Xeons): http://www.techpowerup.com/gpudb/1250/hd-graphics-2500.html

    From this standpoint lower-end Xeons would be wasting nearly 30% of their transistor budget on a GPU they'd never use, since those envisioned platforms always have dedicated discrete GPUs. They have a vital need to deploy those transistors for workstation-class functions like larger caches, ERC memory, etc.

    Intel made those design decisions several years ago. They likely didn't envision the current ironic situation with consumer-class CPUs with Quick Sync out-performing workstation-class Xeons in narrow transcoding tasks.

    Quick Sync is essentially a "one trick pony", but a very useful one. It is limited to certain codecs and certain parameters, e.g, single-pass H.264, MPEG-2, etc. However those are commonly used. But -- if you do multi-pass H.264, e.g. in FCP X select "better quality H.264" (Export->Master File->Settings), it is much slower. The base nMP is faster than the fastest 2013 iMac on that task.

    In theory the nMP's immensely fast GPU could be harnessed for transcoding but that's a different programming model and may take longer to upgrade software for that. Also, page 10 of Intel's IDF13 white paper implies a GPU (while more flexible) still can't achieve the transcoding performance of Quick Sync. https://intel.activeevents.com/sf13/connect/sessionDetail.ww?SESSION_ID=1178 This is because Quick Sync is essentially an on-chip ASIC (Application-Specific Integrated Circuit), whereas a GPU is a more general-purpose device.

    Despite this limitation, a Mac Pro is more broadly useful for video editing than even the highest end iMac (which I have). I am always waiting on the iMac to finish rendering, since that is a largely compute-bound task. More cores would be useful, which are only available on the Mac Pro.
     

Share This Page