OWC & Digilloyd test: iMac Vs Mac Pro

Discussion in 'Mac Pro' started by Loa, Dec 18, 2015.

  1. Loa macrumors 65816

    Loa

    Joined:
    May 5, 2003
    Location:
    Québec
    #1
  2. rawweb macrumors 6502a

    rawweb

    Joined:
    Aug 7, 2015
    #2
  3. h9826790 macrumors 604

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #3
    I wonder if the cMP is really upgrade to use a "proper" SSD and the top current GPU, rather than those old mid level stuff from OWC, will that make any difference.

    In fact, I agree that shows a 6 years old Mac Pro can catch up with the top of line current iMac.

    Also, this tests only take a very short period of time to complete, this area is the iMac's strongest area. The real power of the MP is it can keep the speed for 24/7. If they make another torture test which run few hours, the result may be different.
     
  4. SoyCapitanSoyCapitan macrumors 68040

    SoyCapitanSoyCapitan

    Joined:
    Jul 4, 2015
    #4
    There's also a sluggishness and random jerking that occurs when panning and scrolling around when using Photoshop on an older machine even if you have the best graphics card. Even using Windows on the cMP doesn't solve the performance. The solution is to disable GPU accelerated panning and flicking. This gives you the old school navigation without the flicking motion.

    This issue doesn't occur if you're using a new machine. On my Skylake Photoshop is almost too fast when panning and flicking. It's like sliding around on wet ice.
     
  5. ixxx69 macrumors 6502a

    Joined:
    Jul 31, 2009
    Location:
    United States
    #5
    Those benchmarks highlight some very specific performance aspects of the Macs and Photoshop.

    While I'm sure there are some users whose usage is literally described in the benchmarks, for most usages, my impression is the benchmarks over-represent certain aspects such as SSD performance and amount of installed RAM.

    There's no denying that those incremental component performance improvements over time add up to perceivably faster general system performance as well. However, I think it's worth pointing out that in "real world" usage (for most users), you could pull heavy duty PS all day with the latest iMac, and then the next day with a nMP, and not perceive any performance difference or measurable efficiency gains.

    (also, I think it has to be acknowledged that OWC is trying to sell RAM upgrades)
     
  6. Arron Rouse macrumors regular

    Joined:
    Apr 28, 2010
    Location:
    Chichester, UK
    #6
    The cMP numbers all seem to be within 1% of each other no matter how many cores/processors. To me that points to something dodgy in the benchmark software. It may be that it's only really measuring the GPU or storage.
     
  7. h9826790 macrumors 604

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
  8. Arron Rouse macrumors regular

    Joined:
    Apr 28, 2010
    Location:
    Chichester, UK
    #9
    8GB diglloydHuge test
    cMP 6 core 338.17
    cMP 12 core 338.51
    cMP 8 core 343.01
    cMP 4 core 344.54

    Slight exaggeration -- the slowest is 1.85% slower than the fastest. And the 6 core is faster than the 12 core. If you think that makes sense, I have a bridge you might be interested in purchasing ;)
     
  9. h9826790 macrumors 604

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #10
    Ar... I see what you mean now. You are just talking about the 8G RAM setup, not all of them. From OWC's way to present the result. I assume this test is quite RAM demanding. In fact, when RAM size increase, performance increase. So, for the 8G RAM setup, the bottleneck is not at the CPU but the RAM size, of course the core count won't make any difference.

    However, if you check the 48G RAM setup, then the core count make the difference.

    Also, it seems the test is quite sensitive to the memory bandwidth as well. That's why the 96G result is better than the 128G result.

    For me, they are selling the RAM for this test. GPU and storage is not the main point, but they just try to sell it as well. And they don't even bother to sell their cMP CPU upgrade programme in this benchmark / advertisement.
     
  10. SoyCapitanSoyCapitan macrumors 68040

    SoyCapitanSoyCapitan

    Joined:
    Jul 4, 2015
    #11
    Just bear in mind the size of the files generated by these tests are about 10 times bigger than what nearly any professional Photoshop user would create. It's far from real world stuff.
     
  11. 996085 macrumors 6502

    Joined:
    Oct 5, 2015
    #12
    It tells me the ability to upgrade the cMP keeps it in the running.
     
  12. Arron Rouse macrumors regular

    Joined:
    Apr 28, 2010
    Location:
    Chichester, UK
    #13
    So a fully tricked out cMP with 12 cores at maximum speed and a PCIe SSD can't even manage half the performance of an i7 iMac and that's "in the running"? It's a good job you haven't spent the last few days telling everyone that the i7 iMac is crap. That would be really embarrassing. :rolleyes:
     
  13. 996085, Dec 21, 2015
    Last edited: Dec 21, 2015

    996085 macrumors 6502

    Joined:
    Oct 5, 2015
    #14
    The test doesn't appear to benefit from multiple cores as evidenced by the 4, 6, 8, and 12 core cMP scoring similar results to one another (looking at the 32GB benchmark scores). This means single thread performance is critical to this benchmark. Given the processors in the cMP are based on a six year old technology I'd say they hold their own fairly well. Hell the clock speed of the iMac is 15% higher than that of the highest cMP. Combine these two and it's no surprised the iMac scores better.

    As for the large file test I can't speak to the performance of the Accelsior E2 SSD and how it compares to Apple's Fusion drive (I assume 128GB of SSD is sufficient given the iMac scores appear OK with 128GB but slow down considerably when it comes to 24GB of SSD). In fact, and this is no surprise as it's the same test as the smaller file...just larger), it looks like even the large test is processor dependent because the difference between the 3.3GHz and 4GHz models is almost 60 seconds (almost 79% longer). Given the Fusion drive is the same I'd say this benchmark is highly dependent on the processor speed (as mentioned above). When you compare the 3.3GHz iMac benchmark to the slowest cMP benchmark you see the cMP takes slightly over 32 seconds (24%) longer than the iMac. Given the Mac Pro is based on six year old technology I'd say it's doing very well.

    With that all said you completely missed the point...how poorly would the cMP compete if it couldn't be upgraded? Imagine if it weren't using PCIe SSD? Or upgraded GPUs (not sure how the latter fits in as there is no baseline to compare it with)?

    EDIT: Interestingly when the cMP is configured with 128GB of RAM it comes within 11% of the iMacs 64GB (maximum for the iMac) large score benchmark. Not too shabby in my opinion.
     
  14. SoyCapitanSoyCapitan macrumors 68040

    SoyCapitanSoyCapitan

    Joined:
    Jul 4, 2015
    #15
    In the highly multi processor aware Photoshop benchmark thread the Skylake in the iMac matched the fastest 12 core cMP. I can vouch for it myself.

    Yes the cMP does well for an older computer but most six years old computers that age don't die off as quick as a six year old machine used to way back in the day when we had single core systems and operating systems that were advancing at a pace faster than the hardware they were running on.

    This is all set to change with the coming of VR next year. It's been a long wait and it's going to be a cricket game changer. The cMP will still be a great Mac for getting some upgrades in there but it won't be able to handle what's coming happening on the computer and entertainment scene in the next two years. VR won't just be for gaming. It will encompass almost all entertainment consumption - live concerts, movies, documentaries, social networking, even many jobs could very well be chuffed or created in virtual economies.

    You don't want an old CPU for what's going on. Just a few super efficient, high clock speed cores are more important than just chucking a dozen older slower cores at a problem.
     
  15. 996085 macrumors 6502

    Joined:
    Oct 5, 2015
    #16
    Which highly multi processor aware Photoshop benchmarks are you referring to? Yes, at some point the cMP will be too old to compete and it will need to be replaced. But given the ability to upgrade it it has lasted longer than if it lacked the ability to be upgraded. That speaks volumes to the ability to upgrade.
     
  16. Stacc macrumors 6502a

    Joined:
    Jun 22, 2005
    #17
    These tests only exist to sell memory. I am guessing very few workloads look like their benchmarks.
     
  17. Bytehoven macrumors regular

    Bytehoven

    Joined:
    Dec 1, 2015
    Location:
    Up Shellpot Creek
    #18
    Put a CUDA CORE GTX 680, 980, 980Ti, Titan X in those cMP rigs and run that Photoshop test again.
     
  18. SoyCapitanSoyCapitan macrumors 68040

    SoyCapitanSoyCapitan

    Joined:
    Jul 4, 2015
    #19
  19. 996085 macrumors 6502

    Joined:
    Oct 5, 2015
    #20
    I agree. It seems the test is well suited to increases in memory and with the memory manufacturers name plastered all over it well...that would be a reasonable assumption. However it doesn't change what I wrote...as long as one keeps in mind my comments are wrt this test.
     
  20. fastlanephil macrumors 6502a

    fastlanephil

    Joined:
    Nov 17, 2007
    #21
    Some of us use our cMac Pro for just audio work. The only graphics needs are maybe two or three displays. Gearslutz has a thread for track count benching your Mac with Logic but no one has posted results for the 4 ghz 5K iMac but the 3.5ghz 5K iMac track count looks to be about 110 tracks. A maxed out 3.46 12-core cMac Pro scores at about 210 tracks. A 2013 6-coreMac Pro scores at about 160 so the 2013 8-core Mac Pro would probably score at about 190 or so. It looks like the cMac Pro for now is still a pretty cost effective solution for audio/Logic. Master/slave set ups using VE Pro 5 are another very cost effective solution many audio users that work with instrument templates.
     
  21. 996085 macrumors 6502

    Joined:
    Oct 5, 2015
    #22
  22. SoyCapitanSoyCapitan macrumors 68040

    SoyCapitanSoyCapitan

    Joined:
    Jul 4, 2015
    #23
    You only need the most recent pages
     

Share This Page