Why is the 2016 Macbook Pro comparable to the nMP?

Discussion in 'Mac Pro' started by slughead, Dec 10, 2016.

  1. slughead, Dec 10, 2016
    Last edited: Dec 10, 2016

    slughead macrumors 68040

    slughead

    Joined:
    Apr 28, 2004
    #1
    On the face of it, the 2016 Macbook Pro doesn't have a lot of performance at all. The Radeon 450 has around 1 TFLOPS and the 460 is around 1.9 TFLOP. As I pointed out in this meme, it's basically the same power as Radeons released while George W bush was president.

    The base model of nMP released waaay back in 2013 has the D300 and a quadcore xeon. The D300 has over 2 TFLOPS and there are TWO of them. Already this is twice the processing power as the best rMBP. The CPU scores may be about the same but the GPU is absurd.

    So why then is the Macbook Pro holding a candle to this machine ?

    In some cases, the 'late 2016' Quad-Core MacBook Pro can be as fast as a Quad-Core Mac Pro cylinder. The MacBook Pro was beaten by the iMac 5K in all cases.
    Does that mean you won't ever need a Mac desktop? No. There are pro apps like FCPX and DaVinci Resolve that run much faster on a desktop.

    This is not to say that a $1500 PC with a GTX 1080 (NINE FREAKING TFLOPS) wouldn't set the competition ablaze if Apple actually supported it, but the question remains: Is the reason these machines are in the same ball-park a software issue or is this just magic?
     
  2. thefredelement macrumors 65816

    thefredelement

    Joined:
    Apr 10, 2012
    Location:
    New York
    #2
    I think a lot of it is IO throughput. Once things are loaded the nMP is faster, but for smaller tasks, disk and bus I/O on newer tech may end up faster.
     
  3. BornAgainMac macrumors 603

    BornAgainMac

    Joined:
    Feb 4, 2004
    Location:
    Florida Resident
    #3
    Apple makes the case that you can't combine touch screens with laptops. That a laptop and tablet should be separate devices. They claimed they tested it and it doesn't work.

    Now Apple is marketing the laptop as a desktop replacement and laptop. If they had tested that theory, they would realize the laptop components will overheat and throttle down, the battery will not last so it will require it to be plugged in, they have to use slower components, cooling is more difficult, you are limited with memory, and you can only have 1 GPU and 1 CPU, and you will be limited by the number of cores. Desktops can be made upgradable and fixable by the end user, Desktops are just able to handle heavy loads faster and longer than a laptop.
     
  4. Ph.D., Dec 10, 2016
    Last edited: Dec 10, 2016

    Ph.D. macrumors 6502

    Joined:
    Jul 8, 2014
    #4
    The premise is stretched too thin to be taken all that seriously (along with a little too much bait), and so I really shouldn't wade into this debate, but here are a few points:

    First, I just bought a base model no-touch 13" MBP. It totally blows away my heavily-upgraded MP 3,1 in day-to-day use, and it does that while using around 1/20th of the power and volume. No, I wouldn't do heavy number crunching on the MBP, but for everything else it feels so much faster that it's ridiculous. Even the base model has twice the single-thread speed as a 3,1, but it feels much faster than that due to all the other progress such as in memory and flash speed. No-doubt a high-end MBP would feel much faster still.

    Apple is pushing several boundaries that do make sense. One is focus and efficiency. The new macbook pros and the cylinder (at least back in 2013) push hard to achieve high levels of optimization and efficiency. Another is flash speed, which even on the base MBP model is twice as fast as in the cylinder. Another is IO throughput (TB 3). Another is over-all balance and refinement - the new machines have been honed to a level of engineering refinement that almost only Apple can achieve.

    The net result is that, for typical use cases, a well-optioned MBP can easily do better for most people than a cMP. The cylinder is a different beast, being an early, slightly flawed, now-old attempt to move traditional desktops in the same high-efficiency direction as the MBPs. It's less immediately responsive than some newer computers (e.g. high-end iMac) but is able to crunch indefinitely without serious thermal issues - and that's nearly its only intended use case.
     
  5. slughead thread starter macrumors 68040

    slughead

    Joined:
    Apr 28, 2004
    #5
    The question was more about the reason for a slower computer performing faster, and I'll point out that your 3,1 was also released while GW bush was president (2008) :)

    I was wondering if there is some sort of software tomfoolery or some other explanation.
    --- Post Merged, Dec 10, 2016 ---
    Can you link to a source on this? that's so crazy. Many many windows users are now hooked on touch screens for their laptops. I think they just don't want to flub up the interface in OS X to make it touch friendly (huge backlash when MS did it initially).

    I agree though, they're trying to stuff their pros into overpriced laptops for their desktop work. I have a feeling most "pros" that could make do with a rMBP don't actually need much processing power or GPU for their workflow (and therefore would do great on a cheaper machine)... either that or they're just enthusiasts with gobs of money. Nothing wrong with that.
     
  6. pat500000 Suspended

    pat500000

    Joined:
    Jun 3, 2015
    #6
    It doesn't seem like Apple care how they design their products as long as they sell it.
     
  7. antonis macrumors 68000

    antonis

    Joined:
    Jun 10, 2011
    #7
    All computers are comparable to each other. The rest is just marketing. Regarding the specific comparison, though, I think that 2015 MBP is closer to 2013 MP or anything else with a "pro" label whatsoever.
     
  8. filmak macrumors 6502a

    Joined:
    Jun 21, 2012
    Location:
    between earth and heaven
    #8
    The MBP is more advanced... in Apple terms... everything now is soldered or glued.:)
    A perfection...

    So, everybody stop complaining for your nMPs, after 3 years they have, a new feature, the label of Apple's most upgradable computer.

    Unbelievable...:(
     
  9. Ph.D., Dec 11, 2016
    Last edited: Dec 11, 2016

    Ph.D. macrumors 6502

    Joined:
    Jul 8, 2014
    #9
    I don't think that's true at all. They care a great deal about design decisions, and their products are indeed designed to extraordinary standards of engineering. It's just that some or many people feel trapped by Apple's limited options and complain that their designs are not what they want.
    --- Post Merged, Dec 11, 2016 ---
    No. One could argue that iPhones perform better on lesser hardware than Android due to the fact that iOS is so tightly integrated with Apples' hardware, but Mac OS is another animal that can't manage as tight an integration (after all, you can hack Mac OS onto average PC's), and hence there's no magical software voodoo involved.

    My personal comparison of my MBP to my 3,1 was to point out the gains in speed and especially efficiency that have been achieved over the years by Apple, Intel, Samsung, etc., and hence why a new MBP can now be competitive for many tasks with the now old nMP. No magic, just good engineering.
     
  10. tomvos macrumors 6502

    tomvos

    Joined:
    Jul 7, 2005
    Location:
    In the Nexus.
    #10
    As other have already said, comparing 2013 hardware to 2016 hardware will yield such a result. Especially when comparing the high end 2016 mobile solution to the 2013 low end desktop solution. Not much of a surprise here.

    Besides this, comparing benchmark will give you an answer how well a system performs in a selected benchmark. Nothing else. It does not tell you how good a system runs in a real world usage scenario. That's the inherent problem of benchmarking. If you design a benchmark as close to real use cases as possible, you'd have to include all metrics which are relevant for the task simulated in the benchmark. Such a benchmark would be honest, but you'd have to mix CPU-, GPU-, IO-, Memory- and other performance factors.
    The problem is, such a benchmark would lead to realistic results, but you might not be able to tell if the better result is based on IO- or CPU- or GPU-performance. Because these things are connected in real life. So your 2016 MBP may hold a candle to a 2013 MP (or a upgraded 2010 MP) in benchmarks. Fine. Have fun running benchmark. But real life performance is complex and to be good at it, you have to avoid bottlenecks.

    That's the hardware story. Next up, software side of things. The 2 TFLOPS are peak performance. Have a look at wikipedia how peak performance is calculated. To sum it up, this is marketing. In real usage your CPU and GPU spend a lot of time just idling because they wait for some other task to signal their semaphores. Recent advances in GPU have lead to better usage of GPUs ... but there are still lots of usage scenarios which could be compared to your daily ride to work: Sure, this Porsche is faster than your Toyota. But than, in the daily traffic you don't reach your work faster by driving a Porsche. The same is true for GPUs and Multicore CPUs. If you are blocked by other things or stop signals, it does not matter how fast you could go ... the only thing that matters if how fast you actually go.

    You can see lots of these situations in the benchmarks at barefeats. Whenever systems of quite different capabilities seem to cross the finish line nearly at the same time, you can assume that the benchmark task does not tell you anything about the supposed measurement at all.
     
  11. BornAgainMac macrumors 603

    BornAgainMac

    Joined:
    Feb 4, 2004
    Location:
    Florida Resident
    #11

    http://www.macworld.com/article/313...hief-jony-ive-rules-out-touchscreen-macs.html
    and
    https://www.cnet.com/news/touch-screen-mac-unlikely-says-apples-federighi/
     
  12. slughead thread starter macrumors 68040

    slughead

    Joined:
    Apr 28, 2004
    #12
  13. filmak macrumors 6502a

    Joined:
    Jun 21, 2012
    Location:
    between earth and heaven
    #13
    If we find it a horrible idea then we can be sure they 're working on it...:)
     
  14. theluggage macrumors 68030

    Joined:
    Jul 29, 2011
    #14
    ...because,

    (a) the nMP's raw performance was always entirely contingent on running OpenCL software that could fully exploit the dual GPU for processing. If I recall correctly, even at launch, the i7 iMac of the day matched it on some benchmarks. Its other potential strengths are the i/o throughput and extra stability of ECC RAM that comes with the (expensive) Xeon processor - which is important on those 24-hour render jobs but won't show up on <60-second Photoshop benchmarks that barely warm up the processor. The nMP was never a great "power user"'s general purpose machine - it is a specialist video/3D workstation that never made sense unless you were running an OpenCL-heavy workflow (see the FCPX, DaVinci Resolve benchmarks) that would max-out the hardware and I/O for hours.

    (b) As the article you quoted states), the new flash storage on the 2016 MBP runs rings around the 3-year-old tech in the nMP - which, as well as making the 2016 MBP thrash everything in the disc access tests, could tip the balance on any other benchmark that doesn't fully exploit the nMP's GPUs.

    Anybody here think that the lack of nMP updates for 3 years is not a problem? Even if there haven't been any compelling CPU/GPU upgrades, the SSD clearly needs some love.
     
  15. pmau macrumors 65816

    Joined:
    Nov 9, 2010
    #15
    Also, the chipset had trange PCIe lane mappings for the ports.
    They should just update the chipet, provide only 4 full TB3 port and the fastest SSD and RAM
    If then there would be usable GPU drivers, not only chips, I would even consider buying one.
    ... but the pricing might again be delusional.
     

Share This Page