MBP 15" 2.9ghz gets beat in benchmarks by 2015 2.8ghz?

Discussion in 'MacBook Pro' started by RumorConsumer, Nov 16, 2016.

  1. RumorConsumer macrumors regular

    RumorConsumer

    Joined:
    Jun 16, 2016
    #1
    Hello - trying to interpret the results here - https://browser.primatelabs.com/mac-benchmarks

    You can see a series of benchmarks of 2015 and even 2014 MBPs that are faster than the 2016. Not appreciably, but higher numbers. Why does this happen?
     
  2. fs454 macrumors 65816

    Joined:
    Dec 7, 2007
    Location:
    Los Angeles / Boston
    #2
    It's true, and Intel is to blame. They have pretty much fully stagnated as of the past few years - nothing new to see on all fronts, just samey performance in smaller packages.
     
  3. johngwheeler, Nov 16, 2016
    Last edited: Nov 16, 2016

    johngwheeler macrumors 6502

    Joined:
    Dec 30, 2010
    Location:
    I come from a land down-under...
    #3
    [EDIT: sorry, I didn't notice the title of your thread which showed you are clearly comparing two very similar CPUs from different i7 releases (Haswell & SkyLake). I haven't gone into the individual GeekBench test results for each CPU, but this does seem to be a prime example of CPU stagnation from Intel. The overall 2016 performance may be better however, with faster RAM and SSD]

    Generally because some of the older CPUs are faster than the current ones! When comparing benchmarks, you need to look at the same family and closest clock speed for each generation of CPUs. The year-on-year improvements have been small for the last few years - mostly in the 5-10% region. Because of this, even a simple difference in clock-speed from, for example a 2014 3.1GHz CPU to a 2016 2.9GHz CPU can mean the older chip has faster benchmark values.

    If you consider widely different processor types, such as a dual-core 15W MacBook Air CPU to a quad-core 47W MBP 15 CPU, then the results are also going to be very different.

    The benchmarks are, IMO, becoming less useful as a way of determining actual system performance. Factors such as GPU, internal bus speeds, RAM speeds and especially SSD speeds are likely to have a bigger effect these days than the small differences in CPU speeds. The "growth areas" for laptops are now GPU, SSD speeds and interface speeds (TB3, USB 3.1 etc.). In the tablet/phone world, ARM based designs such as Apples A-series chips have also seen very healthy year-on-year performance improvements that are far larger than the Intel Core family CPUs.
     
  4. Qwe9203 macrumors member

    Joined:
    Nov 18, 2014
    #4
    That Haswell processor was the best. It was an advanced mobile processor. The ones in 2016 are regular mobile processors (cheaper).
     
  5. littlepud macrumors regular

    Joined:
    Sep 16, 2012
    #5
    Skylake is supposed to be 5-10% faster clock-for-clock. However, the Haswell CPUs in the 15" MBPs from 2013-2015 are all Crystalwell chips with Iris Pro, which includes 128 MB eDRAM. It's conceivable that the eDRAM (acting as both graphics memory and L4 cache) makes the Haswell faster than Skylake at some tasks, especially synthetic benchmarks.
     
  6. Wowereit macrumors 6502a

    Joined:
    Feb 1, 2016
    Location:
    Germany
    #6
    1st reason: the Iris Pro's eDRAM cache was also used by the CPU itself
    2nd reason: maximum turbo of the Haswell chip is at 4.0 GHz while Skylake's is at 3.8 GHz
     
  7. Fthree macrumors 6502a

    Joined:
    Mar 14, 2014
  8. RumorConsumer thread starter macrumors regular

    RumorConsumer

    Joined:
    Jun 16, 2016
    #8
    The cache and turbo boost reasons feel very plausible. Thank you!
     
  9. monkeydax macrumors 6502

    Joined:
    Nov 2, 2012
    #9
    What makes you say this?
     
  10. Kris.K macrumors newbie

    Kris.K

    Joined:
    Nov 17, 2016
    #10
    I registered to specifically post on this thread and add some more information to this. I really want to know why, I've just spent a significant amount of money to get a completely upgraded MacBook Pro 15" and if its slower than last years that's pretty ****.

    Geekbench 4 Benchmarks

    Single-CPU Scores
    MacBook Pro (15-inch Retina Mid 2014) - 4538
    MacBook Pro (15-inch Retina Mid 2015) - 4487
    MacBook Pro (15-inch Late 2016) - 4508

    Multi-CPU Scores
    MacBook Pro (15-inch Retina Mid 2014) - 14978
    MacBook Pro (15-inch Retina Mid 2015) - 14744
    MacBook Pro (15-inch Late 2016) - 14058

    Links:
    2014 https://browser.primatelabs.com/v4/cpu/1042571
    2015 https://browser.primatelabs.com/v4/cpu/942360
    2016 https://browser.primatelabs.com/v4/cpu/1047696

    Additional Information:
    I picked the scores by searching the laptop on Geekbench and picking the highest Mac OS. As there are higher benchmarks with these laptops running Windows instead.

    There have been mentions that it is the version of CL and GL that Mac OS runs?

    Also I spoke to an Apple rep on live chat and they said that it may take time for the Mac software to catch up and use the full potential of these latest CPUs. The Apple rep I chatted too also had purchased the same 15" Macbook and was surprised when I mentioned all this information but she had some good answers as to what she thinks and where these CPUs are positioned for the future.
     
  11. leman macrumors 604

    Joined:
    Oct 14, 2008
    #11
    The i7-4980HQ (Haswell) is pretty much the currently fastest mobile consumer CPU on the market. So much for Apple using "outdated" CPUs for two years.

    At the same time, Skylakes have larger cache and are more efficient, so there is a good chance that they will have a bit more thermal headroom. All in all, I'd prefer the Skylake CPU to the Haskell one. But if you already have a 2014-2015 MBP with the top CPU, and your work does not rely on GPU/storage perforamance that much, then the 2016 is not an upgrade for you.
     
  12. johngwheeler macrumors 6502

    Joined:
    Dec 30, 2010
    Location:
    I come from a land down-under...
    #12
    What on earth are you talking about? There is no indication of any such division between "advanced" and "regular" mobile processors in Intel's CPU roadmap. This sounds like some invented justification rather than a reasoned analysis based on fact.

    The Intel processor families are the result of their ongoing development, but the tendency is to incrementally improve the product so that they maintain their market position or improve it (although Intel already has 90% of the PC market, I understand). Yes, Intel may not have to try "too hard" in the face of lacklustre competition, but they are making improvements in their processors with each generation. These improvement may not be focused on clock-speed increases or even CPU benchmarks, but rather on power efficiency, lithography techniques, integrated GPUs and other CPU features.

    If you're going post such assertions you need to provide some evidence.
     
  13. leman macrumors 604

    Joined:
    Oct 14, 2008
    #13
    I think that they meant is to say that the Haskell CPUs were Iris Pro enabled models with the L4 cache. Most likely also binned higher. So yeah, maybe 'advanced' isn't the best term here, but 'premium' might fit.
     
  14. johngwheeler macrumors 6502

    Joined:
    Dec 30, 2010
    Location:
    I come from a land down-under...
    #14
    Agreed. There is more to the CPU than the benchmark results, and there is also more to the computer & its overall performance than the CPU.

    The question should be whether the computer as a whole represents an upgrade for you. If not, then there may be little reason to buy it.

    We all need to face the reality that computer CPUs are not going to get very much faster until there are some major technological advances, which could take many years. Other parts of the computer may well improve, including graphics, memory, cache, storage, display, interfaces & battery life. These are now the differentiators, not Geekbench results. Moore's law (which was actually about the total number of transistors rather than speed) is slowing down and may stagnate.

    The question should really be: is it fast enough for what I want to do?

    If not, then the answer may be to look to higher core counts (& the software that can use them effectively) and General Purpose GPUs for compute power. A laptop computer is limited in what it can achieve, and we may be close to this limit. There are lots of other types of computer available (including cloud services) if you need to do more.
    --- Post Merged, Nov 17, 2016 ---
    But there are SkyLake processors with Iris Pro with eDRAM - it's just that Apple chose not to use them!

    http://ark.intel.com/products/family/88392/6th-Generation-Intel-Core-i7-Processors#@Mobile

    I don't see any "downgrading" of SkyLake processor compared to Haswell. They're not much faster, but they do represent an evolution in the design (not least of which is the 22nm to 14nm lithography and decreased power usage).

    It does look like future quad-core CPUs will not include Iris Pro graphics, which may be why Apple chose just to keep basic iGPUs and add dGPUs to their MBP 15 line. If they don't do it for SkyLake, they will have to do it later unless Intel changes its plans.
     
  15. x-evil-x macrumors 68030

    x-evil-x

    Joined:
    Jul 13, 2008
    #15
    do you want geekbench numbers or real world? Could be something isn't optimized in geekbench for these specifics. but here watch this. The 2016 MacBook pro's are fast. On paper its hard to say real world they are.
     
  16. Wowereit macrumors 6502a

    Joined:
    Feb 1, 2016
    Location:
    Germany
    #16
    He is doing tasks where SSD and GPU affect the result aswell, so there is that.
    Throw in current model's SSD and GPU together with the old CPU and you MIGHT (not saying that you will, but there is a pretty good chance) get results which are even a little bit better.
     
  17. leman macrumors 604

    Joined:
    Oct 14, 2008
    #17
    Yeah, what we need to quantify the real difference between these CPUs is a real-world use case that only uses the CPU, but not the SSD or a GPU. Like some statistical computations. Sadly, even though there are a lot of 460 Pros out there, there still seems to be a lack of proper tests. I'd run some but our machines won't be here before I go on christmas holiday :/
     
  18. Kris.K macrumors newbie

    Kris.K

    Joined:
    Nov 17, 2016
    #18
    I've made a post summarizing what i've found. I've given credit to those that suggested the ideas too. Eg you Leman. :)
     
  19. Bryan Bowler macrumors 68040

    Joined:
    Sep 27, 2008
    #19
    That was the best review I've seen yet! Thanks for sharing.
     
  20. lympero macrumors 6502a

    lympero

    Joined:
    Sep 1, 2008
    Location:
    Arta, Greece
    #20
    Apple didn't use the available high end cpu's for 2016 mbp (6770hq, 6780hq). Instead they used the cheap - crap 6700 and 6820 etc.
    maybe they are saving them for next year refresh or they are just too cheap. Anyway you bought a pretty expensive mbp with crap CPU because Tim wants more money.
     
  21. SoyCapitanSoyCapitan macrumors 68040

    SoyCapitanSoyCapitan

    Joined:
    Jul 4, 2015
    #21
    Higher Turboboost

    Haswell isn't some wimp. It still has the most powerful desktop CPUs.
     
  22. Kris.K macrumors newbie

    Kris.K

    Joined:
    Nov 17, 2016
    #22
    It's all theories. Higher turboboost doesn't explain significantly slower ram.
     
  23. jackoatmon macrumors 6502a

    jackoatmon

    Joined:
    Sep 15, 2011
    #23
    CPUs don't really get faster anymore. They haven't for years now.

    They get smaller and cooler and more efficient.

    Make decisions with that dynamic in mind.
     
  24. RumorConsumer thread starter macrumors regular

    RumorConsumer

    Joined:
    Jun 16, 2016
    #24
    Wow that was great.
     
  25. leman macrumors 604

    Joined:
    Oct 14, 2008
    #25
    Of course its faster in GPU-assisted workflows — it has a faster GPU! But if I run a statistical simulation that does't use the GPU or the SSD, will it be faster? Maybe. We don't know. Because nobody has tested it yet AFAIK.
     

Share This Page