Does 2018 Mac mini 'cheat' Geekbench with high turbo boost

Discussion in 'Mac mini' started by BigBoy2018, Dec 11, 2018.

  1. BigBoy2018, Dec 11, 2018
    Last edited: Dec 11, 2018

    BigBoy2018 macrumors 6502

    BigBoy2018

    Joined:
    Oct 23, 2018
    #1
    So compare -
    1) 2018 Mac mini 3.0 base block i5, with turbo boost to 4.1
    2) 2017 iMac 3.8 base clock i5 with turbo boost to 4.2

    Both have almost the same SINGLE CORE Geekbench score, and I assume that's because in the time it takes to run the Geekbench tests, both machines can remain running at full turbo boost.

    However, lets say I've got a video render or export that takes 30+ minutes ... would the iMac then come out ahead since neither machine could maintain turbo boost that long, and the iMac actually has a higher base clock speed?

    According to the expert at Bare Feats, a lower base clock with a high turbo boost will eventually lag behind, but he doesn't actually test it.

    So ... does anyone know where there are extrended stress tests to determine what, if any, disadvantage there is to having a lower base clock, but a super high turbo boost speed? Thanks!



    Screen Shot 2018-12-11 at 7.05.50 AM.png
     

    Attached Files:

  2. pl1984 macrumors 68000

    Joined:
    Oct 31, 2017
    #2
    The answer is: It depends. IMO Geekbench is a lousy measure of a computers performance for anything other than running Geekbench. Specifically for the reason you state: It doesn't sufficiently stress the system and therefore its scores don't represent what one would see with real world applications that do. I have a 2010 Mac Pro and a 2012 rMBP and transcoding on the Mac Pro is considerably faster than the rMBP despite the latters higher GB score.

    It's my educated guess, and that's all this is since I have no real world experiences with either of these systems, is the iMac will outperform the Mini under such tasks which force the processor to operate near its base frequency (which is 26% faster on the iMac, but take a little off for IPC improvements in the Mini processor). What tasks would those be? Anything which causes the processor to operate near full utilization. Doing so increases heat which will have to be offset reduced clock speed.

    I've engaged in a number of discussions with new Mini owners attempting to get some Handbrake benchmark data to see what frequency the Mini tends to operate at under full CPU utilization. Unfortunately there's just not enough data to draw any conclusions. What I have seen is the Mini, even when utilizing all cores for lengthy periods of times, operates a little bit above the base frequency.
     
  3. Ploki macrumors 68040

    Joined:
    Jan 21, 2008
    #3
    Short answer : yes.
    Long answer: yes, but Mac Mini is relatively cool and can maintain higher turbo for a significant amount of time compared to i.e. MacBook Pro i9.

    Thats why desktops are usually better, because they have better cooling and higher base frequencies and can maintain higher CPU speed for longer.

    I see iMac and Mac Mini as a “cross” between mobile and desktop and performance reflects that.

    imac pro, mac pro have higher sustained values, because of better cooling and workstation chips.
    --- Post Merged, Dec 11, 2018 ---
    I saw 3,7 floating around for the i7. Which is higher above base than MacBook Pros.

    I’ll be sure to do a dev>null sustained operation test to see how well it maintains frequency under load when I receive mine!
     
  4. pl1984 macrumors 68000

    Joined:
    Oct 31, 2017
    #4
    I think I've seen posts to this effect too.
     
  5. Cheapassmac macrumors regular

    Joined:
    Nov 5, 2018
    #5

    Turbo boost doesn't meant sustained max ghz. It's a metric most useful in single threaded tasks. The cpu is limited by both thermal throttles, as well as power consumption. If the CPU is using all it's cores, it won't be running at max ghz on every core (and it wasn't designed to).

    Your mileage will vary on each cpu, but typically turbo boost CPUs under load will hover only slightly above base clock with laptop configs, moderately above base clock on apple desktop configs (this includes the 2018 mac mini, but not earlier), or halfway+ between turbo/base on a super cooled PC.
     
  6. ktcifone macrumors member

    Joined:
    Oct 30, 2009
    #6
    I have a i7/32/512, original Apple config. would like to compare your tests with my 2012 i7 custom config.
     
  7. pl1984, Dec 12, 2018
    Last edited: Dec 12, 2018

    pl1984 macrumors 68000

    Joined:
    Oct 31, 2017
    #7
    I'd be happy to run any test for comparison. However I can only do so with software I have or can download at no cost. Typically the benchmarking I've done is Geekbench (not a fan), Cinebench, and Handbrake using a common source file (typically a Samsung 4k demo) using a common preset. If that works for you let me know what you'd like me to test. I have the following systems:

    • 2013 nMP, hexa core @ 3.5GHz, 16GB, 256GB SSD
    • 2010 cMP, quad core @ 3.46GHz, 32GB, 1TB spinner
    • 2012 rMBP core i7, 16GB, 768GB SSD
    • Z440, quad core, 3.5GHz, 16GB, 1TB SSD
    I have other systems but I think these would be of the most interest to you.
     
  8. philipma1957 macrumors 603

    philipma1957

    Joined:
    Apr 13, 2010
    Location:
    Howell, New Jersey
    #8
    I would altered your title. It is wrong you are blaming Apple Mac mini when it is geekbench fault for a lousy test.

    Is the geekbench test flawed for testing the 2018 Mac mini properly?

    That is correct and true .

    Your title is wrong as it is not the Mac mini’s fault the geekbench testing favors it.

    I have attacked the 2018 mini a lot but as weird as it sounds your title is unfair to the 2018 Mac mini.
     
  9. BigBoy2018 thread starter macrumors 6502

    BigBoy2018

    Joined:
    Oct 23, 2018
    #9
    Agree with you. But ... does intel offer these prcessors with abnormally high turbo boost speeds with the intention of making them ‘look’ better to benchmarks like geekbench?
    If so, then would you agree the word ‘cheat’ still applies?
     
  10. pl1984 macrumors 68000

    Joined:
    Oct 31, 2017
    #10
    Doubtful. Intel defines Max Turbo Frequency as (from ARK for the i7-8700B):

    Max turbo frequency is the maximum single core frequency at which the processor is capable of operating using Intel® Turbo Boost Technology

    I think he's over reacting with your wording and therefore wouldn't put any thought into altering the title.
     
  11. brentsg macrumors 68040

    Joined:
    Oct 15, 2008
    #11
    No, they offer these processors to allow people to make the most of the power and thermal envelopes that they have available.

    Would you prefer that they just run at base frequency all the time, lest they run the risk of people making weird forum threads?
     
  12. BigBoy2018 thread starter macrumors 6502

    BigBoy2018

    Joined:
    Oct 23, 2018
    #12
    Ok, without being snarky, can you explain why some processors have turbo boost speeds of 1ghz or more, while others only have a TB of a few hundred mhz?
     
  13. pl1984 macrumors 68000

    Joined:
    Oct 31, 2017
    #13
    Build variances, cooling variances, and power variances to name three.
     
  14. philipma1957 macrumors 603

    philipma1957

    Joined:
    Apr 13, 2010
    Location:
    Howell, New Jersey
    #14
    if geekbench had a multicore test showing every core for one score
    if geekbench had a single core test showing one core for a second score.

    the thread would not have been written
    btw if you are running 1 core programs or low core programs the current test is okay.

    I use mini's for very lite duty I would not mind really high boost and only 2 or 4 cores as it works for me with my minis. In fact I just purchased a used mid range 2014 with 8gb ram and a 500 gb ssd it does the job I need it to do.
     
  15. brentsg, Dec 13, 2018
    Last edited: Dec 13, 2018

    brentsg macrumors 68040

    Joined:
    Oct 15, 2008
    #15
    I'm unclear if you're talking about an advertised boost of large vs small for different Intel parts, or are you referring to some people having better experience with the same part than others.

    If it's the former, Intel can essentially bin the same or similar parts based upon silicon quality. There's a process on a per-core bases that's happening in the background, where each core requests a certain voltage based upon workload at every specific moment. Then the motherboard is working to meet that requirement, along with the associated current based upon that workload. Lower quality silicon can be sold as less expensive parts with lower expected turbo performance, based upon higher power requirements needed to hit boost targets. Apple doesn't allow any sort of tweaking, but enthusiast motherboards offer very fine tuning of frequency, per-core voltage, and current (so thus package power). Enthusiast boards in the PC space also allow for a selection of motherboards with overbuilt VRMs that are designed to provide higher than usual TDP for extended or indefinite lengths of time.

    Meanwhile there's a whole different set of processes in play to govern the turbo performance/duration, and the best read would be here: https://www.anandtech.com/show/13544/why-intel-processors-draw-more-power-than-expected-tdp-turbo

    In the end we're left with a Mini (or whatever form factor) that uses the parameters described in the article above to govern turbo performance with an individual processor sample, chassis, cooling system, thermal paste application, ambient temperature, VRM quality, and workload.

    There are just a lot of variables in play, and like it or not.. silicon lottery is a real thing. Even within a specific Intel part, there's a lot of variability in regard to quality. In the enthusiast PC scene, people will pay obscene amounts of money for CPUs that are of exceptionally high quality. See: https://siliconlottery.com/

    This silicon lottery is especially important for overclocking, but you can think of Intel's turbo as a sort of automatic overclocking process.
     
  16. ondert macrumors regular

    ondert

    Joined:
    Aug 11, 2017
    Location:
    Italy
    #16
    I quit looking at geekbench scores since it doesn’t tell actual performance of your device but the capacity of the cpu. However, especially the current Mac line-up can’t keep or even refah that performance. Better to check cinebench scores.
     
  17. Spectrum macrumors 65816

    Spectrum

    Joined:
    Mar 23, 2005
    Location:
    Never quite sure
    #17
    In reality, all current processors are actually boosting from a idle base clock of ~1Ghz.
    The fact that an i7MBPro may say 2.3 with boost to 4.1 Ghz versus i7mini 3.2 with boost to 4.1 appears simply due to the stated TDP rating of each processor:
    The MBPro can maintain 2.3 (minimum) cross all cores within 45 Watts.
    The mac mini can maintain 3.2 (minimum) across all cores within 65 watts.
    Other than that, it seems to me that they don't really need to be different CPUs (even though they most likely are).

    Moreover, the misnomer is really saying that the i3 doesn't have a turbo boost. It certainly does compared to its idle speed, it is just that the 'turbo' is capped at 3.6 Ghz, which is at or below its 4-core 65 Watt TDP.

    The main thing that turbo permits, it seems, is the ability to temporarily exceed the TDP rating. Which, as explained eloquently above by brentsg, is dependent on a host of other parameters including, VRM output, cooling, ambient temp, workload, CPU quality, etc.
     
  18. ktcifone macrumors member

    Joined:
    Oct 30, 2009
    #18
    Tested this on my 2018 mini took me 3 minutes at HQ 1080/30 general preset.
    2012 custom hack mini i7 2.3/16GB/240GB SSD Raid 0 was about 8 minutes.
     
  19. BigBoy2018 thread starter macrumors 6502

    BigBoy2018

    Joined:
    Oct 23, 2018
    #19
    That is the crux of my question.

    Even though both have the same turbo boost speed, the processor with the higher base clock speed will ultimately be faster in real world performance, despite the fact that both may score almost identically in a geekbench test. Do we all agree on that?
     
  20. Ploki macrumors 68040

    Joined:
    Jan 21, 2008
    #20
    They don't tho, Mac Mini scores significantly higher on Geekbench (up top 28k on multicore, vs 23k of the i7 MB Pro).
     
  21. pl1984 macrumors 68000

    Joined:
    Oct 31, 2017
    #21
    What was the test you performed?
     

Share This Page