Hypothetical Geekbench score for new MP?

Discussion in 'Mac Pro' started by Mactrunk, Jun 12, 2013.

  1. Mactrunk macrumors regular

    Joined:
    May 12, 2005
    #1
    Just to calm the stormy waters....

    I'm wondering if some of the venerable high IQ members here could give a forecast of the Geekbench scores for the 6,1.

    This is obviously dependent on a huge variety of unknowns.
    I'm just curious about how much more powerful the CPU would be, compared to the current 12 core.
     
  2. theSeb macrumors 604

    theSeb

    Joined:
    Aug 10, 2010
    Location:
    Poole, England
    #2
    GB is not a good indicator of real world performance and trying to come up with scores for a CPU that has not been officially announced yet seems somewhat daft to me. Never mind the fact that we have no idea what the other CPU options will be.

    Based on that I am going to hypothesise that the score will be 30 bajillion.
     
  3. Erasmus macrumors 68030

    Erasmus

    Joined:
    Jun 22, 2006
    Location:
    Hiding from Omnius in Australia
    #3
    Based on the information that Apple provides on their website about the new Mac Pro, I think you'll find that the GeekBench score will be precisely 2.7183 times as high as the present top spec Mac Pro.

    At least that's what my highly complicated and in-depth exponential-based performance extrapolation techniques suggest, one generation to the next.
     
  4. Tutor, Jun 13, 2013
    Last edited: Jun 13, 2013

    Tutor macrumors 65816

    Tutor

    Joined:
    Jun 25, 2009
    Location:
    Home of the Birmingham Civil Rights Institute
    #4
    Many roads can lead to the same place.

    It'll be much, much, much lower. How about less than half of what you estimate. My dual E5-2680s (16 cores total) score about 38,000 [ See my Geekbench profile in the URL in my sig.]. That's four more cores for a higher TDP chip which Apple probably would never have used even if they had timely updated the Mac Pro for Sandy Bridge. Ivy Bridge chips are estimated to yield up to a 15% performance improvement over my Sandy Bridges [{5% to 15% increase in CPU performance when compared clock for clock} http://en.wikipedia.org/wiki/Ivy_Bridge_(microarchitecture) ]. So the math goes like this if we're talking about the 2680s V1 SB: 12/16 = .75; 38,000 x .75 = 28,500 (16 vs 12 cores); 28,500 x 1.15 = 32,775 (max Ivy Bridge improvement). That's no where near 2.7183 x 25,269 [ http://browser.primatelabs.com/mac-benchmarks - click on 64-bit Benchmarks ] = 68,688.72. My 4 E5-4650s, which are the exact equivalents of E5-2680s, score only 58,027 and that 32 actual cores or 64 with hyperthreading.

    But there's another way of looking at this to compare 2012 12-core with 2013 12-core: The average performance increase, according to IXBT Labs and Semi Accurate as well as many other benchmarking sites, at clock to clock is 11.3% compared to the Nehalem Generation, which includes Bloomfield, Clarkdale, and Lynnfield processors. [ http://en.wikipedia.org/wiki/Sandy_bridge ]. The average score for the top of the line 2012 Mac Pro (2nd Nehalem Generation) is 25,269, as shown above. 25,269 x 1.113 = 28,124.397 (Sandy Bridge diff); 28,124.397 x 1.15 = 32,343.06 (Ivy Bridge diff). Using either route leads to the same place: a Geekbench 2 score of between 32,000 to 33,000 for Apple's top of the line Mac Pro if (and it's a big if) those 12 cores are clocked at 3.06 GHz. If they're lower clocked, then the delta will be even lower. Thus, as I've said more times than I care to count, the key to the success of the 2013/2014 Mac Pro will be Apple's pricing it for what it truly is. Apple's smoke and mirrors cannot hide the fact that, if you use any benchmarks that we commonly use for CPU performance, performance delta of 2013/2014 Mac Pro over 2012 Mac Pro, clock per clock, will be, at best, for the top of the line 2013/2014 system 1.28 (= 1.113 x 1.15). Apple's bound by math and the laws of physics just as are its competitors, as are we.

    I love all of my math and science teachers. They all told me I'd do so in my later years. And they were 100% correct.


    Addendum: In my underclocking studies I performed on my WolfPacks, I found that both the Cinebench and Geekbench score were about 10% less that I expected when I used 2 6-core CPU's as opposed to one 6-core CPU. Thus, the benchmark score range that I suggested above should be widened to 32,000 to 36,000 to take into account that all 12 cores are on one CPU as opposed to 2 CPUs. But it remains to be seen whether those 12 cores will be clocked at 3.05 GHz and I think that is highly unlikely given Sandy Bridge's clockings and Ivy Bridge's projected slight speed bump over Sandy Bridge. I apologize for this late correction.
     
  5. itsmrjon macrumors regular

    Joined:
    Jun 11, 2011
    Location:
    Chicago
    #5
    I can give you a hypothetical LINPACK number.

    It will be around 255 GFlop/s +- 30 GFlop/s (not using OpenCL, depending on clock speed).

    Benchmarks like Geekbench are arbitrary. Try giving LINPACK a whirl, real problem, real solution.
    http://www.top500.org/project/linpack/
     
  6. derbothaus macrumors 601

    derbothaus

    Joined:
    Jul 17, 2010
    #6
    Tutor pretty much nailed it. Those are not magic procs that Apple will be using. Slight bumps and the users who have the closest current gen will be closest to the new geekbench score. The removal of the second CPU keeps the leaps to a minimum over the current gen. One is now faster than the 2 previous but it could have been faster.
     
  7. thekev macrumors 604

    thekev

    Joined:
    Aug 5, 2010
    #7
    Minimal gains on the x86 end. I'm not sure whether there are E5-1600s versions that go up to 12 cores. This is significant because E5-2400s are often clocked lower. If they approach 1600s clock speeds, they will be priced into the stratosphere. To me it would seem illogical to use one extremely expensive cpu over 2 less expensive ones if you're chasing core counts either way, but you never know with Apple. If they were going to use AMD, they should have been much clearer on pushing OpenCL to their developers. It has been a little quirky at times.
     
  8. Tutor, Jun 13, 2013
    Last edited: Jun 13, 2013

    Tutor macrumors 65816

    Tutor

    Joined:
    Jun 25, 2009
    Location:
    Home of the Birmingham Civil Rights Institute
    #8
    There's one factor that I forgot to take into account where the score of the new 12-core could be up to 10% higher than I suggest and that is the fact that those 12 cores are on one CPU. In my underclocking studies I performed on my WolfPacks, I found that both the Cinebench and Geekbench score were about 10% less that I expected when I used 2 6-core CPU's as opposed to one 6-core CPU. Thus, the benchmark score range that I suggested above should be widened to 32,000 to 36,000 to take into account that all 12 cores are on one CPU as opposed to 2 CPUs. But it remains to be seen whether those 12 cores will be clocked at 3.05 GHz and I think that is highly unlikely given Sandy Bridge's clockings and Ivy Bridge's projected slight speed bump over Sandy Bridge. I apologize for this late correction. I'll make the change above also.
     
  9. wallysb01 macrumors 65816

    Joined:
    Jun 30, 2011
    #9
    Intel is going to be stretching to put 12 cores on a single processor for Ivy Bridge E5s, so I wouldn't expect very high base clock rates. However, turboboosting may take low threaded work flows into the mid-upper 3's for GHz, just as the E5-2687W sees. So, just guessing, but the E5-2687W v2 might be 12 cores, a base clock rate of 2.8, but with 1-2 cores active it might be turboing to 3.8.

    The problem with this design though, might be the TDP. That E5-2687W is 150W. Ivy bridge will drop general TDPs, but squeezing core count and clock rate to the limit will keep it high. So, can this new Mac Pro go up to 150W? Or will intel offer a lower clock rate 12 core that Apple is targetting? Something more like 2GHz base and upto around 3.0GHz with turbo?

    I think its just pretty hard for people to predict what's going to happen in Intel's upper end 2600 v2s given that we are going to see 10 and 12 cores, when the 2600 v1 only go up to 8. More will of course be known in a few months once we start getting some leaks from Intel.
     
  10. Erasmus macrumors 68030

    Erasmus

    Joined:
    Jun 22, 2006
    Location:
    Hiding from Omnius in Australia
    #10
    Actually, my post above is a joke. A pretty bad and obscure one, but a joke none the less. Joke explanation follows:

    Everyone know that computer performance increases exponentially over time, i.e.. doubles in performance every few years. 2.7183 is the 'e' constant, as in e^x. Therefore, hilarious joke.

    That nobody noticed.

    Feel free to laugh now, though. :p
     
  11. deconstruct60 macrumors 604

    Joined:
    Mar 10, 2009
    #11
    There are no indications that there is any drop. Leaked Core i7 equivalents of what the 1600 v2 will be are the same old 130W TDP.


    http://www.cpu-world.com/news_2013/...idge-E_extreme_CPUs_to_launch_in_Q3_2013.html

    (the 1600's v1 never had the 150W old-ball the matching Core i7 one in the chart.)

    All indications are that for the 1600 series any process improvements in head they are taking in GHz increases and better to shift off the base rate over wider workloads. The expected core count max jumping from 10 to 12
    is highly indicative that going to take "more cores" over lowering TDP any with the 2600.

    What might happen is the 2620 does what the1620 does is keep core count the same and boost clock. ( stay at 6 and significantly move off relatively quite slow 2GHz base rate. )


    Probably not. The 12core is going to be a "lower" rate. But the Ivy Bridge power management isn't like Haswells. The 2680 has a dynamic range of 2.7-3.5 and 8 cores. Adding 4 more cores ( and associated cache ) while also adding 0.2 to the dynamic range is big leap for just a process shrink.
    (it isn't an optimized for that process tech.. )

    I'd expect more like 0.1 bumps on speeds and core counts all around by 2. ( 6's -> 8's , 8's -> 10's and one, maybe two "if you have to ask you can't afford them" special case 12's with lower than the 10's base rates. )
     
  12. TheEasterBunny macrumors 6502

    Joined:
    Jan 22, 2013
    Location:
    Delaware
    #12
    Regardless of the CPU improvements, it will feel much snappier because of it's storage device.
    This will not help GB2 scores, but it will help user feel. More cores do not give you that "zip" clock speeds and drive access do.
    "Feel" and "look" seem to be the direction Apple has been in for a few years now.
     
  13. B... macrumors 68000

    B...

    Joined:
    Mar 7, 2013
    #13
    No, I got it.
     
  14. Umbongo, Jun 15, 2013
    Last edited: Jun 15, 2013

    Umbongo macrumors 601

    Umbongo

    Joined:
    Sep 14, 2006
    Location:
    England
    #14
    No need to predict, it's been leaked.

    E5-2697 V2 - 12-cores, 2.7GHz, 130W TDP
    E5-2695 V2 - 12-cores, 2.4GHz, 115W TDP
    E5-2690 V2 - 10-cores, 3.0GHz, 130W TDP
    E5-2687W V2 - 10-cores, 3.4GHz, 150W TDP

    E5-1660 V2 - 6-cores, 3.6GHz, 130W TDP
    E5-1650 V2 - 6-cores, 3.4GHz, 130W TDP
    E5-1620 V2 - 4-cores, 3.7GHz, 130W TDP

    E5-2697V2, E5-2687W V2, E5-1650V2 and E5-1620 V2 if I were doing it. 8-core CPUs don't have high clockspeeds from the looks of it. The E5-2687W V2 with 4 more cores over the E5-1650 V2 is worth the extra $1,200 to some without sacrificing clockspeed. Obviously there is the 150W issue.
     
  15. wallysb01 macrumors 65816

    Joined:
    Jun 30, 2011
    #15
    Where did you pull that from? I just did a quick google search and didn't come up with a list containing core counts. But I think you're right. For mid-level pricing reasons, I think you should hit the 1650 and save the upper level for the 10/12 core options. If the Mac Pro can't handle 150W, I don't see how the E5-2690 v2 would be a bad choice. It allows for more distinction between the 10 core and the 12 core that way. Also, in a raw GHz counting way, the 2687W would be faster than the 2697 since 10*3.4 > 12*2.7.

    So my guess is that we will see the 2690 v2 and 2697 v2.
     
  16. Erasmus macrumors 68030

    Erasmus

    Joined:
    Jun 22, 2006
    Location:
    Hiding from Omnius in Australia
    #16
    Yay!!!
     
  17. Umbongo macrumors 601

    Umbongo

    Joined:
    Sep 14, 2006
    Location:
    England
    #17
    30mb cache = 12 cores, 25mb = 10 cores, 20mb = 8. The exception being the E5-2643 V2, which is either listed wrong or doesn't have all its cache disabled? It's replacing a 4-core, so will either be 4 or 6.
     
  18. Mactrunk thread starter macrumors regular

    Joined:
    May 12, 2005
    #18
    I'm relieved to see some discourse here.
    After my initial post, I thought I was an idiot for even asking the question.

    Just saw this...
    http://www.primatelabs.com/blog/
    And yes I know GB scores are not the ultimate test etc.
    That being said, I've found the GB info worked well to roughly gauge the power of the various MPs I've had over the years.

    The current 12 core 5,1 is working well for me.
    Nice to know that it is pretty close to the 6,1.

    I'm sure the new MP will be much better when it is finally released.

    That being said, I'm in no hurry to wait for the new MP.
     
  19. MacVidCards Suspended

    Joined:
    Nov 17, 2008
    Location:
    Hollywood, CA
    #19
    Main thread on front page stories.

    Good for a laugh
     
  20. Mactrunk thread starter macrumors regular

    Joined:
    May 12, 2005
    #20
    What does that mean?
    Please inform.
    Why are you laughing?
     
  21. Glen Quagmire macrumors 6502a

    Joined:
    Jan 6, 2006
    Location:
    UK
    #21
    Because he thinks the new Mac Pro is garbage and thinks everyone else should share his opinions.
     
  22. MacVidCards Suspended

    Joined:
    Nov 17, 2008
    Location:
    Hollywood, CA
    #22
    Move your eyes upward from this window and toward the left.

    Look for a button that says "Front Page"

    Click on it.
     
  23. mabaker macrumors 65816

    mabaker

    Joined:
    Jan 19, 2008
    #23
    The new score is 32bit only - 64bit results will be higher.
     
  24. Mactrunk thread starter macrumors regular

    Joined:
    May 12, 2005
  25. GermanyChris macrumors 601

    GermanyChris

    Joined:
    Jul 3, 2011
    Location:
    Here
    #25
    Here is what he means
     

    Attached Files:

Share This Page