Upgraded 3.46GHz 12-core 2010 Mac Pro faster than new 2.7GHz Mac Pro?

Discussion in 'Mac Pro' started by jkheit, Dec 26, 2013.

  1. jkheit Guest

    Joined:
    Aug 27, 2010
    #1
  2. stefmesman macrumors 6502

    Joined:
    Jun 25, 2010
    Location:
    Netherlands
    #2
    Those scores are from june. There was a later benchmark putting the nmp 12 core on roughly 33.000 on geekbench.
     
  3. mrsavage1 macrumors regular

    Joined:
    Feb 1, 2010
  4. jkheit thread starter Guest

    Joined:
    Aug 27, 2010
    #4
    Thanks for that. So even on integer. About 6% faster on floating point. About 19% faster on memory throughput.
     
  5. kennyman macrumors regular

    Joined:
    May 4, 2011
    #5
    I would say that the overall experience and system responsiveness, are better than the old MP, but it is not by a huge margin. You can judge by yourself from the information below (comparing top end nMP and old MP, both 12-Core machines).

    Integer Performance
    Barefeats nMP is 38050
    Old MP is 38126

    Floating Point Performance
    Barefeats nMP is 42041
    old MP is 39718
    Difference is - 2323

    Memory Performance
    Barefeats nMP is 5152
    old MP is 4705
    Difference is - 447

    I was just comparing it to the MacPro on Geekbench 3 online

    http://browser.primatelabs.com/geekbench3/204819

    So, CPU processing power and memory bandwidth are not that far from MP 4,1/5,1 dual setups.
     
  6. jkheit thread starter Guest

    Joined:
    Aug 27, 2010
    #6
    Thanks Kenny. Seems like it's pretty close. 5/10% difference. The video cards are where the difference comes in. It will be interesting to see if the FirePros will be something that can be added to the old Mac Pro. If so, these old souped up Mac Pros may be the machine to have for a bit longer, with the expansion abilities...
     
  7. peabo macrumors regular

    Joined:
    Feb 14, 2008
    #7
    I think that can already be done, but for the price of 2 FirePros you could have a new Mac Pro as they are $3500 each.
     
  8. kennyman macrumors regular

    Joined:
    May 4, 2011
    #8
    Question is; my friend - do you really need 2 fire pros? Do you want to invest on a GPU release in 2012 for USD 3500? Funny thing is now that FCP 10.1 is released, my RED files 4K editing works much better that before on my old Mac. Wonder why it was not released earlier on:confused:
     
  9. Donar macrumors 6502

    Donar

    Joined:
    Jul 12, 2008
    Location:
    Germany
    #9
    Marketing ;)

    "See how fast the new Mac Pro with FCPX 10.1 is, and look here how slow the old MacPro with FCPX 10.0x is."

    If Apple released FCPX 10.1 earlier, people would have realized that FCP X 10.1 is not only faster on the new MacPro, but also on the original MacPro (and even uses two GPU if available).

    Smart move.
     
  10. kennyman macrumors regular

    Joined:
    May 4, 2011
    #10
    hahaha....you are right! I can tell you honestly that I have seen a drastic change on how FCP handles raw red 4K files and even 5K files now with FCP 10.1 update.

    Previously I had to either render red files to ProRes or use the proxy (background rendering to get the chance to view the files). Now it is kind of magic, that same old hardware works.

    Another thing is NVIDIA can also work on their drivers to make sure MAC OS takes advantage of the GPU compute power. They are also behind here, not sure whether they are upset with Apple lol
     
  11. kennyman, Jan 1, 2014
    Last edited: Jan 3, 2014

    kennyman macrumors regular

    Joined:
    May 4, 2011
    #11
    Totally, we can still squeeze some more power out of the 2010/2012 MP for another 2-3 years. Now that the reviews for nMP 12-Core is out, I can confirmed that the raw CPU performance of dual X5690 is still more or less same with E5-2697 v2. In terms of CPU rendering and raw CPU power, we can still keep the old MP. My MP`s spec is below, let me know if you need any benchmark scores or any other info before upgrades.

    MP 2010
    2 Xeon X5690
    48 GB RAM
    1 SSD #on pcie card
    1 GTX 680

    I guess we will just have to wait and see what Maxwell GPUs bring to the table. Hopefully MVC will do his magic and we can have decent single cards with good performance under the hood in OSX.


    Updates, here are some benchmark scores for comparison

    Geekbench 3

    [​IMG]

    Geekbench 3 Floating Point

    [​IMG]

    Cinebench R11.5 with GTX 680

    [​IMG]

    Cinebench R15 with GTX 680

    [​IMG]

    Heaven Benchmark with GTX 680

    [​IMG]
     
  12. kennyman, Jan 3, 2014
    Last edited: Jan 3, 2014

    kennyman macrumors regular

    Joined:
    May 4, 2011
    #12
    Those who have GTX 780, will have higher scores in both Cinebench and Heaven.
     
  13. Upgrader macrumors regular

    Upgrader

    Joined:
    Nov 23, 2014
    #13
    Hey Kenny,

    Re your 48gb RAM, do you have 6x8gb sticks installed? I have the same dual X5690 chips as you with 32gb RAM (4x8gb) and my Geekbench score is around 31,000. Just trying to troubleshoot my score as it should be similar to yours. I’m guessing it’s the RAM config.

    Cheers.
     
  14. ssls6 macrumors 6502a

    ssls6

    Joined:
    Feb 7, 2013
    #14
    The heaven score is a bit fake since you're running mountain lion.
     
  15. Gav Mack macrumors 68020

    Gav Mack

    Joined:
    Jun 15, 2008
    Location:
    Sagittarius A*
    #15
    Not too obsessed with benchmarks, but all round I would imagine the gap bar single core has closed with the Sintech/nMP 2nd gen 1Tb blade option closing the gap on the disk i/o.
     
  16. kennyman, Dec 11, 2014
    Last edited: Dec 11, 2014

    kennyman macrumors regular

    Joined:
    May 4, 2011
  17. kennyman macrumors regular

    Joined:
    May 4, 2011
    #17
    @Upgrader
    Yes, I have Samsung DDR3 6 x 8GB ones. I believe you can achieve better scores by removing the memory sticks. But no need to troubleshoot scores lol

    It always fluctuates, just be happy with the scores. It also depends on OS, but it is all right to have between 31000 to 32000 with similar setup, all good.

    Scores on mavericks

    [​IMG]
     

    Attached Files:

    • 4.jpg
      4.jpg
      File size:
      90.6 KB
      Views:
      5,624
    • 2.jpg
      2.jpg
      File size:
      85.5 KB
      Views:
      208
  18. apphotography macrumors regular

    Joined:
    Nov 19, 2014
    #18
    The D700 Fire Pro cards don't compare well to the options you have for Mac Pro towers.
     
  19. kennyman macrumors regular

    Joined:
    May 4, 2011
    #19
    True, there are more GPU options available currently for the cMP than nMP. I just hope that the nMP 2015/2016 would have the same GPU connectors and same TDP as 2013 MP. i also use a nMP and would like to be able to replace the GPUs in the near future.
     
  20. 666sheep macrumors 68040

    666sheep

    Joined:
    Dec 7, 2009
    Location:
    Poland
    #20
    If you mean to drop 2015/2016 GPU into 2013 nMP – you won't be able to do it, even if the connector will remain the same. 2013 won't have EFI drivers and BIOS for other GPUs than D300/500/700 in its firmware.
     
  21. kennyman macrumors regular

    Joined:
    May 4, 2011
    #21
    Hmm...you are right, I would not be able to change the system definition from MP 6,1 to MP 7,1. That might be a an issue, I think I will just have to live with it.
     
  22. h9826790 macrumors 604

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #22
    It's hard to tell. Of course it will not be an easy task, however, how can a 4,1 use X5690? May be someone can figure out how to turn a 6,1 to a 7,1 later on.
     
  23. Tutor, Dec 12, 2014
    Last edited: Dec 13, 2014

    Tutor macrumors 65816

    Tutor

    Joined:
    Jun 25, 2009
    Location:
    Home of the Birmingham Civil Rights Institute
    #23
    Tock Tic principle controls - it's like clockwork, but the tock precedes the tick

    MP6,1 to MP7,1 ain't going to happen. Why could a 2006 MP (tock) be tricked into thinking that it is a 2007 MP (tick)? Same processor family. Why could a 2009 MP (Nehalem) (tock) be tricked into thinking that it is a 2010 (tick) or fake 2012 (Westmere) MP (just a processor w/in same family swap) ? Same processor family. If Apple had made a Sandy Bridge MP (tock) , then it could be tricked into thinking that it was an Ivy Bridge MP (tick) because they're part of the same Sandy Bridge processor family. See where this is leading - because the Ivy Bridge (tick) was the end of the Sandy Bridge family, those owners are stuck there. Now maybe a Haswell MP (if there is one) (tock) can be tricked into thinking that it's a Broadwell MP (if there is one) (tick) [see discussion here: 1) http://en.wikipedia.org/wiki/Broadwell_(microarchitecture) ], 2) http://en.wikipedia.org/wiki/Westmere_(microarchitecture) and 3) http://en.wikipedia.org/wiki/Mac_Pro .

    P.S. - A MP4,1 can use 5690s just by doing the 4,1 to 5,1 hack/trick [as you've already apparently done] and by installing the 5690s properly.
     
  24. Upgrader macrumors regular

    Upgrader

    Joined:
    Nov 23, 2014
    #24
    Ha, I guess I mean I'm trouble shooting my upgraded set up - I was concerned that maybe I'd over tightened a heat sink or something else was wrong with what I'd done, making the score lower. Just trying to get the absolute best from the unit. Now I've used it on a job though I'm not sure that 1000 missing points really matters. The machine flies although I could do with another 16gb RAM.
     
  25. zebity, Apr 29, 2015
    Last edited: May 17, 2015

    zebity macrumors member

    Joined:
    Aug 1, 2008
    Location:
    Australia
    #25
    Just did the 12 Core X5690 Upgrade

    Hi Mac'ers,

    Just did the dual X5690 3.46 12 Core upgrade on dual 3.06 X5675 12 Core Mid 2012 Mac Pro.

    I have been waiting for processor cost to come down, which they have.

    Geekbench 64 comparision:

    Single-Core: 2617 -> 2792 (6.7% increase)
    Multi-Core: 28451 ->31603 (11.1% increase)

    <Score>
    2792 Single-Core Score 31603 Multi-Core Score
    Section Description Single-Core Multi-Core
    Integer Processor integer performance 3066 37512
    Floating Point Processor floating point performance 2961 39508
    Memory Memory performance 1909 3975

    Geekbench 3.3.2 for Mac OS X x86 (64-bit)
    System Information
    MacPro5,1
    Operating System Mac OS X 10.10.3 (Build 14D136)
    Model MacPro5,1
    Model ID MacPro5,1
    Motherboard Apple Inc. Mac-F221BEC8
    Processor Intel Xeon X5690 @ 3.46 GHz
    2 Processors, 12 Cores, 24 Threads
    Processor ID GenuineIntel Family 6 Model 44 Stepping 2
    L1 Instruction Cache 32.0 KB x 6
    L1 Data Cache 32.0 KB x 6
    L2 Cache 256 KB x 6
    L3 Cache 12.0 MB
    Memory 64.0 GB 1333 MHz DDR3
    BIOS Apple Inc. MP51.88Z.007F.B03.1010071432
    </Score>

    This combined with fact that I get PCI-e slots which I have used for Areca RAID card and Nvidia GTX Titan, mean that I have enough cpu grunt and greater flexibilty with old Mac Pro than new Mac Pro, which provides "Thunderbolt" that is useful for ..... nothing much to my mind.

    It would be great if I could also get 10GBE, but this would require require removal of Areca RAID as you are limited only 2 16 Lane PCI-e cards.

    Instead I have bonded the two 1GBE ports together for greater network bandwidth.

    Zebity
    (Mac Pro 2012 with 2 x 3.46GZ + Areca + Nvidia GTX Titan)
     

Share This Page