1. Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Multi-Core Prime

Discussion in 'Overclocking & Cooling' started by ovidiutabla, Jan 29, 2013.

  1. ovidiutabla

    ovidiutabla

    Joined:
    Feb 21, 2008
    Messages:
    40 (0.02/day)
    Thanks Received:
    12
    i5 3330 @ 3200Mhz - 18/19 sec
    i7 3770k - 12/13 sec

    What is wrong?
  2. Aquinus

    Aquinus Resident Wat-man

    Joined:
    Jan 28, 2012
    Messages:
    5,892 (6.55/day)
    Thanks Received:
    1,890
    Location:
    Concord, NH
    That the 3820 at 4.5Ghz is gives the same scores a a 3.2Ghz i5 and that your 3770k is scoring as high as an Intel 6-core CPU. I find it even more strange that it doesn't scale with HT (at all). Which are usually signs of much needed software optimizations because I can tell you that there is a big gap in multi-threaded performance between the i5-3330 and the i7-3820 and your benchmark isn't show it. It is also showing perfect (over perfect,) performance scaling when you add more cores which is the sign of a software problem.

    Are you outputting the data your generating so we can confirm the validity of the output?

    Just one more thing. The status says that between 0 and 100 million that 5,761,456 primes where found. According to the prime number theorem shouldn't that number be more like 5,428,681?
    Last edited: Feb 7, 2013
  3. Melvis

    Melvis

    Joined:
    Mar 18, 2008
    Messages:
    3,543 (1.53/day)
    Thanks Received:
    510
    Location:
    Australia
    That is one crappy score with this X6 :ohwell: Going by the scores my X6 is 3.5x slower then a Q9400 :wtf:

    Attached Files:

  4. AphexDreamer

    AphexDreamer

    Joined:
    Jun 17, 2007
    Messages:
    7,078 (2.74/day)
    Thanks Received:
    912
    Location:
    C:\Program Files (x86)\Aphexdreamer\
    It doesn't like AMD's I think. My FX6100 did worse than quad as well.
  5. Melvis

    Melvis

    Joined:
    Mar 18, 2008
    Messages:
    3,543 (1.53/day)
    Thanks Received:
    510
    Location:
    Australia
    Agreed, but my FX 8350 did do a pretty good score to be honest. So I don't know :ohwell:
  6. ThE_MaD_ShOt

    ThE_MaD_ShOt

    Joined:
    Nov 29, 2011
    Messages:
    4,776 (4.99/day)
    Thanks Received:
    3,306
    Location:
    Hi! I'm from the Internet
    I too have to agree. I think it hates Amd. I think I should have faired better then a tad under 1 min.

    [​IMG]
    Crunching for Team TPU
  7. Novulux

    Joined:
    Dec 18, 2011
    Messages:
    150 (0.16/day)
    Thanks Received:
    30
    [​IMG]
  8. CrackerJack

    CrackerJack

    Joined:
    Dec 13, 2007
    Messages:
    2,702 (1.12/day)
    Thanks Received:
    448
    Location:
    East TN
    [​IMG]
  9. ovidiutabla

    ovidiutabla

    Joined:
    Feb 21, 2008
    Messages:
    40 (0.02/day)
    Thanks Received:
    12
    The benchmark is developed using and AMD FX 8350 CPU :) It doesn't hate AMD, just single thread performance on AMD is lower than Intel. I work now on a much more CPU intensive benchmark, optimized for perfect parallelism and multithreading. I will be back soon with news!
    AphexDreamer, rickss69 and HammerON say thanks.
  10. Aquinus

    Aquinus Resident Wat-man

    Joined:
    Jan 28, 2012
    Messages:
    5,892 (6.55/day)
    Thanks Received:
    1,890
    Location:
    Concord, NH
    Then AMD's CPU should be doing better because its not single threaded performance that is lacking. There is also no scaling with HT enabled. I wouldn't call that optimized. I'm still going to mention how the 950 scores lower than my 3820 despite faster clocks, a better ipc, and faster memory. In places where you would expect improvement, your benchmark offers none. Your benchmark says that an 8-core AMD CPU is half as fast as an i5 (heck even a core 2 quad) quad-core and it says that an i5 is just as fast as an i7. Everything I've been seeing is telling me that this is not optimized because then it would scale properly.
  11. rickss69

    Joined:
    Aug 23, 2009
    Messages:
    2,431 (1.36/day)
    Thanks Received:
    604
    Location:
    Rockvale TN (Not Australia)
    It should be clear by now that this is for data gathering purposes rather than any real benefit for the one running the "benchmark". ;)
  12. Aquinus

    Aquinus Resident Wat-man

    Joined:
    Jan 28, 2012
    Messages:
    5,892 (6.55/day)
    Thanks Received:
    1,890
    Location:
    Concord, NH
    Obviously, unless you really think a 950 is faster than a 3820 or if a C2Q is faster than a 8350. I haven't been able to see any consistant results, which concerns me with what is actually being calculated. The point is that stuff looks weird and we can't validate what is happening behind the scenes.
  13. ovidiutabla

    ovidiutabla

    Joined:
    Feb 21, 2008
    Messages:
    40 (0.02/day)
    Thanks Received:
    12
    OK, thanks all for testing this. The software is optimized for multicore but the algorithm is not that complex and may not reveal 100% accurate the differences between different CPU's.

    Having this experience and this goal, I developed a new benchmark, witch use a far more complex multithreaded algorithm. Multi Core PI.

    The old benchmark Multi Core Prime is no longer in development. I hope you will find this benchmark more useful and will reveal more accurate performance differences between platforms.

    Multi Core PI calculates PI decimals using Bailey–Borwein–Plouffe formula. The benchmark is using a multithreaded algorithm written in C++ and provide excellent parallelism. Multi Core PI is written in Visual C++ using MFC and Win32API.

    How it works

    A slider will help you set the decimals of PI, from 10.000 to 100.000. Default is 80.000. Just hit Run benchmark button to start benching your CPU.

    Submit to HWBOT

    First, press Take Screenshot button. A screenshot and a XML datafile will be created. Attentio! CPUZ must be running!
    Second, follow the link provided on the dialog and submit your datafile to HWBOT.

    Supported operating systems

    Microsoft Windows XP / Server 2003
    Microsoft Windows Vista / 7
    Microsoft Windows 8 / Server 2012

    His own thread

    http://www.techpowerup.com/forums/showthread.php?p=2841926#post2841926

    Download link

    http://www.pcgamingxtreme.ro/
  14. AphexDreamer

    AphexDreamer

    Joined:
    Jun 17, 2007
    Messages:
    7,078 (2.74/day)
    Thanks Received:
    912
    Location:
    C:\Program Files (x86)\Aphexdreamer\
    [​IMG]
  15. zsolt_93

    Joined:
    Jul 21, 2010
    Messages:
    242 (0.17/day)
    Thanks Received:
    59
    Location:
    Romania
    I am not sure if this test is any better. 15 Seconds difference and the fx has +2 cores and +1300MHz/core.

    Attached Files:

  16. ovidiutabla

    ovidiutabla

    Joined:
    Feb 21, 2008
    Messages:
    40 (0.02/day)
    Thanks Received:
    12
    Is 27% faster. The FX. The test is accurate. Remember, singlethreading performance on Intel is far superior than singlethreading performance on AMD. +2 cores at +1300Mhz don't do more than 27%.

    The FX 8350 requires about 4200MHZ with 8 cores to match i5 at 3200Mhz with 4 cores. Thread / Thread, Intel is about 50% faster. Do the math, how many cores + you have to supply to overcome this?

    So, I think this test is very accurate. Please test some more.

    [​IMG]
    Last edited: Feb 9, 2013
  17. AphexDreamer

    AphexDreamer

    Joined:
    Jun 17, 2007
    Messages:
    7,078 (2.74/day)
    Thanks Received:
    912
    Location:
    C:\Program Files (x86)\Aphexdreamer\
    Is it possible to make a beanchmark that were more FX optimized? To utilize the modules more efficiently?
  18. ovidiutabla

    ovidiutabla

    Joined:
    Feb 21, 2008
    Messages:
    40 (0.02/day)
    Thanks Received:
    12
    It is developed using an FX 8350... is optimized, it uses efficiently all the cores [check task manager].
    Last edited: Feb 9, 2013
  19. rickss69

    Joined:
    Aug 23, 2009
    Messages:
    2,431 (1.36/day)
    Thanks Received:
    604
    Location:
    Rockvale TN (Not Australia)
    My question is how is this benchmark going to help anyone make any kind of decision regarding their hardware? The only thing I have seen thus far is that raising cpu clock speed will result in a quicker conclusion of the test...any other tweaking has little effect.
  20. Aquinus

    Aquinus Resident Wat-man

    Joined:
    Jan 28, 2012
    Messages:
    5,892 (6.55/day)
    Thanks Received:
    1,890
    Location:
    Concord, NH
    Just because it "uses 100% cpu" doesn't mean it's efficiently using all the cores. Your benchmark "uses" all hyper-threading threads but still doesn't yield any significant performance improvement.
  21. ovidiutabla

    ovidiutabla

    Joined:
    Feb 21, 2008
    Messages:
    40 (0.02/day)
    Thanks Received:
    12
  22. Aquinus

    Aquinus Resident Wat-man

    Joined:
    Jan 28, 2012
    Messages:
    5,892 (6.55/day)
    Thanks Received:
    1,890
    Location:
    Concord, NH
    Saying it and doing it are two different things. You know, I'm going to bitterly fight this considering your benchmark says an i7 950 is faster than my i7 3820 and how HT doesn't scale. Those are signs that something is wrong. This still doesn't change the fact that 100% in the CPU task manager doesn't mean multi-core resources are being optimized (even more so since HT doesn't give any improvement.)

Currently Active Users Viewing This Thread: 1 (0 members and 1 guest)

Share This Page