Recent

% | $
Quotes you view appear here for quick access.

Intel Corporation Message Board

you are viewing a single comment's thread.

view the rest of the posts
  • wallisweaver wallisweaver Nov 20, 2011 5:32 PM Flag

    Fabrication Primer for Fanbois-Part I

    You would have us believe that as Intel produces better and better chips that ARM will not only continue to sell their old technology chips, but that they will continue to ramp up sales at the same rate.

    No sane person would buy your argument.

    If this explanation doesn't work for you, let me know and I'll make you a subliminal tape and you can play it while you sleep...

    SortNewest  |  Oldest  |  Most Replied Expand all replies
    • 'If this explanation doesn't work for you, let me know and I'll make you a subliminal tape and you can play it while you sleep'


      It might be better if you learned at least something about design and a little less about fabrication to acquire a processor semiconductor knowledge that was at least trailing reality rather than in fantasy world where it currently is ...


      At 65nm the top ARM processor was the single-core 1 GHz A8 which has a Dhrystone Performance of 2.0 DMIPS / MHz

      Performance (Total) 2,000 DMIPS

      http://www.arm.com/products/processors/cortex-a/cortex-a8.php?tab=Specifications+


      At 40nm the top ARM processor was the 2 GHz quad-core A9 which has a Dhrystone Performance of 2.5 DMIPS / MHz

      Performance (Total) 20,000 DMIPS

      http://www.arm.com/products/processors/cortex-a/cortex-a9.php?tab=Specifications


      At 28nm the top ARM processor will be the 2.5 GHz quad-core A15 (pairing of quads possible to make a cache coherent octo-core) which has a Dhrystone Performance of 3.5 DMIPS / MHz

      Performance (Total) 35,000 DMIPS (70,000 DMIPS for octo-core)

      http://www.arm.com/products/processors/cortex-a/cortex-a15.php?tab=Performance


      So from 65nm to 28nm ARM performance has gone up 17.5 times (35 for clusters).

      Repeating the same exercise for Intel using say the 65nm ~3 GHz Core 2 Quad and the 22nm ~4 GHz Ivy Bridge Octo-core will give you about a 5-6 times improvement (3 times less than ARM).

      Conclusion: Fabrication is a lower order effect on performance compared to design.

      Stop lecturing and pretending to be smarter than you really are, you are embarrassing yourself without even realizing it.

      • 3 Replies to marsavian
      • 'If this explanation doesn't work for you, let me know and I'll make you a subliminal tape and you can play it while you sleep'


        It might be better if you learned at least something about design and a little less about fabrication to acquire a processor semiconductor knowledge that was at least trailing reality rather than in fantasy world where it currently is ...


        At 65nm the top ARM processor was the single-core 1 GHz A8 which has a Dhrystone Performance of 2.0 DMIPS / MHz

        Performance (Total) 2,000 DMIPS

        http://www.arm.com/products/processors/c...


        At 40nm the top ARM processor was the 2 GHz quad-core A9 which has a Dhrystone Performance of 2.5 DMIPS / MHz

        Performance (Total) 20,000 DMIPS

        http://www.arm.com/products/processors/c...


        At 28nm the top ARM processor will be the 2.5 GHz quad-core A15 (pairing of quads possible to make a cache coherent octo-core) which has a Dhrystone Performance of 3.5 DMIPS / MHz

        Performance (Total) 35,000 DMIPS (70,000 DMIPS for octo-core)

        http://www.arm.com/products/processors/c...


        So from 65nm to 28nm ARM performance has gone up 17.5 times (35 for clusters).

        Repeating the same exercise for Intel using say the 65nm ~3 GHz Core 2 Quad and the 22nm ~4 GHz Ivy Bridge Octo-core will give you about a 5-6 times improvement (3 times less than ARM).

        Conclusion: Fabrication is a lower order effect on performance compared to design.

        Stop lecturing and pretending to be smarter than you really are, you are embarrassing yourself without even realizing it.


        ++++++++++++++++++++++++++++++++++++++++++++++++++++++

        Totally agree. Wally is a complete buffoon. What's worse than someone that's dumb, think they are smart, and is arrogant about it? That's why he has no friends and lives here, even in the wee hours of the morning. Intel will do fine, but Wally is a complete loser who knows about 1 tenth of what he think he knows.

      • The ole ARM doesn't need new fabrication argument.

        Otellini says you're wrong. It's says it's going to be about the physics and transistors and not the micro-architecture.

        Hmmmm. Should I go with you or Otellini knowing more about it?

        I'm going with Otellini instead of you, fanboi.

        And you're not going to impress anyone with that lousy 17.5x improvement. Especially since the calculation uses 28nm production which we know will only be 10 percent by the end of 2012.

        Intel's fabrication is going to grind ARM into powder. Tri-gate changed everything and it arrives in Q2...

      • I haven't seen anyone use Dhrystone this century. Even Reinhold says it is not good with todays CPU and compilers.

        Dhrystone probably runs pretty good on ARM with the small ARM CPU caches.

        Is there any reason ARM chose Dhrystone other than the obvious reason that it runs in a small cache?

 
INTC
34.46+0.45(+1.32%)May 29 4:00 PMEDT