0 Replies Latest reply on Jan 27, 2011 11:12 PM by deliezer

    Monster Math Machine



           I am a mathematician and numerical programmer who is trying

      to do my computations out of the house where my kids can't climb

      all over me.  For this I need to build the monster linux laptop, with

      fast disk access to spacious disk, large memory, large cache, etc.


           I am trying to decide between the i5 and the i7, which I have seen

      on Lenovo's website.  A Linux vendor I have used in the past offers only

      the i5.  Is there something known about the compatibility of the i7 with

      Linux, or do you suspect that the vendor is simply slow off the mark?


           I have noticed the hyper-threading feature.  I write

      in C++, but I do not currently know how to write multi-threaded code, and

      do not think my code would be a candidate for this sort of thing, EXCEPT

      that, when things come to fruition, I might want to run several instances of

      this.  Will this take advantage of the hyper-threading/multiple cores, or

      do I need to have multi-threaded code?


          The new burst technology also stood out on the product sheet, but it sounds

      like running a long heavy computation would not make demands on resources in short bursts,

      will the burst technology be useful for my type of problem?


          Other than that, it looks like the main difference between the i5 and the i7 is

      the size of the L3 cache.  It is hard to estimate how much cache I would wind up

      using, as it depends on the data file, but it is easy to guess it would be large, based

      on the number and size of structures I would be using in a reasonably intense set of

      example data. 


          Can you experts give me some guidance as to whether it is worth it for

      me to go with an i7?  Also, I've noticed that there are many different chips

      within the i7 and i5 families, each has a number associated with it, e.g. i7 920.  How can

      I tell what aspect of performance this number affects?  Is it solely clockspeed? 

      And I've noticed that some of these chips come with a name, e.g. Bloomfield, or Lynnfield. 

      Does this reflect some difference in architecture that favors one type of problem or another?


      I appreciate any help you can give me.


                                                                 David Eliezer