locked
Xeon vs Core 2 Quad

    Question

  • My code basically does this.  It loads information from a local database, run some math test on it and stores the results for display at a later time.  The database is about 1gig in size.  This program is multithread with locks at appropriate places for the database reading and the data storing.  Here is my issue.  I am running 16 threads at once on both these computers.

    I am running a 2 x Xeon 3.2 Quad Core at work.  It takes about 25 mins to run. 
    I am running a Core 2 Quad 2.66 at home and it takes about 5 mins to run.

    Both are running Windows XP 32bit, 2Gig ram.

    So my question is, what is so different about those processors that would cause this difference and is there anything I can do to improve the speed on the Xeon?
    Wednesday, March 19, 2008 5:07 PM

All replies

  • Are the two system "exactly" the same with the exception of the processor?.  

     

    Since the database is local do they have the same quality of harddrives?  Same configuration (raid or not, etc)?

     

    Do they has the same background processes running (like virus checkers)?  My work computer has a lot more processes (mostly dealing with security) running in the background than any home system I had.

     

    You might want to profile your appllication to see if the "math" part is taking more of the process time verse the database storage and retrieval part.  If it is the math part then it seems the Core 2 Duo is better at that.  If it is the data access portion I would say you have some other hardware issue.

    Wednesday, March 19, 2008 5:25 PM
  • Processors are the same except the xeon is clocked hire and probably has a lot more cache.
    Thursday, March 20, 2008 3:37 AM

  • If the threads have to share data between them, the Xeon box needs to move data between the processors over the FSB i believe. Not sure how a C2Q does that, probably more efficient.

    You should try it on a AMD box with hypertransport to see if that brings it down even further.

    GJ
    Thursday, March 20, 2008 12:43 PM
  •  Gert-Jan van der Kamp wrote:

    If the threads have to share data between them, the Xeon box needs to move data between the processors over the FSB i believe. Not sure how a C2Q does that, probably more efficient.

    You should try it on a AMD box with hypertransport to see if that brings it down even further.

    GJ


    Didnt think about that.  There is lots of shared data and having to move the data back and forth would seem to cause issues.  I will have to explore that some more.  I've just started profiling the code to see the issue.

    Thanks for all your help.
    Thursday, March 20, 2008 4:54 PM
  •  Gert-Jan van der Kamp wrote:

    If the threads have to share data between them, the Xeon box needs to move data between the processors over the FSB i believe. Not sure how a C2Q does that, probably more efficient.

    You should try it on a AMD box with hypertransport to see if that brings it down even further.

    GJ


    No, all Core2 architecture processors communicate through the FSB. Technology that allows on-die communication can only be found in AMD processors. This technology is called Hyper Transeport. The upcoming Nahalem line of processors from Intel will have Intel's version of AMD's HT link technology.
    Friday, March 21, 2008 6:22 AM
  • any progress with your research? I am on the verge of buying a new server and i am opting among Xeon Quad or Core2 Quad. Your information could be very useful to meSmile
    Thursday, May 15, 2008 10:41 PM
  • Yes I did.  After I ran it out of debug mode, both seemed to run about the same speed for the math portion.  The main difference between the two computers seemed to be the hard drive speed.

    Sorry this doesnt help, but they both seem comperable for my tasks.
    Thursday, May 15, 2008 11:04 PM