[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums

XG Beta 2.00.168 Release - Improved speed.

Posted By: eXtreme Gammon
Date: Tuesday, 14 June 2011, at 11:03 p.m.

In Response To: XG Beta 2.00.168 Release - Improved speed. (Matt Cohn-Geier)

The speed improvement is due to 2 facts:

  • 1. Introduction of the pruning Neural Network
  • 2. new multi-threading system to do low ply match evaluation so the CPU can be maxed out.

    On a match analyze 1 and 2 result in that 100%+ improvement. For RO only 1 has an effect.

    To follow up on Micheal and Miran response. A 2500K is not hyper threaded. I've made statistics on the effect of hyper threading: here are my finding:

    For N-ply evaluation (N>=4) speed due to multi core can be model as Speed=1+(NCore-1)*0.81+(number of core with hyper threading)*0.29.

  • For a i7 (4 core HT) speed=1+3*0.81+4*0.29=4.59 the speed on one core
  • For a core i5 (4 core no HT) speed=1+3*0.81=3.43 the speed on one core

    For RO, there is less conflict in the threading and speed=1+(NCore-1)*0.92+(number of core with hyper threading)*0.34

    Processor speed used to be easy to take into account: twice the GHz, twice the speed! but as processors are nowadays are using Turbo that raise the CPU clock until it's too hot and this make it quite difficult to take into account.

    For instance OCing a sandy bridge to 4.9 GHz (compare to the stock 3.4 GHz) will not yield a 4.9/3.4 improvement but closer to 4.9/3.9 but this depends on your cooling system.

    • Messages In This Thread

     

    Post Response
    Your Name:
    Your E-Mail Address:
    Subject:
    Message:

    If necessary, enter your password below:

    Password:
     

     

    [ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

    BGonline.org Forums is maintained by Stick with WebBBS 5.12.