1 Reply Latest reply on Mar 18, 2009 12:17 PM by Christopher Peters

    Two 2 GHz processors or ONE 2.3 GHz processor?


      Hello Guys,


      Let me thank yo ufor all your help so far! This community has prooved very helpful for me to decide my purchase.


      Now that I have decided to buy Xeon over Core Quad, I need to decide the processor and the Board.


      The cost of 2.3 GHz procesor is 1.5 times the cost of a 2.0 GHz processor. 2.5 GHz is twice the price of 2.0 GHz.


      So, if I install a 2.3 or 2.5 GHz processor, i dont think I'll ever invest this big on adding another processor to the server hence it will always remain a one processor machine.


      I need to install Windows Server 2008 with Hyper-V on it with 3 Virtual Win 2k3 Servers, 1 Vista and 1 Windows-7 client. One of the virtual 2k3 Server will also host Exchange 2007 on it.THIS IS A LAB ENVIROMENT AND NOT A PRODUCTION SERVER.


      So, can you please suggest which combination should I go for? TWO 2.0 GHz processors or ONE 2.3 GHz processor? Which one of them will perform better?


      With Regards,

      Amit Saxena

        • 1. Re: Two 2 GHz processors or ONE 2.3 GHz processor?
          Christopher Peters



          I just checked Vmware's website for 1 socket (1S) performance on virtualiztion workload and could not find any publications on 1S - only 2S and 4S or higher.


          So here is some data on integer perfromance (SPECint_rate 2006) where i have some downbin data that i can offer as a references to help you


          two 3.0GHz Xeon (Xeon X5460):  score 114

          two 2.3GHz Xeon (Xeon E5410):  score 94.4 (18% lower than X5460)

          two 2.0GHz Xeon (Xeon E5405): score 86.1 (25% lower than X5460)


          We have data that shows the benefit of adding a 2nd CPU to the same configuration server with SPECint_rate 2006.

          Adding a 2nd X5460 to a platform running one X5460 is a 75% performance gain.  So assuming that a E5410 and E5405 would see similar improvements (this is my personal extrapolation and not based on any measurements or publications available today.


          one 3.0GHz Xeon (Xeon x5460):       (personal estimate would be 65 - extrapolating that 114 is 75% higher than this number)

          one 2.3GHz Xeon (Xeon E5410):      (personal estimate would be 53 - 18% below above 65 estimate (frequency scaling of )

                                                               (personal estimate would be 54 - extrapolating that the 94.4 number


          so i think in summary, using these estimates and SPECintrate2006

          two 2.0GHz ==> 86.1 (measured data)

          one 2.3GHz ==> 53-54 (pesonal estimate - logic above)


          I suspect that the details you see will be different because your workload is not SPECint but this is the best proxy i can give you to help you make your decision. one final thought, since most users are building production virtualization environment with more resources (CPU, memory, I/O), i think that a 2S platform may be a better approach to model in a test environment.


          Hope this helps.