Announcement

Collapse
No announcement yet.

Understanding CPU Mark

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Understanding CPU Mark

    Hi there, I have a general question on understanding how the "Average CPU Mark" numbers work in a MULTI-SOCKET and MULTI-CORE system (using for server virtualization), and how those numbers are used to compare TOTAL capacity.

    If I have an older system (HP BL680) with the following:
    [Quad CPU] Intel Xeon E7450 @ 2.40GHz
    (quad socket, 6-core), and the Average CPU Mark is 11080

    I have a newer system (Dell 720) with the following
    [Dual CPU] Intel Xeon E5-2680 v2 @ 2.80GHz
    (dual socket, 10-core), and the Average CPU Mark is 22940

    Then...
    Is my system total capacity based on the CPU Mark itself:
    "total system/old" =11080
    "total system/new" =22944
    or
    Based on SOCKET count:
    "total system/old" with 4 x 11080=44320
    "total system/new" with 2 x 22944=45888
    or
    is it based on CORE count:
    "total system/old" with 24 x 11080=265920
    "total system/new" with 20 x 22944=458880
    or
    Something else in combination?

    Thanks.

  • #2
    I don't think there is any standard metric known as "system total capacity".

    You need to ask yourself, if there is capacity, then what type of capacity is it?
    Capacity to store files on the hard drive?
    Capacity to serve up web pages?
    Capacity to encode video?

    If you are asking about capacity to run multiple virtual machines (VMs) then it really depends on how big each VM is and the processing that needs to be done in each VM.

    The CPU benchmark uses all Cores and CPUs that are available. There is no need to multiply the result with anything else.

    Comment


    • #3
      Sorry for the confusion, but I was trying to gauge the "old" vs. "new" systems for their total CPU capacity. There are many other variables on "total" capacity for the mem, NICs, HBAs, i/o, etc.

      Thanks for the information that the # is using all available socket/core. Does that include hyperthreading also?

      I just wanted to be able to gain a basic baseline insight that "cpu:z" has "x" times more capacity than "cpu:y" as part of a general sizing exercise I am starting for moving a whole bunch of VM to new hardware as part of an upgrade project.

      Comment


      • #4
        Yes, Hyperthreading is used if it is available.

        Comment

        Working...
        X