1

Machines involved:

  • Whitebox i5-2500k, 32gb, 4xSSd in raid 10
  • Lenovo TS440 with Xeon 1225v3, 32gb, and 4xSSD in raid 10

Both machines are configured the same software wise. Both machines have a quad core with 4 logical cores.

Both machines run 12 virtual machines.

The biggest thing I notice difference wise, is that with all 12 VMs at 100% usage on both machines, the 2500k shows the host is at 100% cpu usage, while the Lenovo shows it varies between 10 and 30% cpu usage.

Is this just a difference in the age/performance of the proc? Is there a reason that I may be looking past?

Improve this question
5
  • 2
    Although processors look largely the same on paper, you are missing a few important pieces, especially related to a virtual environment... Aside from possible bottlenecks in the mainboards, the Xeon processor has a faster memory bus, supports VT-d and TSX-NI, and a larger cache. The difference here seems extreme, but without knowing what is happening inside the processor, its hard to tell exactly.
    – acejavelin
    Commented Feb 3, 2016 at 22:11
  • 3
    Those CPUs are not slightly different, for your type of work, they are vastly different
    – Ramhound
    Commented Feb 3, 2016 at 22:42
  • I agree the CPU's are quite different in terms of age and market, but after looking into this, I am surprised the performance difference is so marked. The Passmark score is 6457 vs 7027, and a compare at ark.intel.com/compare/75461,52210 shows some differencs, but surprisingly little in terms of attributes - does the TSX-NI instruction set really make that much difference or is there something else (differences in configuration perhaps ?). [ I acknowledge that you can get up to a 40% increase on certain loads, but how likely is this in a real-world environment ??]
    – davidgo
    Commented Feb 4, 2016 at 9:53
  • @davidgo TSX-NI and VT-d, coupled with a larger internal cache and a faster memory bus, can make a significant real-world difference, especially in a virtualize environment with multiple VMs. But I will agree that the extreme difference is a little hard to explain.
    – acejavelin
    Commented Feb 4, 2016 at 13:31
  • There could be other underlying issues here not knowing more about the i5 machines hardware configuration. One thing that immediately comes to mind is many consumer mainboards for the i5 of that generation struggled with dual channel memory support when the memory configuration was maxed out , they would often fall back to single channel which could significantly decrease performance in a virtual environment. That's just the first thing that came to my mind, there are many other possible factors that cannot be taking into account with the information given.
    – acejavelin
    Commented Feb 4, 2016 at 13:41

0

Reset to default

You must log in to answer this question.