locked
Intel turbo boost on VM? RRS feed

  • Question

  • Hi all,

    Currently we are testing regarding CPU performance on a VM vs. physical host. We've noticed that on a VM our software used for testing is 10% slower then on a physical host. Then we noticed that on a VM the core frequency is set at the base frequency of our Intel processors (in task manager, but also in other tools for checking CPU frequency). On the physical host the frequency is higher because of the Intel turbo boost being applied. 

    We've also tested disabling the turbo boost on the physical host. Then the difference in CPU performance between the VM and the physical host is only 3%. 

    Is it as designed that intel turbo boost isn't working on a VM? Or how can we pass the intel turbo boost onto the VM so that it is also being applied there? 

    Kind Regards,
    Joël

     

    Friday, August 10, 2018 7:49 AM

Answers

  • The max turbo speed depends on the number of cores in use, CPU will never turbo to max speed width all cores in use.

    Example

    If all cores are running at 100%, the turbo won't boost to max freq, this usually only happens with a single core at 100%, i.e., using a single threaded app, that's how it's designed.

    VMs rarely, if ever report CPU speed properly, even Hyper-V doesn't do it properly with Windows hosts and clients.

    Best regards,
    Leon


    Blog: https://thesystemcenterblog.com LinkedIn:

    Friday, August 10, 2018 9:23 AM

All replies

  • Hello,

    Unlike physical servers/computers, virtual machines only report the base frequency, they don't report the turbo frequency the same way as the idle one is also not reported.

    So in short, it is working, but the VM is just not reporting it.


    Best regards,
    Leon


    Blog: https://thesystemcenterblog.com LinkedIn:

    Friday, August 10, 2018 8:17 AM
  • Hi,

    Thank you for your reply. 

    But how does that explain the differences in our tests? We can see that on the VM there is almost no difference in turbo boost being turned on or off. On the physical host we do see a big difference. 

    Kind Regards,
    Joël

    Friday, August 10, 2018 9:12 AM
  • The max turbo speed depends on the number of cores in use, CPU will never turbo to max speed width all cores in use.

    Example

    If all cores are running at 100%, the turbo won't boost to max freq, this usually only happens with a single core at 100%, i.e., using a single threaded app, that's how it's designed.

    VMs rarely, if ever report CPU speed properly, even Hyper-V doesn't do it properly with Windows hosts and clients.

    Best regards,
    Leon


    Blog: https://thesystemcenterblog.com LinkedIn:

    Friday, August 10, 2018 9:23 AM
  • But that would be assuming I think that we did run stress tests or something like that and monitoring CPU usage? 

    That is not how we did our tests. We are testing normal usage with software (say a normal website), and we are monitoring performance counters of our software from within our software. 

    Friday, August 10, 2018 11:57 AM
  • It’s difficult to say as it depends a lot on how the application is built. From personal experience I’ve gotten better results while performing stress tests versus any normal tests.

    Blog: https://thesystemcenterblog.com LinkedIn:

    Friday, August 10, 2018 12:07 PM