Announcement

Collapse
No announcement yet.

Intel Turbo Boost Max 3.0 Testing On The Linux 4.11 Kernel

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

  • Intel Turbo Boost Max 3.0 Testing On The Linux 4.11 Kernel

    Phoronix: Intel Turbo Boost Max 3.0 Testing On The Linux 4.11 Kernel

    With the upcoming Linux 4.11 kernel release there is better support for Intel Turbo Boost Max 3.0 after originally it didn't look like this feature would be available for Linux. Under Linux 4.11, my Core i7 6800K + MSI X99A WORKSTATION box is now working with "ITMT" enabled, so here are some quick benchmarks...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    I see nothing interesting here. I guess you should test the same on Windows to see if it's an Intel fake technology, or a linux kernel bugged/limited implementation

    Comment


    • #3
      I think you should try testing a task that is better suited to a variable number of threads. For example, the task of computing the Tutte Polynomial of a moderate sized graph.

      When I first heard about TBM3, I thought it would be applicable to tasks like that. The reason is that, when computing the Tutte Polynomial of a graph, the number of active threads varies from one to dozens (or more, for large graphs if you have enough cores). Currently, without access to a processor with TBM3, I am stuck with the choice between:
      --Disabling all but a few cores and overclocking. This helps the single-threaded parts of the computation go faster, but some branchers of the computation of the Tutte Polynomial could gain a lot from parallelization
      --Keeping all cores enabled and the clock speed significantly lower. This makes some part of the computation go faster, but there is a considerable amount of work that either can't be parallelized, or can't use more than a few threads.

      I would think, without actually having tested it, that Turbo Boost Max 3 would help: all cores could be on and active when they are needed, but could automatically disable themselves during inherently serial portions of the computation, allowing the one or few active threads to bump up in speed a little.

      Hopefully, my department will get a CPU with TMB3 in the near future, and then I will be able to test whether it actually helps to a noticeable degree.

      Comment

      Working...
      X