Announcement

Collapse
No announcement yet.

Explanation about perfirmance test

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

  • Explanation about perfirmance test

    Hi, I was looking into the forum to see if i can get information to the explanation of the command line parameters explanation for performance test but to no avail (maybe not looking/searching hard enough).
    Eg, processes - does it equate to hyper-threaded core or something else. Why processes parameter (or argument) value at 1024, what is the meaning of duration for short, medium and long etc.
    I want to look for some guide explanation these. I sure need some help to direct me to the source which allow me to understand what each of the parameter means (in layman explanation).
    I want to do performance test on a server and its better to understand each parameters mean, so that when I do the test, I can adjust accordingly to observe and better understand/appreciate the output.
    Thanks for any help. Cheers.

  • #2
    The Help file (under the Help menu) in the software is the best place to start.

    Extract is below.

    Click image for larger version

Name:	image.png
Views:	193
Size:	339.6 KB
ID:	56112

    Comment


    • #3
      THank you David.

      Comment


      • #4
        Hi David
        For processors, which is mentioned as:
        How many threads to use when running the CPU tests. Should be set to the total number of cores in the computer for best results (Number of CPU's * number of cores per CPU).
        ​Assuming if I have 2 sockets of 64 cores each (total 128 cores), do i interpret that if processes is set at 128, we are testing the pcores and if i set this to 256 (it is testing pcores with hyperthreading, assuming I run it under a VM instance)?
        Lastly, if I were to set the processes to 256 and beyond (incremental of 12, am I right to interpret that I am testing the cpu with the level of oversubscription?

        Thanks for enlighting me.

        Comment


        • #5
          Number of threads any system can run at the same time is,
          <Number of CPUs> x <Number of P & E cores> x <Number of threads per core, on average>

          PerformanceTest should use this value by default, but you can force it lower of higher if you want.

          Forcing it higher should not result in better performance however. It should be around the same, unless you go way over, then it will get worse.

          Forcing it lower is more interesting. Could be better or worse, depending on the hardware.

          PerformanceTest doesn't force code to run on P-Cores in preference to E-Cores. The operating system does this.

          Comment


          • #6
            This post from a few years ago is also applicable and interesting.
            The Apparent Uselessness of a Second AMD EPYC 7742

            Comment

            Working...
            X