Ansys Free Student Software

Ansys Free Student Software

Software works weird; Processor 3995WX

    • offroader10052
      Subscriber

      Hello.

      I ask for advice.

      I bought an AMD Ryzen Threadripper PRO 3995WX processor and can't use it to its full potential.

      Problem:

      The same task in Ansys CFX is solved at different speeds.

      Each new run of the solution gives a solution at a different speed.

      test 1 - 3200 seconds
      test 2 - 900 seconds
      test 3 - 7000 seconds
      test 4 - 1400 seconds

      How to achieve the same minimum solution time?

      PC specifications
      - AMD Ryzen Threadripper PRO 3995WX (64 physical cores, 128 threads);
      - 128 GB RAM;
      - 512 SSD samsung evo;
      - no throttling occurs, temperature 50-55 °C;
      - Windows 10 x64;
      - Ansys 19.1.

      Did not help:
      - disable hyper trading
      - disable turbo boost
      - reinstall win to another copy
      - ssd replacement
      - Installing another version of Ansys 2021
      - installation of all driver and OS updates.

      I have run successfully this model before on another computer.


      Other programs run fine on this PC with the same repeatable results (benchmarks, video encoding, rendering).

    • Rob
      Ansys Employee

      The same job is taking a different time or different number of iterations or....?  Timings can be affected by compute load, ie what else the system is doing. How many cores are you using?

    • offroader10052
      Subscriber

      1)

      The same job is taking a different time or different number of iterations or....? 

      iterations

       

      2) 

      Timings can be affected by compute load, ie what else the system is doing.

      ok

       

      3) 

      How many cores are you using?

      I tried to solve the problem on the same number of cores.
      First on 4
      Then on 10
      fourteen
      32
      64

    • Rob
      Ansys Employee

      Iteration count should be roughly the same, but will depend on the initial condition being the same and may change for core count if there are any odd effects from there. 

Viewing 3 reply threads
  • You must be logged in to reply to this topic.