General Mechanical

General Mechanical

how to use NVidia RTX5000 GPU for mechanical HPC

    • Houston008
      Subscriber

      trying to use RTX5000 for ansys APDL HPC, but got error messages,

      wonder what i can do to use the RTX5000?

      thanks a lot,



      *** FATAL ***              CP =    0.438  TIME= 08:23:45

       No recommended GPU devices have been detected on machine

       Only Tesla-series, A10/A16/A30/A40/A100, Quadro             

       GP100/GV100/RTX6000/RTX8000 or RTX A4000/A5000/A6000 GPU devices are   

       recommended at this release. For optimal performance, install a     

       recommended GPU device in this machine. If you wish to use an      

       alternative GPU device, please review the recommendations in the     

       section titled "Requirements for the GPU Accelerator in Mechanical    

       APDL" in the Installation Guide for your platform.            

    • Gary Stofan
      Ansys Employee
      You are welcome to try setting the GPU override variable, but it still might not enable the RTX5000
      1.4. Requirements for the GPU Accelerator in Mechanical APDL (ansys.com)
      For NVIDIA GPU cards, the driver version must be 451.82 or newer. For optimal performance on Windows, the TCC (Tesla Compute Cluster) driver mode is recommended when using Tesla series GPU cards. Some limitations exist when using this driver mode. Check your GPU card documentation for more details on how to set this driver mode and the existing limitations.
      To utilize a NVIDIA GPU device that is not on the recommended list of cards, set the following Windows environment variable:
      Variable: ANSGPU_OVERRIDE
      Value: 1
      This is most beneficial when you wish to run on newer NVIDIA GPUs that were not available at the time of release of this version of the Ansys program. If you choose to use this environment variable, you should ensure that the NVIDIA GPU device that you wish to use is sufficiently powerful, in terms of both double-precision compute power and on-card memory, to achieve meaningful acceleration for your simulation. Using this environment variable with an underpowered CPU may actually decelerate your simulation.
    • Houston008
      Subscriber
      Thanks for the note.
      By setting ANSGUP_OVERRIDE=1, was able to use ANSYS APDL with RTX5000 GPU.
      however there is no benefit in speeding up my simulation, which is a 3D nonlinear contact analysis. As mentioned in ANSYS APDL manual.
    • Gary Stofan
      Ansys Employee

      Not directly related to your GPU issue, but in the Analysis Settings branch you may want to try setting Contact Split to Program controlled and see if speed is improved.

      https://ansyshelp.ansys.com/account/secured?returnurl=/Views/Secured/corp/v221/en/wb_sim/ds_advanced_controls.html?q=contact%20split
      Contact Split(DMP)
      When theContact Split(DMP)property is enabled, the contact conditions for the model are split during the distributed solution to generate several smaller contacts. These contacts are then processed across the available cores. Once the solution is complete, the split pairs are merged. This entire process is automatic. Property options include:
      Off(default): Nocontact splitting.
      Program Controlled: The application selects whether to performcontact splitting in order to improve performance.
      On: The application splits the contact regions during the solution. Selecting this option displays the additional propertyMax Number of Splits. This entry specifies the maximum number of splits that can occur for each contact region during the distributed solution. A minimum value of 2 is acceptable.


    • Houston008
      Subscriber
      thanks a lot for the note. yes this would help speeding up simulations.
Viewing 4 reply threads
  • You must be logged in to reply to this topic.