Learning Forum

[1829116]

Hi 
I posted the following question on another platform:
--------------------------------------------------------------------------------------------

Hi
In ANSYS 2021 R2, I am able to set the number of processors and number of GPUs for the FLUENT simulation:

But I am not able to do that in 2022 R2:

Any advice? I am using a machine with 60 processors and 2 GPUs

 

Thanks

-------------------------------------------------------------------------------------------------

I then got the following helpful answer:

Hi 1829116, 
As Erik correctly mentioned since it’s a structural mechanics forum so you can post this question in the fluent forum. but for your answer, you can click on the show more option and select a “parallel setting” and find the option of the “solver gpgpus per machine” to choose. I have highlighted the option so you can see it. 


Regards,
Sampat

-------------------------------------------------------------------------------------------

I do have a followup question thou. Can I change the numer of GPUs after I have started a simulation? Usually I can do that for the number of processes from properties of the solution in Workbench. Any similar method for the number of GPUs

 

Thanks

Hadi

[Rob]

You can't change the number of processes once the solver is running. You can stop iterating, save, close and then when reopening Fluent change the core count. Fluent will automatically repartition to the new core count. 

Note, the GPGPU offload isn't the full gpu solver so read the documentation before switching it on. It has some benefits for certain models but isn't the (new) full gpu solver that's in 2023Rx. 

[1829116]

Thanks for the responese. 

We downloaded ANSYS 2023 R2 and are trying to use our Enterprise License to run the simulation on 2 GPUs. The simulation is a dynamic mesh model and this option (Dynamic Mesh) is disapearing when using the Enterprise Licesnse.

Please advise

 

Thanks

Hadi

 

[Rob]

Dynamic mesh is fine with Enterprise level licence, but may not be available on the full gpu solver. The functionality isn't (yet) as complete as the core solver. 

[1829116]

Thanks for the response. 

My machine has a CPU with 60 processors and 2 GPUs. What do you think is the best way to run my dynamic mesh model? I tried with the premium license to select the paralle setting and select 60 processors with two GPGPU but then the simulation only utilised the CPU and not the GPUs.

 

Thanks,

Hadi 

[Rob]

If it's not in the pure gpu solver then pure cpu. Depending on the cell count you may find using fewer cores to be faster too. 

[1829116]

So if I understand correctly, currently there is no way to utilise GPUs for a dynamic mesh model even in ANSYS 2023 R2?

[1829116]

 

I tried to follow what this video

I assigned the GPU to solve for all my variables (k – omega – pressure coupled – temperature) using the suggested commands. This showed no difference. My solution is still using the CPUs especially for the remeshing after every timestep.

 

[Rob]

How many cpu did you use?