Learning Forum

[AndyHor]

Hi,

 

I am using Workbench R3 Fluent to run a transient problem, my computer is i5 with 4 processors, and 1 graphic card Nvidia GTX1080 TI,

but when I run the calculation in fluent, it used 4 processors only, but not using GPU, how can I make Fluent to make use of using this GPU to run ?

 

I have tried the method from https://forum.ansys.com/forums/topic/using-gpu-in-fluent/

but it could not even run and auto shut down fluent.

 

Please help. Thank you !

[Rob]

There's two parts to this. First, you'd need an extra parallel licence, which with Student & Teaching packs you don't have: you'd need to pay for an additional pack (and possibly the main solver too). Secondly, GPU only gives you acceleration in specific models and (I think) some of the multigrid maths. 

Finally, with the  Student & Teaching node limits, and 4 cores you're hitting 112k cells per core: more parallel will help with speed but another 4 cores will only half the iteration time. 

[AndyHor]

My uni lab is using [ANSYS Academic Teaching Introductory] version, do this have extra parallel licence ? 

 

How much around will the licence cost ? 

[Rob]

You're limited to 4 cores with Teaching and 512k cells. I'd forget using GPU for now.  

If you need a lot more cells/speed then talk to your supervisor about Research licences: that removes the cell cap and gives access to the HPC packs. If your University doesn't have these we can discuss further. 

[AndyHor]

We dont have the much fund to support for a research licences. 

So, can you teach me on way using the GPU plus core ?

 

Thank you

[peteroznewman]

You can't use a GPU on a Teaching license.

[AndyHor]

Dear Rwoolhou, 

My university computer has a research license [ANSYS Academic Research Mechanical and CFD], as illustrate in picture below. How could I speed up the calculate if my present computer only have 4 cores ? 

Please advise, hope to hear from you very soon. 

 

[Rob]

With a Research Key and HPC pack you should be able to turn on gpu. However gpu only speeds up certain parts of the calculation (S2S cluster calculation is the one I tend to remember). Without the HPC you may be able to run 3 parallel & 1 gpu: the 4th cpu core will probably be quicker than using the gpu. 

In terms of speeding up the calculation the obvious one is a bigger computer. However, using a good quality mesh and well posed problem can also speed up the process as each time step will solve more quickly. 

[StePiz2019SAE]

HI, i have a research license [ANSYS Academic Research Mechanical and CFD] and a personal workstation with Intel i9 10980XE (18 cores - 36 threads ) 3.00GHz to boost 4.60GHZ and PNY Quadro P4000 - 8GB. 

Using only physical core (not threads), what clock frequency does the solver use? 

The license can permitted to me to use 16 core in parallel, it's true? Can I use acceleration GPGPU? I use ANSYS for an external aerodynamics problems. On the platform says that the gpu in question allows it to be used on fluent. Can I use all the various fluent situations? isn't it limited to just a few things?

Can I setup fluent launcher with 16 core and 1 gpu? 

Thank you.

[Rob]

You may be able to use a gpu with the Research licence, drop the number of cpu to 15 and try it with one gpu.  The solver will run the cpu at whatever speed the system will let it: remember data traffic, RAM, monitors etc will also influence the apparent speed. 

[StePiz2019SAE]

What do you mean for monitors? Do you mean the setup on fluent?

I have one license research academic on my personal workstation as license server and I can use the license even if I'm offline, what does data traffic have to do with it?