Learning Forum

[Uniform]

I upgraded my hardware from an i7 4790k 16gb 1333mhz ram to ryzen 9 3900x with 32gb 3200mhz ram and when I run my meshes it will only fully utilize my ram. How do I get it to fully utilize all 12 cores at nearly 100%? I am using 19.2.

[Rob]

Assuming they're physical cores and not due to hyperthreading you'll need an HPC pack: you don't get that with the Student or Teaching version of the code. How many cells are in the model, and which solver (Mechanical, Fluent etc) are you using?

And re-reading, unless you're building big meshes in Fluent Meshing don't worry too much about parallel meshing. 

[Uniform]

im using mechanical

[Uniform]

In that case what can I do to make the program to run as fast as possible and utilize as much computing power my computer can provide?

[peteroznewman]

Which ANSYS license are you running under?  Is it a commercial or Research license?

Then the company/university can purchase HPC licenses or HPC Pack licenses.

Maybe they already have.

Reply with the text between the [ ] at the top of the Mechanical window.

 

 

[Uniform]

I am using the academic research HPC

[Rob]

Then you can use 12 cores parallel with one additional pack. I'd check the RAM, you may be paging which will slow the calculation down. Easy way if you're using Win10 is to use Task Manager and check for excessive disc activity. 

As an aside 16GB with 12 cores seems an odd configuration: can you confirm that you have 12 real cores and not 6 real with hyperthreading? 

[Uniform]

I am running 12 real cores with 32gbs of ram

[Rob]

So, if you launch Mechanical with 11-12 cores parallel how does the cpu loading & RAM usage look?  You may find using one less core than the hardware has to be faster depending on operating system demands. 

[Uniform]

How do I launch in parallel

[peteroznewman]

If you mean you are using Mechanical Workbench, then in Mechanical, you go to Tool, Solve Process Settings... click the Advanced button and type in 11 for Number of Cores. I have run the same model with different number of cores and there are diminishing returns as the core count goes up, especially if you have large contact areas. For example, going from 2 to 4 cores might solve in 60% of the time of 2 cores, then going from 4 cores to 8 cores might solve in 75% of the time of 4 cores. Different models will scale differently. Using every core can be slower than using one less as rwoolhou said.

If you get an error when you try to solve with 11 cores, try using 10 cores. I don't recall the number of cores 19.2 allowed without HPC, it may have been 4 or 2. An HPC Pack adds 8 cores to the base number.  If you can't solve with 6 cores, then there is a license issue. Are you sharing the license with other users?  In that case, someone else might be using the license. There are ways to check this.

[Rob]

You may cap at 8 cores with one HPC pack, I know it's 12 at 2019Rx  but can't remember for R19.2   Good excuse to update to the later software too. 

[Omi]

.

In Ansys Electronics you can add more cores to be used by changing the HPC Options.

Go to Simulation -- HPC options -- Configurations -- and then select Edit and set the Tasks and Core numbers to the one you desire.

There is one concept that I cannot understand, what will happen if we set the Tasks number to more than 1? How the simulation will be solved?

.