ANSYS Workbench

The ANSYS Workbench step can be used to start Workbench on a GPU worker with hardware accelerated graphics support. Note that, users can also start ANSYS Workbench using the desktop shortcuts for ANSYS Workbench, however, this method does not use hardware acceleration for 3D. This is useful when you want to use ANSYS Workbench with the built-in HPCBOX connector to create a simulation workflow within ANSYS Workbench and include cluster operations and job submission with RSM to the HPCBOX cluster. Please read more on how to achieve this in the section titled “HPCBOX Connector for ANSYS Workbench”.

Configuring local scratch on workers for structural jobs

To configure local scratch use on the workers for Mechanical jobs, please follow the instructions below:

• In a terminal window within your HPCBOX Remote Desktop run the following commands:

module load ansys
rsmclusterconfig

This opens the RSM config manager and within the manager, follow the instructions as per the screenshot below.

RSM Setup

The screenshot below shows the settings to submit a CFX solver run through RSM to the workers once the HPCBOX component has finished starting the workers.

Configuring alternate Processes Per Node (PPN) for specific CPU architectures

When submitting distributed parallel jobs on specific CPU architectures like AMD’s Epyc architecture, you might get better performance when limiting the number of processors used on every compute worker. This is because of the CPU design and NUMA node architecture and how they access memory. By default, ANSYS RSM on HPCBOX is configured to use all CPU cores on any particular compute-worker, however, this can be changed and the general recommended setting for certain Azure instance sizes are below: HB60rs : 45 PPN HB120rs_v2: 90 PPN

:::info  It must be noted that the effectiveness of this setting is also impacted by the user’s model. :::

To change the default setting, users can follow the instructions below:

module load ansys
rsmclusterconfig

This will open the RSM cluster configuration utility. Within the utility, navigate to HPC Resources -> HPCBOX -> all.q, then select the HPC Resource tab.

Change the value of Distributed Parallel to the corresponding PPN value using the table below:

PPN	Value	AZURE Compute Worker Instance Validity
ALL	orte	HB60rs and HB120rs_V2
45	orte_ppn_45	HB60rs and HB120rs_V2
60	orte_ppn_60	HB120rs_V2/3/4
90	orte_ppn_90	HB120rs_V2/3/4
116	orte_ppn_116	HB120rs_V2/3/4

Click on Apply after changing the value.