Creating a Virtual Desktop with Ansys Fluids
The Ansys Fluids virtual desktop experience is summarized in the following steps:
Create a virtual desktop. A project space administrator creates a virtual desktop with an Ansys Fluids package installed. See Creating a Virtual Desktop in the Administration Guide.
Important notes for Ansys Fluids:
Operating system: Windows or Linux
If you are using SpaceClaim to create or modify your model, you must start with a Windows virtual desktop.
Applications. Select an Ansys Fluids package.
If you are creating a Linux virtual desktop, you must add either KDE Desktop Environment or GNOME Desktop Latest Version to the list of applications to be installed.
GPU workflows: For accelerated graphics, use GNOME. For GPU compute, you can use either GNOME or KDE as long as the correct driver type is selected when specifying hardware.
Note:For a Windows virtual desktop, Ansys recommends that you create the virtual desktop without any applications, and then add the application after the virtual desktop has been created. See Adding Applications to a Virtual Desktop in the User's Guide. Sometimes, Windows updates are triggered on newly created virtual machines, causing application installation to fail.
When creating a Linux virtual desktop, installing Ansys Fluids in a shared file location is not supported. An application can only be installed on Azure Files NFS shared storage during the creation of an autoscaling cluster. You can, therefore, point to a shared file location that already has Ansys Fluids installed (if one exists).
Hardware: The recommendations are outlined below.
Graphics Nodes for Pre/Post
- Standard_NV_A10_v5*
- Standard_NV_v3*
- Standard_NC_T4_v3*
GPU requirement for interactive workflows: Use a virtual machine (VM) with a GPU. Although non-GPU machines may handle low-level interactive tasks, they are primarily suited for non-graphical batch workflows to ensure optimal performance and compatibility.
Large simulation models: A VM size with a graphics card is recommended. These nodes may have enough CPU cores to run small models. On Linux VMs, use GNOME to enable accelerated graphics.
Series selection and memory requirements: Your choice of VM series should be based on the graphics card available. Memory requirements depend on many factors including the mesh size and type, single vs double precision, and the physical model used.
For VDI workflows, always check the total memory available on a VM to ensure it meets the needs of the models being run.
Storage requirement: A minimum of 256 GB of local storage is required for the installation of Ansys Fluids.
Pre- and post-processing: A VM with a single GPU and the required CPU core count is sufficient.
Hyperthreading
The NV series has hyperthreading enabled and it cannot be disabled.
The NC series has hyperthreading disabled.
Compute/Memory Optimized Series
- E series (for example, Standard_E32)
- F series (for example, Standard_F16)
Core count options: Multiple options for core counts are available. Select the VM size based on the model size and workload requirements.
Memory considerations: Memory requirements vary depending on factors such as mesh size and type, single vs double precision, and the physical models used.
Recommendation: VM sizes with lower core counts are available but are not recommended for production workloads.
Hyperthreading:
The E series has hyperthreading enabled and it cannot be disabled.
The F series has hyperthreading disabled.
GPU Nodes for Ansys Fluent Native GPU Solver
- Standard_NC_A100_v4*
- Standard_NV_v3*
- Standard_NC_T4_v3*
- Standard_NC_H100_v5*
- Standard_ND40rs_v2*
Multi-GPU support: Some GPU VM sizes have multiple GPU cards. Ansys Fluent Native GPU Solvers can run on one or more GPU cards. Select Tesla drivers when creating these VMs.
Single GPU compute workloads: Using the GRID driver enables both accelerated graphics and GPU compute.
Driver compatibility for Ansys Fluent 2025 R2:
GPU compute is not supported on virtual machines with GRID drivers.
The available GRID driver versions support CUDA toolkit version 12.2. Ansys Fluent 2025 R2 requires CUDA 12.8 for GPU compute. For more information see Supported GPUs and Drivers in the Fluent User's Guide.
NVIDIA A10G GPUs: When configured for multi-GPU compute, A10G GPUs experience performance loss since they are only compatible with GRID drivers.
Hyperthreading:
The NV series has hyperthreading enabled and it cannot be disabled.
The NC and ND series have hyperthreading disabled.
High Performance Compute
- HC series (for example, Standard_HC44rs)
- HB_v4 series (for example, Standard_HB176-48rs_v4)
- HB_v3 series (for example, Standard_HB120rs_v3)
- HB_v2 series (for example, Standard_HB120rs_v2)
Node size selection: Choose the cluster node size based on model size and workload requirements.
VDI workflows: These VM sizes can also be used for VDI workflows, as they provide multiple core count options to match different use cases.
Hyperthreading: These VM sizes have hyperthreading disabled.
* Has GPUs (graphics processing units)
For GPU compute, follow these guidelines when choosing a GPU driver option:
- Single GPU compute: GRID or Tesla
- Multi-GPU compute: Tesla preferred for best performance, but not available for NVadsA10_v5
Launch a virtual desktop session. Users with access to the project space connect to the virtual desktop to start a virtual desktop session. See Launching a Virtual Desktop Session in the User's Guide.
Run the remote application. Once the desktop is open you can run simulations on the virtual desktop just as you would on your local desktop. For product information refer to the Fluent Documentation and CFX Documentation.
If you need to transfer files to the virtual desktop, see General Guidelines for Transferring Files in the User's Guide.
Note: Currently, launching Aero or Icing from the Fluent Launcher in Ansys Access on Microsoft Azure results in display issues. To work around this issue, see Running Ansys Fluent Aero and Fluent Icing (2024 R1 and 2024 R2).