Breaking the Mesh Barrier: High-Performance Fluid Simulations with DualSPHysics and CloudHPC
In the world of Computational Fluid Dynamics (CFD), we have long been “trapped” by the mesh. Whether you are simulating the splash of a breaking wave or the complex rotation inside a centrifugal pump, the mesh is often the bottleneck—taking days to generate and even longer to compute.
Today, we’re looking at DualSPHysics, a game-changing open-source solver that tosses the mesh aside in favor of particles. When combined with the NVIDIA Tesla T4 GPUs available on CloudHPC.cloud, it creates a high-performance environment that makes complex simulations accessible to everyone.
What is SPH?
Smoothed Particle Hydrodynamics (SPH) is a meshless, Lagrangian method.
- Lagrangian (Particle-based): Imagine you are sitting on a drop of water as it flows. You follow its path, its velocity, and its pressure. In SPH, the fluid is represented by a set of points (particles) that move according to the laws of physics.
- Smoothed: To calculate the properties at any given point, the software “smooths” the data from neighboring particles using a mathematical kernel function.
SPH vs. Finite Volume Method (FVM): What’s the Difference?
Most traditional CFD tools (like OpenFOAM or Ansys Fluent) use the Finite Volume Method (FVM). Here is why SPH is a different beast:
- Mesh vs. No Mesh: FVM requires an Eulerian grid—a fixed “cage” through which fluid flows. Creating this grid for complex moving parts (like a pump impeller) is incredibly labor-intensive. SPH is mesh-free; if you have the 3D geometry, you can start simulating immediately.
- Free Surfaces & Splashing: FVM struggles with “free surfaces” (where water meets air). It requires complex interface-tracking like Volume of Fluid (VoF). SPH handles splashing, droplets, and breaking waves naturally because the particles simply move where the forces push them.
- Deforming Boundaries: In FVM, moving boundaries often require “moving meshes” that can become distorted or fail. In SPH, the particles interact with solid boundaries without any grid-related numerical instability.
Accelerating with NVIDIA Tesla T4 on CloudHPC
DualSPHysics was designed from the ground up to be GPU-native. While a CPU handles tasks one by one, a GPU handles thousands of particle interactions simultaneously.
The NVIDIA Tesla T4 available on CloudHPC.cloud is a perfect match for DualSPHysics for several reasons:
- Turing Architecture: The T4 is built on the Turing architecture, providing optimized CUDA cores that DualSPHysics uses to calculate particle-to-particle interactions at lightning speed.
- Energy-Efficient Performance: Despite its small 70W footprint, the T4 delivers high single-precision (FP32) performance. Since DualSPHysics is often run in single-precision for maximum speed without sacrificing engineering accuracy, the T4 provides an incredible “bang for your buck.”
- Memory Bandwidth: SPH is memory-intensive. The T4’s GDDR6 memory allows for the rapid data transfer required to update the positions of millions of particles every microsecond.
Why Run on CloudHPC?
You don’t need to own a $10,000 server to run these simulations. By using CloudHPC.cloud, you get:
- Instant Access: Spin up a Tesla T4 instance in minutes.
- Scalability: Run multiple “What If?” scenarios in parallel.
- Pre-configured Environment: No need to struggle with CUDA drivers or compiling complex libraries; the platform is ready for DualSPHysics out of the box.
Conclusion
Whether you are simulating a pump, a dam break, or an offshore structure, DualSPHysics offers a level of flexibility that traditional mesh-based solvers simply cannot match. By leveraging the GPU power of the NVIDIA Tesla T4 on the cloud, you can turn a week of processing into a few hours of work.
Ready to try it? Head over to CloudHPC.cloud and start your first particle-based simulation today.
CloudHPC is a HPC provider to run engineering simulations on the cloud. CloudHPC provides from 1 to 224 vCPUs for each process in several configuration of HPC infrastructure - both multi-thread and multi-core. Current software ranges includes several CAE, CFD, FEA, FEM software among which OpenFOAM, FDS, Blender and several others.
New users benefit of a FREE trial of 300 vCPU/Hours to be used on the platform in order to test the platform, all each features and verify if it is suitable for their needs
