Memory Usage
RS3 uses an efficient algorithm to analyze 3D FEM analysis with different mesh options available. Depending on the mesh settings, RS3 will optimize the memory usage in the engine for the analysis. For large and complex 3D models, it is important to pay attention to the memory usage of the software with the different settings.
Below are the recommended memory space and hard drive disk space for the following analysis options:
RAM
Minimum requirement : 32 GB
Recommended for 4-noded tetrahedra: 64 GB
Recommended for 10-noded tetrahedra: 128 GB
Hard Drive
Recommended: 1TB SSD.
Examples
Below shows an example of memory usage for 4 nodes (3 mil of dofs) and 10 nodes (~25mil of dofs) calculation using iterative solver.
| 4 node 6 mil element, 1mil nodes (elastic) | 10 nodes, 6 mil elements, 8.4 mil nodes (elastic) | |
| Peak Memory (Gb) | 17.0 | 128 |
| Memory use stable (Gb) | 14.9 | 81.0 |
| Model data (Gb) | 11.6 | 45.6 |
| Global matrix size (Gb) | 2.6 | 20.0 |
| Peak in global assembling (Gb) | 3.6 | 34.0 |
| preconditioner (Gb) | 1.3 | 12.0 |
| Peak in preconditioner (Gb) | 3.5 | 42.8 |
| Total time (mins) | 20.7 | 792.1 |
| Total times reading input files (mins) | 2.1 | 3.8 |
Total time computing (mins) | 17.7 | 778.7 |
| Total time solving matrix (mins) | 16.6 | 535.5 |
| Total time read writing results (mins) | 0.9 | 9.6 |
Using 10 node increase memories 7.5 times. Also 10 nodes use virtual memory on hard drive so it would be much slower (35 times). Time ratio used to solve matrix is smaller in 10 nodes than in 4 nodes because of the different algorithm used in solving 10 node matrix.
Using SSR or advanced material models that requires additional parameters to be store at gauss points will increase memory requirements.