The machine ran out of memory. Please choose a larger machine by increasing the number of computing cores or reduce the fineness of your mesh.
This error occurs when the machine selected for your simulation runs out of memory. This means that the machine selected was too small for the simulation to be able to run until it is finished.
What Could Be the Possible Reason?
The machine can run out of memory because of several reasons, as stated below:
- The size of the flow domain: The flow domain is too large for the machine to run a simulation, therefore the simulation requires a larger machine.
- Element size: The element size is too small which produces a large mesh size.
- The selected number of cores: The number of cores you selected for the simulation run is too small.
What Can I Do Now?
You can solve the problem in several ways, as stated below:
CFD and Structural Analysis
- Use the Standard mesher: The standard mesher is an automatic mesher that produces a three-dimensional unstructured mesh that consists of tetrahedral or hexahedral elements. This mesh ranges from 1-very coarse to 10-very fine.
You can do this by going to the Mesh and changing the Algorithm to Standard. This will change the meshing algorithm to SimScale’s standard meshing algorithm which takes into account the geometric properties of the model.
- For mesh related solutions, the mesh needs to be regenerated and a new simulation needs to be run. Note that the coarser the mesh, the faster it is to generate.
- After the number of cores has been increased, a new simulation run will need to be created to implement the change to the number of assigned cores.
- Enlarge the size of elements: The larger the size of the elements, the lower the number of elements required. You can do this by decreasing the fineness level of the mesh. It is better to decrease the mesh by one level to find the balance between good mesh and enough memory space.
This can be done by going to the Mesh and decreasing the Fineness level. The fineness level will be increased in according to the features of the geometry. It is important to remember that the decreasing the fineness level will make the mesh generate faster.
- Increase the number of cores: The larger number of cores assigned will increase the memory available for the simulation. Currently, SimScale supports up to 96 cores.
You can do this by going to Simulation control and changing the number in the Number of processors.
PWC and LBM
For an LBM or PWC simulation, if the user encounters this error, the solution is to reduce the Fineness of the mesh and this can be done in the Mesh settings of the simulation.