[haiteng]

Hello! I am implementing a two-dimensional URANS simulation with a mesh size of ~150k and a time step size of 0.01 s. The case is run on 16 CPU cores with single precision. In the beginning of the simulation, the case runs quite fast, and 1 s physical time (100 time steps) only takes an elapsed time of 11 min. However, as the calculation proceeds, the simulation speed decreases greatly. Now, it takes an elapsed time of 70 min for 1 s physical time! I haven't changed anything on the case. It is just there and keeps running. Could anyone give me hand out of this?

Here are two guesses from me which needs substantiation from professionals:
1. "Data Sampling for Time Statistics" is turned on, will this slow down computational speed as time evolves?
2. After each iteration, CPU usage drops sharply to 8% and then rises to 50% (which is 16 CPU on my computer) when the next iteration is calculated. This phenomenon repeats for each iteration (shown in the figure). Why does this occur? Will this hamper the simulation speed?

Many thanks!

[haiteng]

Figure mentioned in last post is attached here. Many thanks!