Cavity 3D : 4.1 versus 3.0.1 performance issue
 

Hi guys,

I’m using the cavity-3D case with a 3M cells mesh.

I’m currently having a performance issue while using 4.1 version.

Here is the output I got with 4.1.

Starting time loop

Time = 0.0001

Courant Number mean: 0 max: 0
smoothSolver: Solving for Ux, Initial residual = 1, Final residual = 1.68989e-06, No Iterations 3
smoothSolver: Solving for Uy, Initial residual = 0, Final residual = 0, No Iterations 0
smoothSolver: Solving for Uz, Initial residual = 0, Final residual = 0, No Iterations 0
DICPCG: Solving for p, Initial residual = 1, Final residual = 0.0477591, No Iterations 153
time step continuity errors : sum local = 2.75823e-08, global = 1.4455e-23, cumulative = 1.4455e-23
DICPCG: Solving for p, Initial residual = 0.0857006, Final residual = 9.88484e-07, No Iterations 528
time step continuity errors : sum local = 1.13433e-12, global = 6.91952e-24, cumulative = 2.13745e-23
ExecutionTime = 4.56 s ClockTime = 4 s

While I got the following with 3.0.1:

Time = 0.0001

Courant Number mean: 0 max: 0
smoothSolver: Solving for Ux, Initial residual = 1, Final residual = 1.68989e-06, No Iterations 3
smoothSolver: Solving for Uy, Initial residual = 0, Final residual = 0, No Iterations 0
smoothSolver: Solving for Uz, Initial residual = 0, Final residual = 0, No Iterations 0
DICPCG: Solving for p, Initial residual = 1, Final residual = 0.0477591, No Iterations 153
time step continuity errors : sum local = 2.75823e-08, global = -1.14771e-23, cumulative = -1.14771e-23
DICPCG: Solving for p, Initial residual = 0.0857006, Final residual = 0.00406192, No Iterations 149
time step continuity errors : sum local = 4.79613e-09, global = 9.74005e-24, cumulative = -1.73708e-24
ExecutionTime = 2.59 s ClockTime = 6 s

Does anyone know why It would take twice the number of iteration to make the same step in 4.1 than in 3.0.1 ?

Thanks in advance,

Ludo

In 4.1 you arrive at very small final residual (9.*e-7) compared to the 3.0.1 case (4.*e-3). So starting from the same initial residual, achieving a much smaller final one will take more numerical effort (i.e. iterations).
So, apparently the behaviour of 4.1 and 3.0.1 is different for the case you ran. Since the case has not been posted, there is nothing more to say.
If you make OpenFOAM to reach the same final residual, the number of iterations will most probably be quite comparable.