Parallel Computing-Solution didn't match with those run locally
 

Forumers,

I used simpleFoam to calculate the car-like body - Ahmed body. I used snappyHexMesh to create the grids and it turns out that the total number of cell is 1.3 million. By any chance, i chose to compute parallel in TACC Stampede. The settings in constant and system are totally same in both cases, locally and parallel.

However, the velocity fields computed in parallel are entirely nonsense b/c the velocity could achieve to 1km/s, which is unattainable in car aerodynamics.

Can any masters tell me what's going on here?

Info:

6 mpi tasks by using decomposePar;