Why parallel running does not reduce time?


Just Got Here
Can Any one help me about the following issue? Thanks in advance!

I used nwchem to optimize the ethanol with the exact same input file but different numbers of cpu:

The input file ethanol_opt.nw goes as follow:

echo

start molecule

title "ethanol"
charge 0

geometry units angstroms print xyz autosym
  C       -8.30097        2.76242        0.01298
C -6.81962 3.11668 0.03691
H -8.85443 3.41915 0.71675
H -8.44508 1.70474 0.31942
H -8.70929 2.90452 -1.01004
H -6.40793 3.00539 1.06576
H -6.69124 4.17959 -0.25769
O -6.10974 2.31333 -0.86567
H -5.98622 1.43184 -0.42574
end

basis
 * library 6-31G*
end

dft
 xc b3lyp
mult 1
end

stepper
   maxiter 200
end

task dft optimize



The command to run the jobs are as follow:
(1) mpirun -np 1 nwchem ethanol.nw > ethanol.out 2>&1 &
(2) mpirun -np 2 nwchem ethanol.nw > ethanol.out 2>&1 &
(3) mpirun -np 4 nwchem ethanol.nw > ethanol.out 2>&1 &
(4) mpirun -np 8 nwchem ethanol.nw > ethanol.out 2>&1 &
(5) mpirun -np 16 nwchem ethanol.nw > ethanol.out 2>&1 &
(6) mpirun -np 32 nwchem ethanol.nw > ethanol.out 2>&1 &


The job running results are as follow:
(1) mpirun -np 1 nwchem ethanol.nw > ethanol.out 2>&1 &
The job finished with "Total times cpu: 209.3s wall: 121.1s"

(2) mpirun -np 2 nwchem ethanol.nw > ethanol.out 2>&1 &

The job failed with some error info. as follow:
rtdb_seq_get: type mismatch "dft:theory" in ./molecule.db: arg=1000, db=0
------------------------------------------------------------------------
task:energy: theory not specified 0
------------------------------------------------------------------------
------------------------------------------------------------------------
current input line :
33: task dft optimize
------------------------------------------------------------------------
------------------------------------------------------------------------
An error occured in the Runtime Database
------------------------------------------------------------------------

(3) mpirun -np 4 nwchem ethanol.nw > ethanol.out 2>&1 &

The job failed with some error info. as follow:
 maximum number of geometry optimization 
iterations exceeded: restartable shutdown will occur
------------------------------------------------------------------------
dft optimize failed 0
------------------------------------------------------------------------
------------------------------------------------------------------------
current input line :
33: task dft optimize
------------------------------------------------------------------------
------------------------------------------------------------------------
This type of error is most commonly associated with calculations not reaching
convergence criteria
------------------------------------------------------------------------

(4) mpirun -np 8 nwchem ethanol.nw > ethanol.out 2>&1 &

The job failed with some error info. as follow:
 maximum number of geometry optimization 
iterations exceeded: restartable shutdown will occur
------------------------------------------------------------------------
dft optimize failed 0
------------------------------------------------------------------------
------------------------------------------------------------------------
current input line :
33: task dft optimize
------------------------------------------------------------------------
------------------------------------------------------------------------
This type of error is most commonly associated with calculations not reaching
convergence criteria
------------------------------------------------------------------------

(5) mpirun -np 16 nwchem ethanol.nw > ethanol.out 2>&1 &

The job failed with some error info. as follow:
 maximum number of geometry optimization 
iterations exceeded: restartable shutdown will occur
------------------------------------------------------------------------
dft optimize failed 0
------------------------------------------------------------------------
------------------------------------------------------------------------
current input line :
33: task dft optimize
------------------------------------------------------------------------
------------------------------------------------------------------------
This type of error is most commonly associated with calculations not reaching
convergence criteria
------------------------------------------------------------------------

(6) mpirun -np 32 nwchem ethanol.nw > ethanol.out 2>&1 &
The job failed with some error info. as follow:
 
maximum number of geometry optimization
iterations exceeded:  restartable shutdown will occur
------------------------------------------------------------------------
dft optimize failed 0
------------------------------------------------------------------------
------------------------------------------------------------------------
current input line :
33: task dft optimize
------------------------------------------------------------------------
------------------------------------------------------------------------
This type of error is most commonly associated with calculations not reaching
convergence criteria
------------------------------------------------------------------------

So, why using more than one cpu does not reduce the running time but make the job failed with error?
By the way, when use multiple cpu, the output file becomes very large, eg, 142MB for "-np 32", how can the file size being reduced?

Thank you~


Forum >> NWChem's corner >> Running NWChem