Hi, everyone, I have successfully compiled the cp2k in openmp+mpi. When I start the cp2k by following mpirun command, I find the times for every SCF step are different remarkablly, and also find that the CPU usage on the computing node are not always 800%(There are two 4-core CPU on each node). Does anyone have idea about it or have any experience to promote the paralleling efficiency. The command: /opt/mvapich2-1.8.1/bin/mpirun_rsh -ssh -hostfile nodes.$$ -np 10 MV2_ENABLE_AFFINITY=0 OMP_NUM_THREADS=8 cp2k.psmp TiO2.inp >TiO2.out (I have 10 nodes in the nodesfile) The OUT PUT for SCF: ----------------------------------- OT --------------------------------------- Step Update method Time Convergence Total energy Change ------------------------------------------------------------------------------ 1 OT DIIS 0.15E+00 78.8 0.00489573 -11358.8478798550 -1.14E+04 2 OT DIIS 0.15E+00 49.5 0.00418778 -11408.3162227805 -4.95E+01 3 OT DIIS 0.15E+00 74.5 0.00337817 -11430.9534193832 -2.26E+01 4 OT DIIS 0.15E+00 48.2 0.00386960 -11458.1345581625 -2.72E+01 5 OT DIIS 0.15E+00 47.1 0.00203908 -11483.8037596735 -2.57E+01 6 OT DIIS 0.15E+00 63.8 0.00156578 -11514.2831610125 -3.05E+01 7 OT DIIS 0.15E+00 66.8 0.00114899 -11541.2744732691 -2.70E+01 8 OT DIIS 0.15E+00 47.8 0.00088405 -11557.1215180177 -1.58E+01 9 OT DIIS 0.15E+00 62.4 0.00074951 -11566.0261626878 -8.90E+00 10 OT DIIS 0.15E+00 69.6 0.00063936 -11573.9602882755 -7.93E+00 75 OT DIIS 0.15E+00 85.0 0.00001131 -11617.3170395951 -8.91E-03 90 OT DIIS 0.15E+00 45.1 0.00000688 -11617.3824121651 -1.08E-03 Best wishes Jianfeng