MD after running about 2000 steps, electronic steps cost more time,even longer than 1000s

[jumin... at gmail.com]

Hi, the output file is too large to upload. I Interrupt this task, and 
restart from the filename.restart, it can run fast at the every steps .
I attach the end of the my interrupt task.

*******************************************************************************
 ENSEMBLE TYPE                =                                             
 NVT
 STEP NUMBER                  =                                             
2148
 TIME [fs]                    =                                     
 2148.000000
 CONSERVED QUANTITY [hartree] =                             
 -0.190313285471E+04

                                              INSTANTANEOUS             
AVERAGES
 CPU TIME [s]                 =                     1126.39               
 55.62
 ENERGY DRIFT PER ATOM [K]    =          0.144496897105E+02   
0.106968597441E+02
 POTENTIAL ENERGY[hartree]    =         -0.190310055425E+04 
 -0.190315711107E+04
 KINETIC ENERGY [hartree]     =          0.424781016194E+00   
0.388406053299E+00
 TEMPERATURE [K]              =                     311.580             
 284.898


my restart job:
*******************************************************************************
 ENSEMBLE TYPE                =                                             
 NVT
 STEP NUMBER                  =                                             
2879
 TIME [fs]                    =                                     
 2879.000000
 CONSERVED QUANTITY [hartree] =                             
 -0.190313525184E+04

                                              INSTANTANEOUS             
AVERAGES
 CPU TIME [s]                 =                       33.54               
 45.19
 ENERGY DRIFT PER ATOM [K]    =          0.125381875450E+02   
0.112253094524E+02
 POTENTIAL ENERGY[hartree]    =         -0.190313820972E+04 
 -0.190315644925E+04
 KINETIC ENERGY [hartree]     =          0.370838312434E+00   
0.391549366589E+00
 TEMPERATURE [K]              =                     272.012             
 287.204
 *******************************************************************************

在 2015年4月30日星期四 UTC+8下午12:32:45，jumi... at gmail.com写道：
>
> Dear  cp2k users and developers,
>          I run MD at NVT, however, after run about 2000 steps, the job run 
> slowly and slowly. if I restart the job from the slow step point, the job 
> can run fast like at start time.  anyone have the similar problem?
>          I attach my input. 
>
>
>  mixed_precision    : F
>
>   ----------------------------------- OT 
> ---------------------------------------
>
>   Step     Update method      Time    Convergence         Total energy   
>  Change
>   
> ------------------------------------------------------------------------------
>      1 OT DIIS     0.15E+00   94.8     0.00003762     -1903.1123236014 
> -1.90E+03
>      2 OT DIIS     0.15E+00 1387.5     0.00002286     -1903.1127142782 
> -3.91E-04
>      3 OT DIIS     0.15E+00  143.4     0.00001048     -1903.1129615906 
> -2.47E-04
>      4 OT DIIS     0.15E+00  264.3     0.00000683     -1903.1130262224 
> -6.46E-05
>
>   *** SCF run converged in     4 steps ***
>
>
>   Electronic density on regular grids:      -1312.0000002470       
> -0.0000002470
>   Core density on regular grids:             1311.9999997705       
> -0.0000002295
>   Total charge density on r-space grids:       -0.0000004765
>   Total charge density g-space grids:          -0.0000004765
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20150430/3902bfa9/attachment.htm>