Diverged energy during DFPT calculation.

[abalone]

I am using Phonopy with DFPT to calculate the phonon dispersion of silicene.
The supercell I used is 3*3 as reported in literature.
However, I found that during DFPT calculation the energy will diverge during the second DOF.

Linear response progress:
Degree of freedom: 2/ 2
N E dE d eps ncg rms rms(c)
RMM: 1 -0.235284990488E-01 -0.23528E-01 -0.55648E-03 801 0.504E-01
RMM: 2 -0.219603695460E-02 0.21332E-01 -0.16654E-03 618 0.103E-01 0.578E-02
RMM: 3 0.374028335112E+03 0.37403E+03 -0.39783E-04 649 0.436E-02 0.141E-02
RMM: 4 0.748098316495E+07 0.74806E+07 -0.65498E-05 668 0.157E-02 0.757E-03
RMM: 5 0.149619663716E+12 0.14961E+12 -0.95111E-06 676 0.664E-03 0.184E-03
RMM: 6 0.299239327433E+16 0.29922E+16 -0.60764E-07 695 0.171E-03 0.347E-04
RMM: 7 0.598478654865E+20 0.59845E+20 -0.56977E-08 690 0.455E-04 0.151E-04
RMM: 8 0.119695730973E+25 0.11969E+25 0.11406E-08 693 0.198E-04 0.821E-05
RMM: 9 0.239391461946E+29 0.23938E+29 -0.12927E-08 632 0.207E-04 0.567E-05
RMM: 10 0.478782923892E+33 0.47876E+33 0.98422E-09 648 0.159E-04 0.395E-05
RMM: 11 0.957565847784E+37 0.95752E+37 -0.46883E-09 672 0.113E-04 0.299E-05
RMM: 12 0.191513169557E+42 0.19150E+42 0.62530E-09 663 0.101E-04 0.229E-05
RMM: 13 0.383026339114E+46 0.38301E+46 0.53222E-09 685 0.756E-05 0.139E-05
RMM: 14 0.766052678227E+50 0.76601E+50 0.10554E-08 701 0.598E-05 0.696E-06
RMM: 15 0.153210535645E+55 0.15320E+55 0.18425E-08 711 0.509E-05 0.362E-06
RMM: 16 0.306421071291E+59 0.30641E+59 0.19133E-08 747 0.418E-05 0.176E-06
RMM: 17 0.612842142582E+63 0.61281E+63 0.19257E-08 772 0.389E-05 0.903E-07
RMM: 18 0.122568428516E+68 0.12256E+68 0.19931E-08 780 0.383E-05 0.544E-07
force on displaced ion 1 direction 3 : -0.000 -0.000 -1.850


In the OUTCAR file it was shown that the PAW double counting will be extremely large
PAW double counting = 0.00000000******************
entropy T*S EENTRO = 0.00000000

I am wondering if there is any solution to this problem.

[abalone]

I found this problem was system dependent. I tested it on another cluster will a different compilation of vasp. There is no problem.

They are the same input file and same vasp file (version 5.3.3 purchased from MD).