help to solve P4-error for vasp question

[bendix]

When running some cases like NEB-case, I often meet the below error!
Try to solve this error about one month, but i don't success until to now!
The errors let to me crazy...
Can anyone help me, please?

my lib-machines
compiler:IFC 9.1.040
lib.:MKL 9.0
CPU: Intel core 2 duo E6600
distr.:Fedora core 6
MPI: mpi-1.2.7p1 + mpiexec-0.82
PBS: torque


error information:
17624: p4_error: interrupt SIGx: 15
p0_17624: (63417.773438) net_send: could not write to fd=4, errno = 32
p3_24409: p4_error: interrupt SIGx: 13
rm_l_1_17627: p4_error: interrupt SIGx: 15
rm_l_1_17627: (63415.160156) net_send: could not write to fd=5, errno = 32
p3_24409: (63419.808594) net_send: could not write to fd=5, errno = 32
mpiexec: Warning: task 0 exited with status 1.
mpiexec: Warning: task 2 died with signal 1936280415 (Unknown signal 1936280415).
mpiexec: Warning: task 3 exited with status 116.

[admin]

sorry, this error is not due to vasp but probably due to some
communication problem. Please in any case make sure that VASP has been compiled with the same integer byte-length as your MPI .

[bendix]

I find about this question recently.
When used ISPIN=2 in NEB-case, this error exhibits right now.
Just close spin(ISPIN=1 or comment ISPIN), VASP can run NEB-case.
I try this parameter (ISPIN) of NEB in the two different clusters, all machines have the same error!

why? Please tell me!
thanks~~

I use the parameters in INCAR, below...
SYSTEM=adsorption process
ISTART=0
ICHARG=2
ISIF=2
IBRION=1
NPAR=8
ISMEAR=0
SIGMA=0.1
LREAL=Auto
ALGO=Fast
LPLANE=.TRUE.
NSIM=4
NSW=30
PREC=Accurate
IMAGES=4
SPRING=-5
ISPIN=2
<span class='smallblacktext'>[ Edited Sat Jun 30 2007, 09:56AM ]</span>

[admin]

sorry, I cannot reproduce this error. Using LAM and mpirun, there is no difference whether a NEB calculation is done for a magnetic or a non-polarized system.