Hi Mary,
8 GPUs is a lot to use you probably won't get optimal scaling unless you
have very good interconnect and only 1 GPU per node. Some things to try /
consider:
>|--------------------- INFORMATION ----------------------
>
>| GPU (CUDA) Version of PMEMD in use: NVIDIA GPU IN USE.
>
>| Version 12.0
>
>|
>
>| 03/19/2012
You should update your copy of AMBER since there have been many tweaks and
bug fixes. Do:
cd $AMBERHOME
./patch_amber.py --update
Run this until it stops saying there are updates (about 3 or 4 times). Then
make clean
./configure gnu
make
./configure -mpi gnu
make
./configure -cuda gnu
make
./configure -cuda -mpi gnu
make
>begin time read from input coords = 400.000 ps
>Number of triangulated 3-point waters found: 35215
>Sum of charges from parm topology file = -0.00000042
>Forcing neutrality...
This happens with the CPU code sometimes - often when the inpcrd / restart
file does not contain box information when a periodic simulation is
requested. Does it run ok with the CPU code? - Alternatively it may just
be running so slow over 8 GPUs that it hasn't even got to 500 steps yet to
print anything. Try it with just one GPU and see what happens.
All the best
Ross
/\
\/
|\oss Walker
---------------------------------------------------------
| Assistant Research Professor |
| San Diego Supercomputer Center |
| Adjunct Assistant Professor |
| Dept. of Chemistry and Biochemistry |
| University of California San Diego |
| NVIDIA Fellow |
|
http://www.rosswalker.co.uk |
http://www.wmd-lab.org |
| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
---------------------------------------------------------
Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Mar 14 2013 - 09:00:04 PDT
