| 5:16:50 PM PST - Wed, Nov 28th 2012 |
|
I compiled NWChem 6.1 on a xeon 8 cores cluster connected by a gigabit ethernet port. I used mpich2-1.3.2p1, l_fc_c_9.1.052 and l_mkl_p_10.0.5.025. The compilation was fine, however when I try to run a demanding calculation (in a single node/8cores), after a few time the CPUs go down and remain working with a very low % (below 10%), even they disappear. I am mounting the home directory by NTFS and I tried putting the scratch files in the node but a same behaviour is observed.
Any idea what is wrong?
Thanks.
|