Top Page > Browsing

Memory Usage Date: 2006/11/03 21:22 Name: Kevin Hi, I'm trying to relax an oxide on a metal slab. The cell is ~ 5 x 5 x 20 Ang, 10 of which is a vacuum gap. The memory requirements for this seem to be huge! At the moment I only have 16 Gb of ram which is all used up straight away (Siesta uses about 6BG for the same calculation). I ran the memory leak test and all seems ok. Is there any way to limit the memory usage? maybe ignore points in the Vacuum? Thanks, Kevin

Page: [1]

Re: Memory Usage ( No.1 ) Date: 2006/11/04 01:32 Name: T.Ozaki Hi, It seems to be true that OpenMX requires a relatively large memory size, and this is an unsolved issue. However, if you are using a linear scaling solver, the memory requirement seems to be reasonable, since the Green functions have to be stored. Regards, TO Re: Memory Usage ( No.2 ) Date: 2006/11/04 06:41 Name: JessK Hi, I would like to add that I have the same problem using O(N) method. My system is tetragonal bulk, 8x8x32 Angstrom, 160 atoms, cutoff 250 Ry. I tried different O(N) methods but all of them crashed before first MD step. I used both EM64T Intel machine (24 CPUs, 4 GB/node memory) and SGI altix (64 CPUs, 64 Gb memory in total). Are there some parameters which can be tuned? Thanks, JK P.S. In fact, with SIESTA I could manage to run this job even using traditional diagonalization method on 12 CPUs.

Page: [1]