Hi, I'm trying to understand how best to compare performance of Gromacs 4.6.1 on CPU only versus CPU plus GPU on a cluster. Each node has is 2 x 8 Core Sandy Bridge + 1 Nvidia M2090.
I've been trying to use the d.dppc benchmark (with modifications). I would like to profile both standard cutoff and pme electrostatics on the CPU and GPU (as this is building on benchmarking work done on the CPU only which used these benchmarks). While I have successfully run modifications of the d.dppc benchmark utilising both the CPU and GPU I have some questions about the different options. As a beginner with MD and Gromacs could somebody elaborate on the table comparing Group and Verlet schemes on this page: http://www.gromacs.org/Documentation/Cut-off_schemes In particular the table shows which non-bonded interaction features are supported by both the group and verlet schemes. In the table we see four different 'cutoff' non-bonded interaction features listed: unbuffered cut-off scheme buffered cut-off scheme exact cut-off cut-off If plain 'cut-off' is not one of the three more specific schemes listed above it, how does it differ from each of them? The d.dppc input file as provided is: title = dppc cpp = /lib/cpp integrator = md nsteps = 5000 nstlist = 10 nstxout = 0 nstvout = 0 nstxtcout = 0 nstlog = 0 dt = 0.002 constraints = all-bonds nstenergy = 0 ns_type = grid coulombtype = cut-off rlist = 1.0 rvdw = 1.0 rcoulomb = 1.8 tcoupl = berendsen tc_grps = dppc sol tau_t = 0.1 0.1 ref_t = 323 323 Pcoupl = no Could you say which of the four cutoff types mentioned above these settings equate to and how these setting should be changed to equate to each of the other three cutoff types? -- View this message in context: http://gromacs.5086.x6.nabble.com/Different-Cutoff-Options-tp5008455.html Sent from the GROMACS Users Forum mailing list archive at Nabble.com. -- gmx-users mailing list gmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
