Running Parallel Gaussian Jobs

Much work went into benchmarking and trying to understand Gaussian in the past 6 months. It is by far the code which uses the majority of the CPU cycles in the HPCC. One should note that MOST Gaussian routines scale very poorly as the number of processors increase. In fact it is not uncommon to have a Gaussian job run SLOWER than the single processor time as more processors are requested. For this reason Gaussian users are HIGHLY encouraged to verify the speedup and efficient use of the system running Gaussian. See http://www.nd.edu/~rich/Gaussian for additional information. Please don't blindly assume that more processor equals more speed.

(Back to HPCC Batch Main Page)

Modified 12/12/02