Ceilidh ... The same input file produces different single point results on different computers

Firefly and PC GAMESS-related discussion club

The same input file produces different single point results on different computers

When trying to calculate the single point energy the different results are obtained. It seems that this situation arises when f basis functions are used.
For example, enegies of Ce3+ cation calculated on Core 2 Duo T6400 (Windows Vista 32 bit)and Core 2 Quad Q6600 (Windows XP 32 bit) processors are equal to -473.5549152219 and -473.5521347899 , respectively. Moreover, the discrepancy is observed at the first SCF step. Here are fragments of output:

Core 2 Duo
DIRECT SCF CALCULATION, SCHWRZ=T FDIFF=T
SCHWARZ INEQUALITY OVERHEAD: 2775 INTEGRALS, T= 0.27
NONZERO BLOCKS
ITER EX TOTAL ENERGY E CHANGE DENSITY CHANGE DIIS ERROR INTEGRALS SKIPPED
* * * INITIATING DIIS PROCEDURE * * *
1 0 -451.522662200 -451.522662200 2.701716401 4.500263113 5530192 72195
2 1 -451.121695620 0.400966580 1.314656872 3.831350192 6267931 68357

Core 2 Quad
DIRECT SCF CALCULATION, SCHWRZ=T FDIFF=T
SCHWARZ INEQUALITY OVERHEAD: 2775 INTEGRALS, T= 0.23
NONZERO BLOCKS
ITER EX TOTAL ENERGY E CHANGE DENSITY CHANGE DIIS ERROR INTEGRALS SKIPPED
* * * INITIATING DIIS PROCEDURE * * *
1 0 -451.520599268 -451.520599268 2.701730061 4.500297771 5530192 72195
2 1 -451.117962565 0.402636702 1.311269564 3.877987209 6274069 68321

What doest T value mean, which is different in this two cases?

Both tasks were in sequential mode. The same is true for parallel runs. The same discrepancy is observed for direct and conventional runs, for tasks with or without ECPs, for different functionals, etc.
For lighter atoms without f functions the energies are identical for two machines (water molecule, Na and Al cations were tested).

Input and output files are attached.

This message contains the 55 kb attachment
[ ce.zip ]

Fri Jan 22 '10 6:33pm

This message read 967 times