Hello MS users,
i noticed that, when running a Geometry Optimization for a methane molecule with DMol³, the resulting structure differs between serial(1 core) and parallel(8 core) calculations.
Parallel runs return exact angles of 109.471°, serial runs show slightly different angles. However, the last structure in the parallel-trajectorie is exactly the same as the result from the serial run.
This happens with GGA-PW91, GGA-BP and GGA-LDYP, however not with LDA methods.
Why do, on parallel runs, the last frame from the xtd file and the xsd file differ, while they are the same on serial runs?
And why do parallel runs differ from serial ones at all?
I used default 'fine' settings for all tests.
thanks in advance, Martin
i noticed that, when running a Geometry Optimization for a methane molecule with DMol³, the resulting structure differs between serial(1 core) and parallel(8 core) calculations.
Parallel runs return exact angles of 109.471°, serial runs show slightly different angles. However, the last structure in the parallel-trajectorie is exactly the same as the result from the serial run.
This happens with GGA-PW91, GGA-BP and GGA-LDYP, however not with LDA methods.
Why do, on parallel runs, the last frame from the xtd file and the xsd file differ, while they are the same on serial runs?
And why do parallel runs differ from serial ones at all?
I used default 'fine' settings for all tests.
thanks in advance, Martin
