# direct vs. in-core calculations

*From*: Kevin Moore <kmoore.,at,.ncsc.org>
*Subject*: direct vs. in-core calculations
*Date*: Thu, 29 Apr 93 11:08:37 EDT

>Richard,
>
>I beg to differ with your numbers. Using N^4/4 one calculates:
>
># functions memory
>----------- ------
> 100 -> 25 MW
> 200 -> 400 MW
> 300 -> 2025 MW
>
>However, the Gaussian manual says, and my own experience support it,
>that for closed shell RHF calculations memory scales as N^4/8 plus
>some additional memory, say 500,000 words. (G92 User's Guide p.88.)
>My own experience has been that the additional .5 MW is too small,
>a more realistic number is 100N*N. Using this formula one calculates:
>
># functions memory
>----------- ------
> 100 -> 12.5 + 1 = 13.5 MW
> 200 -> 200 + 4 = 204 MW
> 300 -> 1012.5 + 9 = 1021.5 MW
It has been a while since I have sat down and worked this out (so someone
can correct me if I am wrong), but I believe the best approximation is:
N^4/8 + N^3/4 + N^2/2
This means that the table would be:
# functions memory
----------- ------
100 -> 12.8 Mw
200 -> 202.0 Mw
300 -> 1019.3 Mw
Generally, I use N^4/8 + N^3/4 and then add about 0.5 Mw onto the end. In
gaussian, you can run the calculation with 'scf=incore' and if you have
not given it enough memory, it will fail at the beginning of the scf. You
can then look at the output file and see how much memory is needed (the
error includes a line saying how many words are needed). Since it is at
the beginning of the scf, the cpu time loss is minimal.
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
|| Kevin Moore North Carolina Supercomputing Center ||
|| Scientific Support Analyst 3021 Cornwallis Rd. ||
|| (919) 248-1179 Research Triangle Park, NC 27709 ||
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++