Firefly and PC GAMESS-related discussion club


 
Learn how to ask questions correctly  
 
 
We are NATO-free zone
 



Re^4: How to speed up the HESSIAN calculations

Siddheshwar Chopra
sidhusai@gmail.com


Dear Luca,
I get your point.. Yes you said it right...I can try NVIB=1.. But the same doubt I had before giving it a try....Will it effect the frequency data drastically?? I will have to check both.. Also the manual says, the change would be 10-100 wavenumbers.. Also Luca you have mentioned that "Run the job on more CPU cores"...Could you please have a look at the SAMPLE.out I have attached and let me know if I am using the full capability of my system or not? As far as I know, on my dual core system, I specify -np 2...and on my i3 processor system,  I use -np 4. Is it fine?? Also, by lowering the accuracy, do you mean I should relax the OPTOL from 10E-7 (in SAMPLE.out) to a higher value??? Then I guess, I might not get the global minimum...

Regards,

On Tue Jan 8 '13 6:07pm, Luca Maidich wrote
-------------------------------------------
>Dear Siddheshwar,
>as Thomas said already there aren't many ways of speeding up hessian evaluation in your case. Just to add other options you can:
>- Change to NVIB=1 in $FORCE to halve the number of single point energy calculations (from 6N+1 to 3N+1) with only a slight difference in the frequencies, but I suggest you to check yourself if the difference is not dramatic by running both calculations, i.e. with NVIB=1 and NVIB=2.
>- Run the job on more CPU cores even if this is not something connected to your input file.
>Other options, to the best of my knowledge, require lowering the accuracy as Thom pointed out...

>Kind Regards
>Luca


[ Previous ] [ Next ] [ Index ]           Wed Jan 9 '13 11:50am
[ Reply ] [ Edit ] [ Delete ]           This message read 679 times