Alex Granovsky
gran@classic.chem.msu.su
running Firefly in extended XP mode with eight cores per each XP group and assuming 50 atomic molecule and single pass numgrad run you will
need 50*6 + 1 eight-core compute boxes i.e. 301*8 cores.
This is doable but very time-consuming. Look at the
http://pubs.acs.org/doi/abs/10.1021/ct400647w
Best regards,
Alex
On Thu Feb 18 '16 2:16pm, GrEv wrote
------------------------------------
>Dear Alex,
>Many thanks for your reply.
>So, in principle it should be possible to do a geometry optimization for large systems (50-100 atoms) using numerical gradients with the extended XP mode if enough CPU-cores are available (to evaluate all the gradients at one bout). Otherwise there shouldn't be other obstacles, should they?
>Best regards,
>Evgeniy
>On Wed Feb 17 '16 1:37am, Alex Granovsky wrote
>----------------------------------------------
>>Hi,
>>see
>>http://classic.chem.msu.su/gran/gamess/numgrad.htmlhttp://classic.chem.msu.su/gran/gamess/numgrad.htmlhttp://classic.chem.msu.su/gran/gamess/numgrad.htmlhttp://classic.chem.msu.su/gran/gamess/numgrad.html
>>http://classic.chem.msu.su/gran/gamess/xpmodes.htmlhttp://classic.chem.msu.su/gran/gamess/xpmodes.htmlhttp://classic.chem.msu.su/gran/gamess/xpmodes.htmlhttp://classic.chem.msu.su/gran/gamess/xpmodes.html
>>Kind Regards,
>>Alex Granovsky
>>
>>
>>
>>On Mon Feb 15 '16 2:25pm, GrEv wrote
>>------------------------------------
>>>Dear All,
>>>I am interested in numerical geometry optimization with FF at the XMCQDPT2 level.
>>>I wonder how it is arranged, in particular whether it is parallelized with respect to all distortions needed to evaluate the gradients.
>>>Thanks!
>>>Best regards,
>>>Evgeniy