Firefly and PC GAMESS-related discussion club

 
Learn how to ask questions correctly  
 
 

 

Alex Granovsky
gran@classic.chem.msu.su

we have analyzed this question some time ago and our conclusion
was that it would make sense for standard summation of PT series
but does not make any sense for Resolvent fitting approach to
summation, and that the latter is faster when running on standard
CPUs.

Kind regards,
Alex Granovsky

P.S. Note, GTX 6xx are typically slower than GTX 4xx
on double precision math so this does not worth efforts at all.


On Mon Jan 28 '13 3:15pm, Roman Kroik wrote
-------------------------------------------
>Hi! I have some questions about CUDA and XMCQDPT2 calculations. Is it reasonable to use XMCQDPT2 code on the nVidia GPUs through CUDA in order to speed up calculations? And what about using the newest nVidia video cards, i.e. GTX 6xx, for CUDA calculations, is it possible?