On Jun 15, 2013, at 2:36 AM, FabrÃcio Bracht <fabracht1.gmail.com> wrote:
> Hello. I was just reading a thread about the use of terachem with amber and
> found some numbers comparing the performance of terachem to gaussian, for
> qm/mm simulations. The author of the thread described the performance
> difference as being 231.65sec using 1GPU for a 164 atoms (ATP,Mg2+, etc)
> 6-31g/rhf minimization, and 1321.2sec for gaussian running on 4 cores.
> I was wondering if anyone has updated numbers regarding the performance of
> terachem for qmmm simulations.
Even with GPUs the computational efficiency will be dominated by the QM part of the calculation, so pure reported timings for terachem on systems related to your quantum region will probably be a good order-of-magnitude estimate. The actual timings will of course depend on your computer and chemical system. However, the GPU performance variability is significantly lower than standard CPUs...
Getting timing comparisons to Gaussian will be difficult, seeing as publishing timing comparisons with other codes violates Gaussian's license agreement. Best case scenario you risk losing access to Gaussian from now on. Worst case, seek a lawyer ;).
HTH,
Jason
--
Jason M. Swails
Quantum Theory Project
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Jun 15 2013 - 14:30:02 PDT
