GTX 980 currently delivers roughly GTX780 performance at AMBER that I
expect to improve on by about 30% in the short-term.
On Fri, Jan 16, 2015 at 12:09 PM, Ross Walker <ross.rosswalker.co.uk> wrote:
> Hi Orion,
>
> The GTX780 is now end of line so you may have trouble purchasing a second
> one. It is fine to mix different GPUs in the same box, that's what I like
> so much about NVIDIA's unified drivers, but I would not recommend running a
> single calculation across two GPUs of different specs - this is untested
> and your mileage will vary. However, you can run two independent
> simulations at the same time, one on each GPU and these will run fine
> without interfering with each other.
>
> I'd recommend going with a GTX980 since that's what replaced the 780 - the
> underlying hardware (the chip) is very different though so I really would
> not recommend spanning it for a single calculation - you'll also need to
> recompile everything with CUDA 6.5 for the 980 to work. Your other option
> is to see if you can still get hold of a 780 - some places may still have
> them in stock. Exxact (mchen.exxactcorp.com) has a few still available I
> believe.
>
> One thing you should check is that your power supply is beefy enough for
> the two GPUs. I would say 1000W (silver) is probably the minimum and 1200W
> Gold or Platinum would probably be a better choice.
>
> Hope that helps.
>
> All the best
> Ross
>
> > On Jan 15, 2015, at 9:34 PM, orion shih <shih.orion.gmail.com> wrote:
> >
> > Hi,
> >
> > Our group is using single GTX780 accelerated pmemd and planning to
> purchase
> > a second GPU to speed up the simulation more. As for the buying choice,
> is
> > it recommended to buy another GTX780 or we can buy better ones if we want
> > to run single job across two GPUs? Since all benchmark data of multiple
> GPU
> > tests use the same GPUs I am not sure if two different GPUs can run in
> > parallel efficiently.
> >
> > Thanks for any help
> >
> >
> > Orion Shih, Ph.D.
> > Postdoctoral Associate
> > National Synchrotron Radiation Research Center
> > 101 Hsin-Ann Road, Room L204
> > Hsinchu 30076, Taiwan
> > Tel : +886-3-578-0281, ext-5204
> > E-mail:shih.orion.nsrrc.org.tw or shih.orion.gmail.com
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Jan 16 2015 - 12:30:04 PST
