That should work magnificently well.
On May 18, 2017 2:47 PM, "Adrian Roitberg" <roitberg.ufl.edu> wrote:
> Thanks!
>
> Not quite what I needed, but it helps.
>
> Exact case is:
>
> State of the art new iMac, and they want to tun amber in gpus. So, the
> only option is to buy an external box, and go through thunderbolt.
>
> thanks
>
>
>
>
> On 5/18/17 5:40 PM, Ross Walker wrote:
> > What do you mean by 'external'?
> >
> > Do you mean sitting in a breakout box with a PCI-e cable going to the
> motherboard of the server? If yes that works fine - performance is
> identical since it's logically equivalent to having the GPUs plugged into
> the physical PCI-e slots. Multi-GPU as long as they can do P2P and sit on
> the same PLX switch should also be the same. That said the connections for
> these break out boxes can be super fragile. Plug them in in the rack, get
> it working reliably and then NEVER touch anything in the vicinity again is
> my advice. If it's in a rack people keep accessing, recabling etc it will
> have you pulling your hair out.
> >
> > If you mean something entirely different then the answer is I've never
> tried it but would be interested to understand what you are referring to.
> >
> > All the best
> > Ross
> >
> >> On May 18, 2017, at 5:23 PM, Adrian Roitberg <roitberg.ufl.edu> wrote:
> >>
> >> Hi there
> >>
> >> Without answering with: "Why would you want to do do that???"
> >>
> >> Has anyone played with running amber on a server but using and external
> >> GPUs ? I would be interested in any results you can share, be it price,
> >> how hard it was to do, and performance.
> >>
> >> Thanks
> >>
> >> adrian
> >>
> >>
> >>
> >> --
> >> Dr. Adrian E. Roitberg
> >> University of Florida Research Foundation Professor
> >> Department of Chemistry
> >> University of Florida
> >> roitberg.ufl.edu
> >> 352-392-6972
> >>
> >>
> >> _______________________________________________
> >> AMBER-Developers mailing list
> >> AMBER-Developers.ambermd.org
> >> http://lists.ambermd.org/mailman/listinfo/amber-developers
> >
> > _______________________________________________
> > AMBER-Developers mailing list
> > AMBER-Developers.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber-developers
>
> --
> Dr. Adrian E. Roitberg
> University of Florida Research Foundation Professor
> Department of Chemistry
> University of Florida
> roitberg.ufl.edu
> 352-392-6972
>
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Thu May 18 2017 - 15:00:04 PDT
