Re: [AMBER] Which GPU is recommended?

From: Alexander Metz <alexander.metz.pharmazie.uni-marburg.de>
Date: Wed, 22 Oct 2014 11:54:31 +0100

Hi Ross,

it seems the linked GTX780 model (EVGA GeForce GTX780 SuperClocked 3GB) is mostly out of stock.
Therefore, I would like to ask how important the choice of this particular model is.
Which other GTX780 models could be recommended,
or are there certain features to look for or avoid?

Thanks,


Alex
 

+++++
Alexander Metz
Institute of Pharmaceutical Chemistry
Philipps-University Marburg
Marbacher Weg 6, 35032 Marburg, Germany
Tel.: + 49 6421 282 5930
E-mail: alexander.metz.pharmazie.uni-marburg.de


+++++



________________________________

Hi Qiao, Please see the following page for info on GPU accelerated AMBER - this
includes a list of supported hardware (there are also benchmarks available
for the various GPU models). http://ambermd.org/gpus/ and http://ambermd.org/gpus/recommended_hardware.htm#hardware At the bottom of this page is a shopping list for Amazon should you wish
to build your own system. In your case your best option being budget conscious is a GTX780 (or GTX980 once we release an update with support for it - in a few weeks). I
do NOT recommend the GTX780TI models since experience shows that they can
be unreliable (the CAS memory refresh rate is too high). The GeForce cards
tend to be okay for extended use - if you plan on putting them in a
cluster though you might want to consider purchasing complete nodes
through a vendor that supports GeForce cards in production systems - For
example the systems offered by Exxact on the above page - which then
include a 3 year warranty - otherwise you just have the standard component
warranty on the GPU itself (1 year?). Hope that helps, All the best
Ross On 10/16/14, 10:29 PM, "xueqiaoup.gmail.com" <xueqiaoup.gmail.com> wrote: >Hi,
>
>I want to buy GPU to perform MD simulations. Because of the budget, I
>would choose one among tesla M2090(single-precision/double-precision
>1331G/665G), tesla K10(4577G/190G). GTX780Ti(5.04T/210G),
>GTX780(4T/1.3T). For MD simulations, which parameter is more important?
>the single-precision floating-point or the double-precision
>floating-point? Besides, I have no idea whether the normal GPU for PC
>like GTX780 or TITAN can work for very long time in HPC cluster. In the
>end, which GPU above is more suitable?
>Thanks!
>
>Qiao Xue >
>
>
>xueqiaoup.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 Wed Oct 22 2014 - 04:00:03 PDT
Custom Search