Cheers Ross. I missed it on the "supported GPUs" table, I was looking at
the max system sizes table which stops at K40.
If anyone needs me to test system-size limits on a K80 and confirm that it
is the same as a K40, I'll be able to do that in a few weeks when the K80s
that the local systems biology gang have ordered show up, we use the same
cluster.
Josh
On 9 November 2015 at 22:09, Ross Walker <rosscwalker.gmail.com> wrote:
>
> > On Nov 9, 2015, at 12:43 PM, Josh Berryman <
> the.real.josh.berryman.gmail.com> wrote:
> >
> > "I would like to get some inputs from somebody with some experience with
> > drugs and MD."
> >
> > Big mistake. Might be OK just looking at the simulations, but don't get
> > high and start messing around with an expensive shared supercomputer.
>
> ;-)
>
> > Of course you could get a personal GPU in which case you are hurting
> noone,
> > 5 million atoms should fit on a Tesla K40, a K80 is twice as big although
> > it hasn't made it onto the "supported cards" list here:
> > http://ambermd.org/gpus/#supported_gpus . I'd be tempted to have a go
> at
> > that myself, it sounds like a fun calculation. Off-topic: a K80 looks
> like
> > it is just two K40s in a box, does anyone know if you can use it OK as a
> > single card?
> >
>
> Huh? - From that page:
>
> Hardware Version 3.0 / 3.5 (Kepler I / Kepler II)
> • Tesla K20/K20X/K40/K80
>
> Says K80 there and benchmarks are on the GPU benchmark page for K80. A K80
> is essentially two K40s on the same card hooked together with PCI-E switch
> and then down clocked a little to fit within the 300W power envelope of a
> PCI-E card. They work very well although cost a LOT! - Titan-X or 980TI is
> probably more cost effective and has 12GB so can handle bigger simulations
> than K40 or K80 if that is a concern.
>
> The K80 will appear to the OS as two distinctly different GPUs. So having
> 4 in a box is actually just like having 8 GPUs in a box. On a single K80
> you can run 2 x 1 GPU jobs or 1 x 2 GPU job - dealers choice.
>
> All the best
> Ross
>
> /\
> \/
> |\oss Walker
>
> ---------------------------------------------------------
> | Associate Research Professor |
> | San Diego Supercomputer Center |
> | Adjunct Associate Professor |
> | Dept. of Chemistry and Biochemistry |
> | University of California San Diego |
> | NVIDIA Fellow |
> | http://www.rosswalker.co.uk | http://www.wmd-lab.org |
> | Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
> ---------------------------------------------------------
>
> Note: Electronic Mail is not secure, has no guarantee of delivery, may not
> be read every day, and should not be used for urgent or sensitive issues.
>
>
> _______________________________________________
> 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 Mon Nov 09 2015 - 23:30:03 PST
