Yes, that has been true for GROMACS for a few years. Low-latency
communication is essential if you want a whole MD step to happen in around
1ms wall time.
Mark
On Nov 5, 2013 11:24 PM, "Dwey Kauffman" wrote:
> Hi Szilard,
>
> Thanks.
>
> >From Timo's benchmark,
> 1 node142 ns/day
> 2
Hi Szilárd and all,
Thanks very much for the information. I am more interested in getting
single simulations to go as fast as possible (within reason!) rather than
overall throughput. Would you expect that the more expensive dual
Xeon/Titan systems would perform better in this respect?
Cheers
Hi Szilard,
Thanks.
>From Timo's benchmark,
1 node142 ns/day
2 nodes FDR14 218 ns/day
4 nodes FDR14 257 ns/day
8 nodes FDR14 326 ns/day
It looks like a infiniband network is "required" in order to scale up when
running a task across nodes. Is it correct ?
Dwey
--
View t
On Tue, Nov 5, 2013 at 9:55 PM, Dwey Kauffman wrote:
> Hi Timo,
>
> Can you provide a benchmark with "1" Xeon E5-2680 with "1" Nvidia
> k20x GPGPU on the same test of 29420 atoms ?
>
> Are these two GPU cards (within the same node) connected by a SLI (Scalable
> Link Interface) ?
Note that
Hi Timo,
Can you provide a benchmark with "1" Xeon E5-2680 with "1" Nvidia
k20x GPGPU on the same test of 29420 atoms ?
Are these two GPU cards (within the same node) connected by a SLI (Scalable
Link Interface) ?
Thanks,
Dwey
--
View this message in context:
http://gromacs.5086.x6.nab
Timo,
Have you used the default settings, that is one rank/GPU? If that is
the case, you may want to try using multiple ranks per GPU, this can
often help when you have >4-6 cores/GPU. Separate PME ranks are not
switched on by default with GPUs, have you tried using any?
Cheers,
--
Szilárd Páll
29420 Atoms with a some tuning of the write out and communication intervals:
nodes again: 2 x Xeon E5-2680v2 + 2 x NVIDIA K20X GPGPUs @ 4fs vsites
1 node 212 ns/day
2 nodes 295 ns/day
--
gmx-users mailing listgmx-users@gromacs.org
http://lists.gromacs.org/mailman/listinfo/gmx-users
* Please
Brad,
These numbers seems rather low for a standard simulation setup! Did
you use a particularly long cut-off or short time-step?
Cheers,
--
Szilárd Páll
On Fri, Nov 1, 2013 at 6:30 PM, Brad Van Oosten wrote:
> Im not sure on the prices of these systems any more, they are getting dated
> so th
just a small benchmark...
each node - 2 x Xeon E5-2680v2 + 2 x NVIDIA K20X GPGPUs
42827 atoms - vsites - 4fs
1 node142 ns/day 2 nodes FDR14 218 ns/day
4 nodes FDR14 257 ns/day
8 nodes FDR14 326 ns/day
16 nodes FDR14 391 ns/day (global warming)
best,
timo
--
gmx-users mailing list
Have you used/considered any cloud approaches?
--
View this message in context:
http://gromacs.5086.x6.nabble.com/Hardware-for-best-gromacs-performance-tp5012124p5012154.html
Sent from the GROMACS Users Forum mailing list archive at Nabble.com.
--
gmx-users mailing listgmx-users@gromacs.org
Im not sure on the prices of these systems any more, they are getting dated
so they will be on the low end price wise. I have a 30,000 ish atom lipid
system for all my simulations so this might be helpful:
System 1
CPU - dual 6 core xeon @ 2.8 GHz
GPU - 2x GTX 680
50 ns/day
System 2
CPU - dual 4
11 matches
Mail list logo