[lammps-users] LAAMPS benchmarks

We have done a head to head benchmark comparison of the newest AMD and Intel processors. We would like to share the results with the forum as they are pretty interesting.

Here is the link: http://www.advancedclustering.com/company-blog/molecular-dynamics-amd-vs-intel.html

Let us know if you have any questions, or have a more relevant benchmark that you would like to see.


your benchmarks do not contain any tests running across multiple nodes.
this is the main application of LAMMPS and thus your tests are not of
much use. also, you don't comment on important things like
memory/processor affinity settings etc.
contrary to your findings, i see an ca. 10% advantage of using LAMMPS
with hyperthreading on a single node using X5550 nehalems (not that it
matters much).

finally, when using intel compilers, enabling ansi-aliasing for files
that do not violate the ansi-c aliasing rules is also a significant
optimization, using MKL's fftw wrapper not. i generally found it
detrimental to the performance, compared to a properly compiled fftw2.