Hi Albert,
> it seems that for general biology system size (60K), most MD tools
> are with efficiency around 20 ns/day.... I thought that some could much
> faster.....
No that's pretty normal and when you actually do apples to apples
comparisons there tends not to be that much difference between the various
codes. Note for a 60K atom system a single GTX580 will likely get you around
20 to 25ns/day, you can probably get to 30 or so with a GTX680 and maybe 35
with multiple GPUs. So no need to spend millions on massive supercomputers.
;-)
I'd be wary of marketing type benchmarks which is what you often see touted
around. If you are seeing things going much much faster than 20ns/day then
likely they are cheating somehow, using hacks like H-mass repartitioning to
get artificially long time steps, or using very short cutoffs, or extremely
coarse FFT grids or being extremely lax on the precision of the math
involved etc. Essentially when looking at any benchmarks you need to look
very carefully at what the exact simulation parameters used were and use
your experience with MD to determine whether they are reasonable or not.
All the best
Ross
/\
\/
|\oss Walker
---------------------------------------------------------
| Assistant Research Professor |
| San Diego Supercomputer Center |
| Adjunct Assistant 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
Received on Thu Jul 05 2012 - 09:30:03 PDT
