Differences

This shows you the differences between two versions of the page.

--- cluster:175 [2018/09/25 12:39]
hmeij07 [Lammps]
+++ cluster:175 [2018/09/26 12:35]
hmeij07 [Lammps]
@@ Line 70: / Line 70: @@
 </code>
-== Lammps (PMMA) ==
+==== Lammps (PMMA) ====
-Material called PMMA (https://en.wikipedia.org/wiki/Poly(methyl_methacrylate). The PMMA simulations require the calculation of molecular bonds, which is not implemented in GPU hence more CPU cores are required than the Coillod example.
+Using a material called PMMA (https://en.wikipedia.org/wiki/Poly(methyl_methacrylate) aka acrylic glass or plexiglas ("safety glass"). The PMMA simulations require the calculation of molecular bonds, which is not implemented in GPU hence more CPU cores are required than the Coillod example. The optimal ratio cpu:gpu appears to be 4-6:1.
-a) 1GPU with X CPUs
-#cpus   ns/day
+^  gpu  ^  cpus  ^  ns/day  ^  quad  ^  ns/day/node  ^
-           89.6
+|  1 P100  |  4  |  89  |  x4  |  356  |
-           61.6
+|  1 GTX  |  6  |  90  |  x4  |  360  |
-           34.3
+|  1 K20  |  6  |  47  |  x4  |  188  |
-b) 4 GPUs with 4 CPUs
- ns/day
-----------
-compare these results to
- K20 GPU + 4 CPUs  37 ns/day
-K20 GPU + 6 CPUs  47 ns/day
-GTX GPU + 4 CPUs  73 ns/day
- GTX GPU + 6 CPUs  90 ns/day
+That means the P100 works as well as the GTX. The K20 works at 50% the performance level of the others implying the cpu calculations are the bottleneck.

DokuWiki