DokuWiki

User Tools

Site Tools

Trace:
cluster:175

Differences

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

Link to this comparison view

Both sides previous revision Previous revision
Next revision 		Previous revision
Next revision Both sides next revision
cluster:175 [2018/09/25 12:35]
hmeij07 		cluster:175 [2018/09/26 20:08]
hmeij07 [Gromacs 2018.3]
Line 70: Line 70:
 </code> </code>
  
-=== WZ ===+==== Lammps (PMMA) ====
  
-a) 1GPU with X CPUs+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.
    
-#cpus   ns/day +^  gpu  ^  cpus  ^  ns/day  ^  quad  ^  ns/day/node  ^ 
-4           89.6 +|  1 P100  |  4  |  89  |  x4  |  356  | 
-2           61.6 +|  1 GTX  |  6  |  90  |  x4  |  360  | 
-1           34.3+|  1 K20  |  6  |  47  |  x4  |  188  |
  
-b) 4 GPUs with 4 CPUs +That means the P100 works as well as the GTX. The K20 works at 50% the performance level of the others which is impressive for this old gpu.
-92 ns/day+
  
----------- 
-compare these results to  
  
-1 K20 GPU + 4 CPUs  37 ns/day 
-1 K20 GPU + 6 CPUs  47 ns/day 
-1 GTX GPU + 4 CPUs  73 ns/day 
-1 GTX GPU + 6 CPUs  90 ns/day 
  
----- 
-Test dir location 
-/home/heme/work/lmp-wz 
- 
- 
--------------- 
-I am running a material called PMMA (https://en.wikipedia.org/wiki/Poly(methyl_methacrylate) ) 
- 
-The reason for the different benchmark than yours is that the PMMA simulations require the calculation of molecular bonds, which is not implemented in GPU.  
  
 ==== Gromacs ==== ==== Gromacs ====
Line 134: Line 118:
  
 </code> </code>
 +
 +==== Gromacs 2018.3 ====
 +
 +The multidir not running in Gromacs 2018 is a bug in the code clashing with the call MPI_Barrier (communication timing error).  It is fixed in Gromacs 2018.3 so we have some multidir results although gpu utilization has room for improvements (about 25% used, same as on GTX).
 +
 +<code>
 +
 +# multidir -gpu_id 0123 with four simultaneous gromacs processes
 +
 +-np  8 -ntomp  4 -npme 1 -maxh 0.1 -pin on -nb gpu
 +01/md.log:Performance:       36.692        0.654
 +02/md.log:Performance:       36.650        0.655
 +03/md.log:Performance:       36.623        0.655
 +04/md.log:Performance:       36.663        0.655 or about 146 ns/day/node (quad)
 +
 +-np 16 -ntomp  8 -npme 1 -maxh 0.1 -pin on -nb gpu
 +01/md.log:Performance:       25.151        0.954
 +02/md.log:Performance:       25.257        0.950
 +03/md.log:Performance:       25.247        0.951
 +04/md.log:Performance:       25.345        0.947 or about 100 ns/day/node (quad)
 +
 +multidir -gpu_id 00112233 with eight simultaneous gromacs processes
 +sharing the gpus, 2 processes per gpu
 +
 +-np  8 -ntomp  4 -npme 1 -maxh 0.1 -pin on -nb gpu
 +Error in user input:
 +The string of available GPU device IDs '00112233' may not contain duplicate
 +device IDs
 +
 +</code>
 +
 +That last error when loading multiple processes per gpu is *not* according to their documentation.
 +
  
 ==== What to Buy ==== ==== What to Buy ====
cluster/175.txt · Last modified: 2018/11/29 18:00 by hmeij07

Page Tools

Except where otherwise noted, content on this wiki is licensed under the following license: CC Attribution-Share Alike 4.0 International
CC Attribution-Share Alike 4.0 International Donate Powered by PHP Valid HTML5 Valid CSS Driven by DokuWiki