This shows you the differences between two versions of the page.
Both sides previous revision Previous revision Next revision | Previous revision Last revision Both sides next revision | ||
cluster:109 [2013/02/02 13:46] hmeij [Lammps GPU Testing (EC)] |
cluster:109 [2013/02/02 15:30] hmeij [Lammps GPU Testing (EC)] |
||
---|---|---|---|
Line 4: | Line 4: | ||
===== Lammps GPU Testing (EC) ===== | ===== Lammps GPU Testing (EC) ===== | ||
- | * 32 cores | + | * 32 cores E2660 |
* 4 K20 GPU | * 4 K20 GPU | ||
* workstation | * workstation | ||
Line 10: | Line 10: | ||
- | Same tests using lj/cut, eam, lj/expand, and morse: AU.reduced | + | Same tests (12 cpu cores) |
- | CPU only 9 mins 10 secs | + | CPU only 6 mins 1 secs |
1 GPU 1 mins 1 secs (a 5-6 times speed up) | 1 GPU 1 mins 1 secs (a 5-6 times speed up) | ||
- | 2 GPUs 0 mins 46 secs (never saw 2nd GPU used, problem set too small?) | + | 2 GPUs 1 mins 0 secs (never saw 2nd GPU used, problem set too small?) |
- | Same tests using a restart file and using gayberne: GB | + | Same tests (12 cpu cores) |
CPU only 1 hour 5 mins | CPU only 1 hour 5 mins | ||
- | 1 GPU 3 mins and 33 secs (a 18-19 times peed up) | + | 1 GPU 5 mins and 15 secs (a 18-19 times peed up) |
- | 2 GPUs 2 mins (see below) | + | 2 GPUs 2 mins |
+ | |||
+ | Above results seems overall a bit slower that at other vendor, but same pattern. | ||
+ | |||
+ | Francis' | ||
+ | |||
+ | ^3d Lennard-Jones melt: for 10,000 steps with 32,000 atoms^^^^^^ | ||
+ | |CPU only| -np 1 | -np 6 | -np 12 | -np 24 | -np 36 | | ||
+ | |loop times| | ||
+ | |GPU only| 1xK20 | 2xK20 | 3xK20 | 4xK20 | | ||
+ | |loop times| | ||
+ | ^3d Lennard-Jones melt: for 100,000 steps with 32,000 atoms^^^^^^ | ||
+ | |GPU only| 1xK20 | 2xK20 | 3xK20 | 4xK20 | (-np 1-4) | | ||
+ | |loop times| | ||
+ | |||
+ | * Serial' | ||
+ | * GPU's serial time matches MPI -np 24 and can be further reduced to 10s, a 3x speed up | ||
===== Lammps GPU Testing (MW) ===== | ===== Lammps GPU Testing (MW) ===== |