Return to my home page
  Contents

Performance's analysis of Beowulf class supercomputers

Comparative of NAMD performance

Processors	1	2	4	8	16	32	64	128	160	192
Cray T3E
Time	x	6.12	3.10	1.60	0.810	0.397	0.212	0.123	0.098	x
Speedup	x	(1.97)	3.89	7.54	14.9	30.3	56.8	97.9	123	x
SGI O2000
Time	8.28	4.20	x	x	0.542	0.271	0.152	x	x	x
Speedup	1.0	1.96	x	x	15.3	30.5	54.3	x	x	x
ASCI-RED
Time	28.0	13.9	7.24	3.76	1.91	1.01	0.50	0.279	0.227	0.196
Speedup	1.0	2.01	3.87	7.45	14.7	27.9	56.0	100	123	143
HP735/125 (ATM)
Time	24.1	12.4	6.39	3.69	x	x	x	x	x	x
Speedup	1.0	1.94	3.77	6.54	x	x	x	x	x	x
Cluster Linux PII (fast Ethernet)
Time	12.34	6.42	3.28	1.69	1.01	0.62	x	x	x	x
Speedup	1.0	1.92	3.77	7.30	12.21	19.90	x	x	x	x

Execution time in seconds for a system of 36573 atoms, with a cut-off of 12A.

Conclusions:

• Scalability of results. Fairly good, but could be best using a Gb net, better than Fast Ethernet.
• Comparation between the result of PII table and above table. It looks that you need the double of processors to reach the same performance, than a O2000, and the same processors to reach a T3E.
• SGI O2000/Linux ratio; Linux Cluster with the double of processors.

  Time: for 1 procs in O2000: 1.28; for 2 procs in O2000: 1.28; for 16 procs in O2000: 0.87.

  Excellent! Beat O2000 with se same prize!

• Cray T3E/Linux ratio; Linux Cluster with the same number of processors

  Time: for 2 procs in T3E : 0.95; for 4 procs in T3E : 0.94; for 8 procs in T3E : 0.94; for 16 procs in T3E : 0.8; for 32 procs in T3E : 0.68

  SIMPLY MARVELOUS!!!

Return to my home page
  Contents