Next: The different Supercomputers Up: Performance of Supercomputers Previous: The LINPACK Benchmark

Analysis of the TOP500

In Table 3.3 we have sorted all different performance types showing up in the TOP500 and include the number of their appearances. Figure 3.2 shows individual as well as accumulated performances with respect to the positon in the TOP500. Figure 3.3 shows the same data for the first hundred entries only. The first 10 systems already contribute [l]First 10 Systems: 364 Gflop/s 364 Gflop/s [tex2html_wrap3904] (24.8%) to the list. The first 42 systems achieve [l]First 42 Systems: 733 Gflop/s half of the total performance in the TOP500! On the contrary, the last 100 [l]Last 100 Systems: 58 Gflop/ssystems contribute only 58 Gflop/s (4%).

Unfortunatly the [tex2html_wrap3906] values are not complete in our list. Looking at them in Figure 3.4 you can see that the values of [tex2html_wrap3908] are quite widespread. Low values are typical for vector machines where MPPs often show quite high ones. This indicates that you have to run bigger problems on MPPs to achieve good efficiencies. Figure 3.5 shows [tex2html_wrap3910] values. The difference between vector machines and MPPs is not so big, nevertheless MPPs tend to have higher [tex2html_wrap3912] values, which shows that they often have bigger memory. This is due to the fact, that their DRAM memory chips are cheaper than the SRAM memory chips of the vector machines.

[figure441]

[figure446]

[figure451]

[figure456]

[fullpage461]

References

8
Reinhold P. Weicker, A detailed look at some popular benchmarks, Parallel Computing, 17, 10&11, 1153-1172, 1991

9
Kaivalya M. Dixit, The SPEC benchmark, Parallel Computing, 17, 10&11, 1195-1209, 1991

10
M. Berry, D. Chen, P. Koss, D. Kuck, S. Lo, Y. Pang, L. Pointer, R. Roloff, A. Sameh, E. Clementi, S. Chin, D. Schneider, G. Fox, P. Messina, D. Walker, C. Hsiung, J. Schwarzmeier, K. Lue, S. Orszag, F. Seidl, O. Johnson, R. Goodrum and J. Martin, The PERFECT Club benchmarks: effective performance evaluation of computers, Intl. J. Supercomputer Appls., 3, 3, 1989, 5-40

11
F. H. McMahon, The Livermore Fortran Kernels test of the numerical performance range, in: J. L. Martin, Performance Evaluation of Supercomputers, Elsevier Science B.V., North-Holland, Amsterdam, 1988, 143-186

12
D. H. Bailey, J. Barton, T. Lasinski, and H. D. Simon (editors), The NAS parallel benchmarks, Technical Report 103863, NASA Ames Research Center, Moffett Field, CA 94035, July 1993

13
D. H. Bailey, E. Barszcz, L. Dagum, and H. D. Simon, NAS parallel benchmarks results 3-94, RNR Technical Report RNR-94-006, NASA Ames Research Center, Moffett Field, CA 94035, March 1994

14
Hockney, R. and Berry, M. (editors), Public International Benchmarks for Parallel Computers, Comp. Sci. Dept., Univ. of Tennessee, CS-93-213, Knoxville, TN, November, 1993

15
Alan H. Karp and Ken Miura and Horst Simon, Gordon Bell Prize 1992, IEEE Computer, 26, 1, 77-82, January, 1993

16
J. J. Dongarra, H. W. Meuer, and E. Strohmaier,
TOP500 Supercomputer Sites, Technical Report RUM 34/93, Computing Centre, Univ. of Mannheim, 68131 Mannheim,
Germany, November 1993

17
J. J. Dongarra, Performance of Various Computers Using Standard Linear Equations Software, Computer Science Department, University of Tennessee, CS-89-85, 1994