Skip to main content

Table 7 Analysis of the clusters Gflops performance and speedup by using the HPL benchmark with \(NB\,=\,32\), different numbers of nodes and different proportion of the systems total main memory utilized

From: Mobile clusters of single board computers: an option for providing resources to student projects and researchers

Proportion of total main memory of the system utilized N Nodes used NB P Q Time to solve the linear system Gflops\(^{1}\) Speedup\(^{2}\)
\(\approx\) 70 % 5696 1 32 1 1 589.17 s 0.209 1.00
8064 2 32 1 2 983.76 s 0.355 \(\approx\) 1.69
11,392 4 32 1 4 1641.22 s 0.600 \(\approx\) 2.78
16,128 8 32 2 4 2166.95 s 1.291 \(\approx\) 6,17
\(\approx\) 75 % 6112 1 32 1 1 722.29 s 0.210 1.00
8640 2 32 1 2 1153.95 s 0.372 \(\approx\) 1.77
12,224 4 32 1 4 1977.85 s 0.615 \(\approx\) 2.92
17,280 8 32 2 4 2617.08 s 1.315 \(\approx\) 6.26
\(\approx\) 80 % 6496 1 32 1 1 859.81 s 0.212 1.00
9216 2 32 1 2 1503.04 s 0.347 \(\approx\) 1.77
13,024 4 32 1 4 2328.03 s 0.632 \(\approx\) 2.98
18,432 8 32 2 4 3055.37 s 1.367 \(\approx\) 6.44
\(\approx\) 85 % 6912 1 32 1 1 1037.84 s 0.212 1.00
9792 2 32 1 2 1705.12 s 0.367 \(\approx\) 1.73
13,856 4 32 1 4 2782.23 s 0.637 \(\approx\) 3.00
19,584 8 32 2 4 3688.51 s 1.358 \(\approx\) 6.40
\(\approx\) 90 % 7328 1 32 1 1 1246.73 s 0.210 1.00
10,368 2 32 1 2 1993.88 s 0.372 \(\approx\) 1.77
14,656 4 32 1 4 3274.85 s 0.641 \(\approx\) 3.05
20,768 8 32 2 4 4370.31 s 1.367 \(\approx\) 6.50
\(\approx\) 95 % 7744 1 32 1 1 1578.72 s 0.196 1.00
10,944 2 32 1 2 2699.97 s 0.323 \(\approx\) 1.64
15,488 4 32 1 4 4927.79 s 0.502 \(\approx\) 2.56
21,920 8 32 2 4 10,326.42 s 0.680 \(\approx\) 3.46
  1. \(^{1}\) The Gflops are rounded to three decimal places behind the decimal point
  2. \(^{2}\) The Speedup is rounded to two decimal places behind the decimal point