HEP Workloads: Reproducibility of Benchmark Results
This Twiki page is aimed at internal discussions of the
HEPiX Benchmarking Working Group about the spread in the results of repeated benchmark runs.
It does definitively not provide any final numbers.
Spread of 50 repeated benchmark runs (latest HEP software release) |
Hardware model: c01-010-156.gridka.de (SL7) - Turbo Boost enabled in BIOS |
Workload |
Version |
No. of |
Total scores |
Stdev |
(Max-Min) |
Success |
Av. |
events |
copies |
Min |
Max |
Mean |
Stdev |
/ Mean |
rate |
runtime |
ALICE gen-sim |
v0.7 |
|
20 |
0.769 |
0.893 |
0.868 |
0.024 |
0.027 |
0.143 |
100% |
170s |
|
40 |
0.884 |
0.939 |
0.901 |
0.012 |
0.014 |
0.061 |
70% |
283s |
v0.6 |
|
20 |
0.700 |
0.891 |
0.865 |
0.031 |
0.036 |
0.220 |
94% |
305s |
|
40 |
0.884 |
1.322 |
0.912 |
0.051 |
0.056 |
0.524 |
81% |
431s |
ATLAS gen-bmk |
v0.7 |
5 |
20 |
493.4 |
510.9 |
502.6 |
4.649 |
0.009 |
0.035 |
|
294s |
40 |
771.2 |
840.4 |
802.2 |
17.89 |
0.022 |
0.086 |
461s |
v0.3 |
5 |
20 |
476.5 |
518.2 |
502.8 |
8.721 |
0.017 |
0.083 |
439s |
40 |
713.0 |
837.9 |
795.1 |
23.97 |
0.030 |
0.157 |
581s |
ATLAS sim-bmk (8 threads per copy) |
|
v0.6 |
10 |
5 |
0.127 |
0.131 |
0.128 |
0.001 |
0.005 |
0.027 |
|
3737s |
ATLAS sim-bmk (4 threads per copy) |
|
v0.16 |
5 |
5 |
0.083 |
0.086 |
0.085 |
0.001 |
0.009 |
0.039 |
|
1603s |
10 |
0.092 |
0.093 |
0.093 |
0.000 |
0.000 |
0.013 |
2726s |
v0.9 |
5 |
0.082 |
0.085 |
0.084 |
0.001 |
0.009 |
0.039 |
1621s |
10 |
0.093 |
0.094 |
0.093 |
0.000 |
0.003 |
0.014 |
2697s |
v0.7 |
5 |
0.082 |
0.085 |
0.084 |
0.001 |
0.009 |
0.035 |
1740s |
10 |
0.093 |
0.094 |
0.093 |
0.000 |
0.034 |
0.013 |
2831s |
ATLAS reco-bmk |
CPU_score (All) |
|
5 |
20 |
42.82 |
46.00 |
45.26 |
0.662 |
0.009 |
0.070 |
|
1070s |
(ESDtoAOD) |
38.86 |
41.98 |
41.28 |
0.661 |
0.016 |
0.076 |
(HITtoRDO) |
0.439 |
0.469 |
0.467 |
0.006 |
0.014 |
0.065 |
(RAWtoESD) |
1.132 |
1.453 |
1.418 |
0.077 |
0.054 |
0.226 |
(RDOtoRDOTrigger) |
2.050 |
2.114 |
2.087 |
0.019 |
0.009 |
0.031 |
CPU_score (All) |
40 |
62.70 |
73.15 |
65.68 |
2.562 |
0.039 |
0.159 |
|
1783s |
(ESDtoAOD) |
57.73 |
68.14 |
60.61 |
2.534 |
0.042 |
0.172 |
(HITtoRDO) |
0.625 |
0.630 |
0.628 |
0.001 |
0.002 |
0.008 |
(RAWtoESD) |
1.747 |
1.773 |
1.756 |
0.007 |
0.004 |
0.015 |
(RDOtoRDOTrigger) |
2.583 |
2.960 |
2.680 |
0.121 |
0.045 |
0.141 |
CMS gen-sim ttbar (4 threads per copy) |
CPU_score |
v0.13 |
25 |
5 |
0.221 |
0.229 |
0.225 |
0.002 |
0.009 |
0.034 |
|
634s |
througput_score |
0.850 |
0.881 |
0.863 |
0.008 |
0.009 |
0.036 |
CPU_score |
10 |
0.246 |
0.248 |
0.247 |
0.001 |
0.002 |
0.009 |
|
1093s |
througput_score |
0.954 |
0.965 |
0.960 |
0.003 |
0.003 |
0.011 |
CPU_score |
v0.5 |
25 |
10 |
0.244 |
0.250 |
0.247 |
0.001 |
0.004 |
0.026 |
|
1104s |
througput_score |
0.925 |
0.971 |
0.956 |
0.005 |
0.006 |
0.048 |
CPU_score |
v0.5 |
20 |
10 |
0.248 |
0.252 |
0.250 |
0.001 |
0.004 |
0.019 |
|
1012s |
througput_score |
0.951 |
0.973 |
0.963 |
0.004 |
0.004 |
0.023 |
CPU_score |
v0.5 |
100 |
10 |
0.247 |
0.252 |
0.250 |
0.001 |
0.004 |
0.021 |
|
4272s |
througput_score |
0.951 |
0.997 |
0.976 |
0.013 |
0.013 |
0.021 |
CPU_score |
v0.4 |
100 |
10 |
0.247 |
0.358 |
0.253 |
0.017 |
0.059 |
0.438 |
|
4264s |
througput_score |
0.977 |
1.075 |
0.989 |
0.016 |
0.016 |
0.099 |
CMS digi (4 threads per copy) |
CPU_score |
v0.9 |
100 |
5 |
1.106 |
1.114 |
1.099 |
0.015 |
0.013 |
0.049 |
|
768s |
througput_score |
3.394 |
4.315 |
3.867 |
0.277 |
0.071 |
0.237 |
CPU_score |
10 |
1.319 |
1.337 |
1.327 |
0.005 |
0.00.36 |
0.014 |
|
1140s |
througput_score |
4.440 |
5.233 |
5.027 |
0.293 |
0.058 |
0.158 |
CPU_score |
v0.5 |
100 |
10 |
4.349 |
4.437 |
4.393 |
0.027 |
0.006 |
0.021 |
|
410s |
througput_score |
4.114 |
4.239 |
4.190 |
0.027 |
0.006 |
0.020 |
CPU_score |
v0.4 |
50 |
10 |
2.176 |
2.225 |
2.193 |
0.013 |
0.006 |
0.022 |
|
399s |
througput_score |
4.136 |
4.253 |
4.184 |
0.026 |
0.006 |
0.028 |
CMS reco (4 threads per copy) |
CPU_score |
v0.12 |
100 |
5 |
0.625 |
0.646 |
0.639 |
0.005 |
0.008 |
0.031 |
|
1126s |
througput_score |
2.469 |
2.548 |
2.521 |
0.021 |
0.008 |
0.031 |
CPU_score |
10 |
0.743 |
0.757 |
0.750 |
0.003 |
0.004 |
0.014 |
|
1427s |
througput_score |
2.936 |
2.985 |
2.967 |
0.013 |
0.004 |
0.016 |
CPU_score |
v0.2-v0.3 |
100 |
10 |
0.745 |
0.757 |
0.752 |
0.002 |
0.003 |
0.016 |
|
1774s |
througput_score |
2.940 |
2.989 |
2.973 |
0.011 |
0.004 |
0.017 |
CPU_score |
v0.1 |
50 |
10 |
0.738 |
0.749 |
0.744 |
0.002 |
0.003 |
0.015 |
|
844s |
througput_score |
2.913 |
2.955 |
2.933 |
0.012 |
0.004 |
0.014 |
LHCb gen sim |
v0.1 |
|
20 |
104.9 |
109.6 |
108.4 |
1.042 |
0.009 |
0.043 |
|
1242s |
|
40 |
114.9 |
117.1 |
116.1 |
0.338 |
0.003 |
0.019 |
|
2119s |
KV-bmk |
v1.0 |
|
20 |
29.10 |
30.38 |
30.62 |
0.336 |
0.011 |
0.049 |
|
223s |
|
40 |
35.44 |
36.75 |
35.93 |
0.227 |
0.006 |
0.037 |
|
339s |
Hardware model: c01-010-155.gridka.de (SL6) - Turbo Boost enabled in BIOS |
Workload |
Version |
No. of |
Total scores |
Stdev |
(Max-Min) |
Success |
Av. |
events |
copies |
Min |
Max |
Mean |
Stdev |
/ Mean |
rate |
runtime |
ALICE gen-sim |
v0.7 |
5 |
20 |
0.838 |
0.864 |
0.850 |
0.008 |
0.009 |
0.030 |
90% |
174s |
40 |
0.919 |
0.981 |
0.938 |
0.018 |
0.019 |
0.073 |
80% |
279s |
v0.6 |
|
20 |
0.812 |
0.874 |
0.849 |
0.011 |
0.012 |
0.064 |
94% |
176s |
|
40 |
0.799 |
0.981 |
0.932 |
0.030 |
0.032 |
0.196 |
84% |
312s |
ATLAS gen-bmk |
v0.7 |
5 |
20 |
478.8 |
506.5 |
479.6 |
6.537 |
0.013 |
0.056 |
|
307s |
40 |
703.5 |
818.1 |
758.1 |
37.08 |
0.049 |
0.151 |
|
433s |
v0.3 |
5 |
20 |
456.5 |
516.6 |
499.2 |
9.692 |
0.019 |
0.120 |
|
324s |
40 |
653.3 |
834.5 |
752.4 |
43.35 |
0.058 |
0.241 |
|
443s |
ATLAS sim-bmk (8 threads per copy) |
|
v0.7 |
10 |
5 |
0.129 |
0.137 |
0.133 |
0.012 |
0.009 |
0.059 |
|
3493s |
ATLAS sim-bmk (4 threads per copy) |
|
v0.9 |
5 |
5 |
0.082 |
0.086 |
0.085 |
0.001 |
0.012 |
0.055 |
|
1610s |
|
v0.9 |
5 |
10 |
0.095 |
0.098 |
0.097 |
0.001 |
0.006 |
0.025 |
|
2644s |
|
v0.7 |
5 |
5 |
0.083 |
0.087 |
0.085 |
0.001 |
0.011 |
0.042 |
|
1583s |
10 |
0.095 |
0.097 |
0.096 |
0.000 |
0.005 |
0.024 |
|
2668s |
ATLAS reco-bmk |
CPU_score (All) |
|
5 |
20 |
45.37 |
46.81 |
46.32 |
0.359 |
0.008 |
0.031 |
|
1003s |
(ESDtoAOD) |
41.59 |
43.05 |
42.62 |
0.368 |
0.009 |
0.034 |
(HITtoRDO) |
0.438 |
0.470 |
0.461 |
0.008 |
0.017 |
0.070 |
(RAWtoESD) |
1.358 |
1.431 |
1.397 |
0.018 |
0.013 |
0.052 |
(RDOtoRDOTrigger) |
1.684 |
1.976 |
1.844 |
0.082 |
0.044 |
0.158 |
CPU_score (All) |
40 |
66.09 |
78.22 |
70.39 |
3.468 |
0.049 |
0.172 |
|
2040s |
(ESDtoAOD) |
61.24 |
73.33 |
65.49 |
3.454 |
0.053 |
0.185 |
(HITtoRDO) |
0.629 |
0.647 |
0.634 |
0.004 |
0.007 |
0.029 |
(RAWtoESD) |
1.728 |
1.773 |
1.752 |
0.013 |
0.007 |
0.026 |
(RDOtoRDOTrigger) |
2.431 |
2.603 |
2.511 |
0.047 |
0.019 |
0.069 |
CMS gen-sim ttbar (4 threads per copy) |
CPU_score |
v0.8-v0.9 |
100 |
10 |
0.258 |
0.290 |
0.264 |
0.006 |
0.025 |
0.119 |
|
4110s |
througput_score |
1.005 |
1.051 |
1.022 |
0.009 |
0.009 |
0.044 |
CPU_score |
v0.5 |
20 |
10 |
0.258 |
0.262 |
0.264 |
0.004 |
0.014 |
0.060 |
|
860s |
througput_score |
0.985 |
1.022 |
0.999 |
0.007 |
0.007 |
0.036 |
CPU_score |
v0.5 |
100 |
10 |
0.258 |
0.274 |
0.263 |
0.003 |
0.012 |
0.060 |
|
860s |
througput_score |
0.985 |
1.029 |
1.009 |
0.012 |
0.012 |
0.043 |
CPU_score |
v0.4 |
100 |
10 |
0.257 |
0.318 |
0.265 |
0.013 |
0.051 |
0.229 |
|
4028s |
througput_score |
1.004 |
1.074 |
1.020 |
0.013 |
0.012 |
0.069 |
CMS digi (4 threads per copy) |
CPU_score |
v0.5 |
100 |
10 |
4.077 |
4.477 |
4.348 |
0.101 |
0.023 |
0.093 |
|
385s |
througput_score |
3.840 |
4.189 |
4.076 |
0.087 |
0.021 |
0.086 |
CPU_score |
v0.4 |
50 |
10 |
1.892 |
2.130 |
2.030 |
0.080 |
0.040 |
0.117 |
|
425s |
througput_score |
3.315 |
3.996 |
3.676 |
0.240 |
0.065 |
0.185 |
CMS reco (4 threads per copy) |
CPU_score |
v0.2-v0.3 |
100 |
10 |
0.740 |
0.772 |
0.758 |
0.009 |
0.012 |
0.042 |
|
1786s |
througput_score |
2.724 |
2.979 |
2.874 |
0.083 |
0.029 |
0.089 |
CPU_score |
v0.1 |
50 |
10 |
0.717 |
0.750 |
0.731 |
0.009 |
0.013 |
0.044 |
|
928s |
througput_score |
2.558 |
2.825 |
2.635 |
0.073 |
0.028 |
0.101 |
LHCb gen sim |
v0.1 |
|
20 |
103.4 |
109.2 |
106.9 |
1.176 |
0.011 |
0.054 |
|
1156s |
|
40 |
117.0 |
122.6 |
119.7 |
0.960 |
0.008 |
0.047 |
|
1935s |
KV-bmk |
v1.0 |
|
20 |
30.78 |
31.26 |
31.12 |
0.120 |
0.004 |
0.015 |
|
304s |
|
40 |
35.19 |
36.58 |
35.85 |
0.330 |
0.009 |
0.039 |
|
340s |
Spread of 20 repeated benchmark runs (latest HEP software release) |
Hardware model: c01-010-156.gridka.de (SL7) - Turbo Boost disabled in BIOS |
Workload |
Version |
No. of |
Total scores |
Stdev |
(Max-Min) |
Success |
Av. |
events |
copies |
Min |
Max |
Mean |
Stdev |
/ Mean |
rate |
runtime |
ALICE gen-sim |
v0.6 |
|
20 |
0.669 |
0.758 |
0.670 |
0.030 |
0.043 |
0.128 |
95% |
223s |
|
40 |
0.694 |
0.811 |
0.738 |
0.029 |
0.040 |
0.159 |
85% |
372s |
ATLAS gen-bmk |
v0.3 |
5 |
20 |
417.395 |
430.162 |
424.785 |
3.303 |
0.008 |
0.030 |
100% |
337s |
5 |
40 |
629.448 |
707.317 |
672.776 |
17.461 |
0.026 |
0.116 |
100% |
498s |
ATLAS sim-bmk (8 threads per copy) |
|
v0.4 |
10 |
5 |
0.087 |
0.111 |
0.107 |
0.007 |
0.061 |
0.216 |
100% |
4243s |
CMS gen-sim ttbar (4 threads per copy) |
CPU_score |
v0.4 |
20 |
10 |
0.213 |
0.217 |
0.215 |
0.001 |
0.005 |
0.020 |
100% |
1030s |
througput_score |
0.821 |
0.838 |
0.831 |
0.001 |
0.005 |
0.020 |
LHCb gen sim |
v0.1 |
|
20 |
95.261 |
96.051 |
95.578 |
0.231 |
0.002 |
0.008 |
100% |
1311s |
|
40 |
105.988 |
107.131 |
106.636 |
0.254 |
0.002 |
0.011 |
100% |
2187s |
HS06 (10 runs) |
|
|
40 |
412.4 |
418.0 |
414.2 |
1.520 |
0.004 |
0.014 |
|
|
Hardware model: c01-010-155.gridka.de (SL6) - Turbo Boost disabled in BIOS |
Workload |
Version |
No. of |
Total scores |
Stdev |
(Max-Min) |
Success |
Av. |
events |
copies |
Min |
Max |
Mean |
Stdev |
/ Mean |
rate |
runtime |
ALICE gen-sim |
v0.6 |
|
20 |
0.752 |
0.788 |
0.769 |
0.009 |
0.012 |
0.047 |
100% |
199s |
|
40 |
0.763 |
0.922 |
0.867 |
0.045 |
0.052 |
0.183 |
85% |
323s |
ATLAS gen-bmk |
v0.3 |
5 |
20 |
413.165 |
446.711 |
439.067 |
8.343 |
0.019 |
0.076 |
100% |
332s |
5 |
40 |
575.117 |
740.849 |
663.477 |
46.827 |
0.071 |
0.250 |
100% |
498s |
ATLAS sim-bmk (8 threads per copy) |
|
v0.4 |
10 |
5 |
0.047 |
0.121 |
0.111 |
0.017 |
0.155 |
0.661 |
100% |
3936s |
CMS gen-sim ttbar (4 threads per copy) |
CPU_score |
v0.4 |
20 |
10 |
0.231 |
0.241 |
0.237 |
0.002 |
0.010 |
0.041 |
100% |
961s |
througput_score |
0.870 |
0.901 |
0.891 |
0.008 |
0.009 |
0.035 |
LHCb gen sim |
v0.1 |
|
20 |
95.360 |
98.875 |
97.249 |
0.890 |
0.009 |
0.036 |
100% |
1250s |
|
40 |
109.450 |
115.184 |
113.674 |
1.166 |
0.010 |
0.050 |
100% |
2044s |
HS06 (10 runs) |
|
|
40 |
414.6 |
421.2 |
417.8 |
2.133 |
0.005 |
0.016 |
|
|
Spread of 20 repeated benchmark runs (latest HEP software release) |
Hardware model: c01-010-156.gridka.de |
Workload |
Date |
No. of |
Total scores |
Stdev/ |
(Max-Min)/ |
Success |
Av. |
events |
copies |
Min |
Max |
Mean |
Stdev |
Mean |
Mean |
rate |
runtime |
ALICE gen-sim v0.4 |
2019-02-27 |
|
1 |
0.053 |
0.056 |
0.055 |
0 |
0 |
0 033 |
100% |
|
|
20 |
0.822 |
0.884 |
0.870 |
0.013 |
0.015 |
0.072 |
100% |
|
|
32 |
0.794 |
0.921 |
0.876 |
0.039 |
0.045 |
0.137 |
90% |
|
|
40 |
0.800 |
0.910 |
0.884 |
0.030 |
0.034 |
0.126 |
80% |
|
ATLAS gen-bmk |
2019-03-02 |
5 |
1 |
28.571 |
29.672 |
29.394 |
0.319 |
0.011 |
0.041 |
100% |
|
5 |
20 |
477.579 |
516.113 |
501.354 |
8.729 |
0.017 |
0.077 |
100% |
|
5 |
32 |
581.384 |
699.518 |
635.287 |
32.660 |
0.051 |
0.186 |
100% |
|
5 |
40 |
739.762 |
839.345 |
797.658 |
21.559 |
0.027 |
0.125 |
100% |
433s |
2019-03-04 |
25 |
40 |
538.660 |
605.196 |
578.275 |
14.382 |
0.025 |
0.115 |
100% |
449s |
2019-03-05 |
100 |
40 |
654.203 |
744.648 |
707.456 |
25.138 |
0.036 |
0.128 |
100% |
634s |
2019-03-06 |
500 |
40 |
607.589 |
667.795 |
642.080 |
15.961 |
0.025 |
0.094 |
100% |
1896s |
2019-03-06 |
2500 |
40 |
643.047 |
680.619 |
658.053 |
9.299 |
0.014 |
0.057 |
100% |
8539s |
CMS gen-sim ttbar v0.2 (4 threads per copy) |
CPU_score |
2019-03-14 |
20 |
10 |
0.246 |
0.251 |
0.249 |
0 |
0 |
0.019 |
100% |
884s |
througput_score |
10 |
0.943 |
0.963 |
0.959 |
0.007 |
0.007 |
0.027 |
100% |
884s |
LHCb gen sim |
2019-03-01 |
|
1 |
5.589 |
6.417 |
6.221 |
0.243 |
0.039 |
0.133 |
100% |
|
|
20 |
108.310 |
109.489 |
108.972 |
0.339 |
0.003 |
0.011 |
100% |
|
|
32 |
112.797 |
114.795 |
113.612 |
0.554 |
0.005 |
0.018 |
100% |
|
|
40 |
115.924 |
116.967 |
116.286 |
0.262 |
0.002 |
0.009 |
100% |
|
Docker versus Singularity (average of 20 runs) |
Workload |
Copies |
Threads per copy |
Singularity |
Docker |
ATLAS sim |
5 |
8 |
0.118 |
0.118 |
LHCb gen-sim |
20 |
1 |
91.79 |
91.67 |
40 |
1 |
104.3 |
104.7 |
Spread of 20 repeated benchmark runs |
Workload |
Date |
No. of |
Hardware model |
Runtime |
copies |
c01-010-156.gridka.de |
c01-025-131.gridka.de |
c01-028-182.gridka.de |
ALICE gen-sim |
2019-02-07 |
1 |
0.0329...0.0333 |
0.0312...0.0321 |
0.0260...1.6667 |
O(3') |
16 |
|
|
0.0221...1.9999 |
O(4') |
20 |
0.0273...2.9986 |
0.0414...1.9484 |
|
O(4') |
24 |
|
|
0.0776...1.3623 |
O(6') |
32 |
0.0179...2.0441 |
0.0312...2.5807 |
0.0473...1.4164 |
O(6') |
40 |
0.0403...1.2904 |
0.0127...2.7695 |
|
O(8') |
ATLAS gen-bmk |
2019-02-08 |
1 |
28.0112...29.3255 |
|
|
O(4') |
16 |
|
|
289.7805...329.9172 |
O(6') |
20 |
490.4275...508.0936 |
367.2221...419.9154 |
|
O(5') |
24 |
|
|
368.9814...439.7616 |
O(7') |
32 |
598.5838...708.9543 |
494.9507...605.0028 |
436.1996...531.0295 |
O(6') |
40 |
751.6043...842.3921 |
563.2879...680.9631 |
|
O(7') |
LHCb gen-sim |
2019-02-07 |
1 |
5.5044...6.2811 |
5.6604...6.3401 |
4.2387...4.3335 |
O(16') |
16 |
|
|
59.2598...60.5848 |
O(25') |
20 |
104.1588...106.9673 |
91.3763...97.5531 |
|
O(19') |
24 |
|
|
63.6833...64.9746 |
O(35') |
32 |
110.3778...112.3555 |
103.4095...106.3811 |
67.3353...68.6971 |
O(30') |
40 |
113.1566...113.8484 |
109.2494...110.8703 |
|
O(35') |
|
HS06 scores of the systems under test are provided for information only: |
HS06 |
|
16 |
|
|
261 |
|
20 |
374 |
333 |
|
|
24 |
|
|
305 |
|
32 |
447 |
390 |
327 |
|
40 |
465 |
416 |
|
|
Hardware models |
Hostname |
Processor(s) |
Number of |
OS |
cores |
logical processors |
c01-010-155.gridka.de |
2x Intel Xeon E5-2660v3 (2.6 GHz) - Haswell |
20 |
40 |
SL6 |
c01-010-156.gridka.de |
2x Intel Xeon E5-2660v3 (2.6 GHz) - Haswell |
20 |
40 |
SL7 |
c01-025-131.gridka.de |
2x Intel Xeon E5-2630v4 (2.2 GHz) - Broadwell |
20 |
40 |
SL7 |
c01-028-182.gridka.de |
2x Intel Xeon E5-2665 (2.4 GHz) - Sandy Bridge |
16 |
32 |
SL7 |
--
ManfredAlefExternal - 2019-02-06