Intel Skylake S2 M2 C20
- Processor: Intel(R) Xeon(R) Gold 6148 CPU
- Base frequency: 2.4 GHz
- Number of sockets: 2
- Number of memory domains per socket: 2
- Memory domain specs: 3-channel DDR4-2666
- Number of cores per socket: 20
- Number of HWThreads per core: 2
- MachineState output: json
+----------+---------------------------------+
| Compiler | icc (ICC) |
|----------|---------------------------------|
| Version | icc (ICC) 19.0.5.281 20190815 |
+----------+---------------------------------+
Optimizing flags: -fast -xHost -qopt-streaming-stores=always -std=c99 -ffreestanding -qopenmp
All results are in GB/s.
Summary results:
+--------------------------------------------+
| Single core | 19.90 (SDaxpy) |
| Memory domain | 60.17 (Sum with 8 cores) |
| Socket | 118.95 (Sum with 7 cores) |
| Node | 235.68 (Sum with 8 cores) |
+--------------------------------------------+
Results for scaling within a memory domain:
#nt Init Sum Copy Update Triad Daxpy STriad SDaxpy
1 7.59 15.73 10.90 17.16 14.15 19.05 15.97 19.90
2 15.03 30.11 21.22 33.14 27.56 37.27 31.06 37.99
3 22.44 42.05 31.54 45.16 39.94 46.93 43.71 47.50
4 29.42 51.27 40.60 50.06 46.43 49.61 48.47 50.15
5 36.58 56.83 46.62 53.01 49.07 50.84 50.60 50.98
6 43.19 59.14 49.06 52.68 50.05 50.81 50.78 50.89
7 47.74 60.01 50.15 52.02 50.00 50.23 50.56 51.21
8 48.06 60.17 50.02 51.34 49.73 49.50 49.86 50.38
9 47.21 59.69 49.78 51.12 50.04 50.97 50.05 50.81
10 46.68 59.26 49.66 50.72 49.44 50.22 49.84 50.25
Results for scaling across memory domains. Shown are the results for the number of memory domains used (nm) with columns number of cores used per memory domain.
Init:
#nm 1 2 3 4
1 7.59 14.41 22.13 29.56
2 15.03 28.21 43.48 57.87
3 22.44 41.83 64.60 85.55
4 29.42 53.88 84.85 111.34
5 36.58 70.01 104.13 133.58
6 43.19 81.43 119.56 140.62
7 47.74 87.82 129.62 148.63
8 48.06 88.28 129.51 155.62
9 47.21 89.12 133.71 164.44
10 46.68 88.34 129.19 160.26
Sum:
#nm 1 2 3 4
1 15.73 31.44 46.86 62.49
2 30.11 59.31 87.70 115.54
3 42.05 83.90 127.24 165.79
4 51.27 101.29 152.35 200.41
5 56.83 112.84 168.77 222.43
6 59.14 117.32 174.67 231.62
7 60.01 118.95 177.60 235.36
8 60.17 118.60 176.35 235.68
9 59.69 117.66 176.57 235.54
10 59.26 115.71 172.44 227.87
Copy
#nm 1 2 3 4
1 10.90 20.98 32.08 42.78
2 21.22 40.65 62.30 82.84
3 31.54 59.74 91.86 121.12
4 40.60 77.42 118.78 157.48
5 46.62 90.25 138.06 181.78
6 49.06 94.60 145.03 189.95
7 50.15 97.39 145.73 193.45
8 50.02 97.41 144.54 196.85
9 49.78 98.29 147.63 195.91
10 49.66 94.93 140.04 182.10
Update
#nm 1 2 3 4
1 17.16 34.32 50.85 67.86
2 33.14 65.57 98.41 130.33
3 45.16 89.41 134.43 178.62
4 50.06 98.46 148.28 198.61
5 53.01 104.73 156.23 209.10
6 52.68 103.68 156.29 206.93
7 52.02 102.93 151.41 202.06
8 51.34 101.75 150.78 205.47
9 51.12 101.48 151.34 204.73
10 50.72 98.38 145.60 190.68
Triad
#nm 1 2 3 4
1 14.15 27.62 42.26 56.38
2 27.56 53.77 81.03 107.94
3 39.94 77.19 117.30 155.71
4 46.43 92.40 138.25 186.38
5 49.07 97.68 145.42 193.33
6 50.05 97.26 148.91 192.74
7 50.00 96.76 145.43 191.47
8 49.73 97.82 141.88 196.57
9 50.04 97.58 146.94 196.14
10 49.44 94.07 140.20 184.75
Memory bandwidth scaling within one memory domain:
The following plots illustrate the the performance scaling over multiple memory domains using different number of cores per memory domain.
Memory bandwidth scaling across memory domains for init:
Memory bandwidth scaling across memory domains for sum
Memory bandwidth scaling across memory domains for copy
Memory bandwidth scaling across memory domains for Triad
