Intel CascadeLake S2 M1 C20
- Processor: Intel(R) Xeon(R) Gold 6248
- Base frequency: 2.5 GHz
- Number of sockets: 2
- Number of memory domains per socket: 1
- Memory domain specs: 6-channel DDR4-2933
- Number of cores per socket: 20
- Number of HWThreads per core: 2
- MachineState output: NA
+----------+---------------------------------+
| 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 | 17.26 (SDaxpy) |
| Memory domain | 119.27 (Sum with 14 cores) |
| Socket | 119.27 (Sum with 14 cores) |
| Node | 237.31 (Sum with 14 cores) |
+----------------------------------------------+
Results for scaling within a memory domain:
#nt Init Sum Copy Update Triad Daxpy STriad SDaxpy
1 7.22 11.01 10.04 14.69 12.91 16.87 14.41 17.26
2 14.31 21.77 19.77 29.22 25.36 32.96 28.31 33.84
3 21.55 33.87 29.53 44.43 37.75 49.90 41.98 50.98
4 28.40 46.56 38.70 57.12 49.15 63.62 54.48 65.49
5 35.51 60.57 48.00 69.72 60.42 75.38 66.25 76.68
6 42.57 73.73 56.53 79.95 69.65 82.55 75.34 82.74
7 49.33 85.12 64.26 85.46 77.80 87.63 82.66 87.66
8 56.27 94.51 71.20 88.35 84.19 91.14 88.51 91.58
9 62.92 102.52 77.06 91.63 88.90 94.28 92.33 94.89
10 69.45 109.14 82.31 94.08 92.27 96.41 94.40 96.21
11 75.08 114.24 86.70 96.08 94.43 97.70 95.45 96.64
12 77.88 116.92 89.16 96.70 95.19 97.34 95.38 96.15
13 79.75 119.18 91.50 97.69 96.04 97.61 95.51 96.26
14 80.79 119.27 92.55 97.93 95.81 97.06 95.12 95.59
15 81.23 118.12 93.13 97.95 95.46 96.54 94.88 95.16
16 81.03 116.55 92.77 97.19 94.56 95.51 94.18 94.21
17 81.48 114.96 93.18 97.30 94.86 95.37 94.36 94.14
18 81.54 112.78 93.06 96.97 94.62 95.08 94.35 93.91
19 81.50 111.32 92.86 96.70 94.50 94.94 94.29 93.76
20 81.51 110.14 92.82 96.30 94.51 94.76 94.34 93.70
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
1 7.22 14.53
2 14.31 28.87
3 21.55 43.29
4 28.40 56.87
5 35.51 70.96
6 42.57 84.32
7 49.33 97.65
8 56.27 110.23
9 62.92 121.68
10 69.45 132.16
11 75.08 139.55
12 77.88 141.18
13 79.75 144.63
14 80.79 148.80
15 81.23 150.96
16 81.03 152.89
17 81.48 154.81
18 81.54 156.02
19 81.50 157.31
20 81.51 157.90
Sum:
#nm 1 2
1 11.01 21.95
2 21.77 43.62
3 33.87 67.41
4 46.56 93.25
5 60.57 120.60
6 73.73 146.70
7 85.12 169.84
8 94.51 188.40
9 102.52 204.55
10 109.14 217.89
11 114.24 227.75
12 116.92 233.63
13 119.18 236.93
14 119.27 237.31
15 118.12 236.68
16 116.55 232.79
17 114.96 228.12
18 112.78 225.93
19 111.32 222.02
20 110.14 220.54
Copy
#nm 1 2
1 10.04 20.04
2 19.77 39.55
3 29.53 58.85
4 38.70 77.16
5 48.00 95.38
6 56.53 112.19
7 64.26 127.62
8 71.20 141.11
9 77.06 153.03
10 82.31 163.09
11 86.70 171.25
12 89.16 177.51
13 91.50 181.52
14 92.55 184.00
15 93.13 185.30
16 92.77 184.93
17 93.18 185.47
18 93.06 185.26
19 92.86 185.00
20 92.82 184.13
Update
#nm 1 2
1 14.69 29.38
2 29.22 58.66
3 44.43 88.66
4 57.12 114.75
5 69.72 139.62
6 79.95 160.17
7 85.46 171.96
8 88.35 178.15
9 91.63 184.90
10 94.08 190.05
11 96.08 193.87
12 96.70 195.49
13 97.69 197.14
14 97.93 197.92
15 97.95 198.00
16 97.19 196.58
17 97.30 196.87
18 96.97 196.33
19 96.70 195.86
20 96.30 194.99
Triad
#nm 1 2
1 12.91 25.71
2 25.36 50.58
3 37.75 75.15
4 49.15 97.98
5 60.42 119.94
6 69.65 137.95
7 77.80 154.90
8 84.19 167.69
9 88.90 177.30
10 92.27 183.40
11 94.43 188.41
12 95.19 190.04
13 96.04 191.17
14 95.81 190.91
15 95.46 190.07
16 94.56 188.92
17 94.86 189.01
18 94.62 188.65
19 94.50 188.55
20 94.51 188.29
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
