AMD Rome S2 M1 C32
- Processor: AMD EPYC 7502 32-Core Processor
- Base frequency: 2.5 GHz
- Number of sockets: 2
- Number of memory domains per socket: 1
- Memory domain specs: 8-channel DDR4-3200
- Number of cores per socket: 32
- Number of HWThreads per core: 2
- MachineState output: NA
+----------+-------------------------------------------------------------------+
| Compiler | AMD clang |
+----------+-------------------------------------------------------------------+
| Version | AMD clang version 10.0.0 (CLANG: AOCC_2.2.0-Build#93 2020_06_25) |
+----------+-------------------------------------------------------------------+
Optimizing flags: -Ofast -fnt-store=aggressive -std=c99 -fopenmp
All results are in GB/s.
Summary results:
+------------------------------------------------+
| Single core | 37.19 (Triad) |
| Memory domain | 131.07 (Triad with 32 cores) |
| Socket | 131.07 (Triad with 32 cores) |
| Node | 268.63 (Update with 32 cores) |
+------------------------------------------------+
Results for scaling within a memory domain:
#nt Init Sum Copy Update Triad Daxpy STriad SDaxpy
1 23.42 8.24 33.60 31.04 37.19 34.80 36.45 34.83
2 23.33 16.45 43.66 41.16 46.64 41.35 44.32 40.77
3 23.35 24.61 44.22 41.85 46.89 41.41 43.89 40.30
4 23.42 32.36 44.56 42.42 46.67 41.20 43.68 40.10
5 23.44 36.48 44.23 42.41 46.52 41.16 43.58 40.05
6 23.44 36.08 44.30 42.75 46.23 40.91 43.34 39.80
7 23.45 36.40 44.35 42.78 45.92 40.64 43.12 39.55
8 23.45 37.35 44.31 42.71 45.82 40.46 43.06 39.41
9 26.38 41.92 49.79 47.72 51.26 45.30 48.23 44.14
10 29.31 46.36 55.21 52.61 56.45 49.90 53.09 48.68
11 32.24 50.78 60.64 57.51 61.53 54.56 57.94 53.18
12 35.17 55.09 66.02 62.41 66.39 59.03 62.58 57.53
13 38.10 59.46 71.41 67.15 71.25 63.57 67.34 61.95
14 41.04 63.71 76.73 71.88 75.86 67.87 71.74 66.06
15 43.96 67.93 81.91 76.27 80.23 71.98 76.11 70.12
16 46.88 72.14 86.80 80.98 84.47 75.94 80.20 73.98
17 49.80 76.02 91.38 84.73 88.43 79.77 84.22 77.82
18 52.74 79.90 95.12 88.52 92.09 83.32 87.89 81.46
19 55.66 83.80 98.93 92.05 95.64 86.78 91.55 84.87
20 58.61 87.35 102.67 95.74 99.16 90.17 94.95 88.32
21 61.54 91.10 105.94 98.60 102.45 93.30 98.49 91.53
22 64.46 94.68 109.96 102.28 105.69 96.70 101.74 94.84
23 67.40 98.21 112.96 105.40 108.59 99.48 104.69 97.82
24 70.31 101.84 116.76 109.15 111.56 102.54 107.76 100.81
25 73.21 105.07 117.69 111.64 114.27 105.24 110.64 103.62
26 76.08 107.66 119.46 114.28 116.90 107.62 113.57 106.37
27 79.01 110.70 120.85 115.96 119.33 110.18 116.27 108.96
28 81.92 113.37 123.62 119.16 122.01 112.79 118.73 111.69
29 84.75 116.06 123.83 120.65 124.36 114.92 121.25 114.06
30 87.59 118.76 125.00 122.83 126.71 117.35 123.66 116.45
31 90.58 121.35 126.63 123.91 129.03 119.55 126.00 118.77
32 93.26 123.97 128.48 127.66 131.07 121.55 128.13 120.99
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 23.42 46.77
2 23.33 46.82
3 23.35 46.79
4 23.42 46.85
5 23.44 46.88
6 23.44 46.85
7 23.45 46.88
8 23.45 46.88
9 26.38 52.74
10 29.31 58.59
11 32.24 64.46
12 35.17 70.32
13 38.10 76.16
14 41.04 82.01
15 43.96 87.89
16 46.88 93.72
17 49.80 99.56
18 52.74 105.40
19 55.66 111.27
20 58.61 117.09
21 61.54 122.90
22 64.46 128.91
23 67.40 134.71
24 70.31 140.56
25 73.21 146.32
26 76.08 152.13
27 79.01 157.75
28 81.92 163.64
29 84.75 169.35
30 87.59 175.22
31 90.58 180.74
32 93.26 186.82
Sum:
#nm 1 2
1 8.24 16.49
2 16.45 32.96
3 24.61 49.13
4 32.36 64.82
5 36.48 73.11
6 36.08 71.88
7 36.40 72.84
8 37.35 74.79
9 41.92 83.96
10 46.36 92.78
11 50.78 101.58
12 55.09 110.41
13 59.46 118.82
14 63.71 127.24
15 67.93 136.01
16 72.14 144.33
17 76.02 151.95
18 79.90 159.67
19 83.80 167.40
20 87.35 174.18
21 91.10 182.22
22 94.68 188.97
23 98.21 196.53
24 101.84 203.26
25 105.07 209.47
26 107.66 215.87
27 110.70 220.98
28 113.37 226.12
29 116.06 232.20
30 118.76 237.18
31 121.35 242.62
32 123.97 245.19
Copy
#nm 1 2
1 33.60 66.73
2 43.66 87.42
3 44.22 88.11
4 44.56 89.12
5 44.23 88.40
6 44.30 88.62
7 44.35 88.68
8 44.31 88.59
9 49.79 99.63
10 55.21 110.44
11 60.64 121.23
12 66.02 132.04
13 71.41 142.93
14 76.73 153.29
15 81.91 163.78
16 86.80 173.68
17 91.38 182.83
18 95.12 190.52
19 98.93 198.12
20 102.67 205.28
21 105.94 212.20
22 109.96 219.52
23 112.96 225.40
24 116.76 232.64
25 117.69 234.61
26 119.46 237.96
27 120.85 240.56
28 123.62 244.52
29 123.83 247.83
30 125.00 250.41
31 126.63 253.05
32 128.48 255.54
Update
#nm 1 2
1 31.04 62.57
2 41.16 82.75
3 41.85 83.74
4 42.42 85.21
5 42.41 85.46
6 42.75 86.16
7 42.78 86.15
8 42.71 86.02
9 47.72 96.33
10 52.61 106.12
11 57.51 116.27
12 62.41 126.04
13 67.15 135.49
14 71.88 145.10
15 76.27 154.58
16 80.98 165.12
17 84.73 173.57
18 88.52 179.82
19 92.05 188.06
20 95.74 194.69
21 98.60 202.60
22 102.28 209.72
23 105.40 217.46
24 109.15 226.22
25 111.64 232.44
26 114.28 235.60
27 115.96 241.23
28 119.16 244.22
29 120.65 251.44
30 122.83 254.98
31 123.91 260.32
32 127.66 268.63
Triad
#nm 1 2
1 37.19 75.36
2 46.64 93.37
3 46.89 93.65
4 46.67 93.42
5 46.52 93.14
6 46.23 92.63
7 45.92 91.94
8 45.82 91.66
9 51.26 102.53
10 56.45 112.89
11 61.53 122.99
12 66.39 132.84
13 71.25 142.56
14 75.86 151.81
15 80.23 160.68
16 84.47 168.87
17 88.43 176.90
18 92.09 184.11
19 95.64 191.20
20 99.16 197.97
21 102.45 204.88
22 105.69 211.16
23 108.59 216.78
24 111.56 222.79
25 114.27 228.33
26 116.90 233.01
27 119.33 238.29
28 122.01 243.19
29 124.36 248.62
30 126.71 253.58
31 129.03 258.09
32 131.07 261.49
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
