AMD Rome S2 M4 C32
- Processor: AMD EPYC 7502 32-Core Processor
- Base frequency: 2.5 GHz
- Number of sockets: 2
- Number of memory domains per socket: 4
- Memory domain specs: 2-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 | 36.34 (Triad) |
| Memory domain | 41.73 (Update, 2 cores per domain) |
| Socket | 163.07 (Triad, 2 cores per domain) |
| Node | 324.81 (Triad, 2 cores per domain) |
+-----------------------------------------------------+
Results for scaling inside a memory domain:
#nt Init Sum Copy Update Triad Daxpy STriad SDaxpy
1 23.42 8.26 33.50 32.72 36.35 35.56 35.51 35.75
2 23.43 16.48 40.61 41.74 41.24 40.08 40.50 39.57
3 23.28 24.52 40.42 39.59 39.49 38.06 38.37 37.72
4 23.44 31.98 40.14 39.22 38.80 37.37 37.90 36.98
5 23.43 37.09 39.75 38.96 38.42 37.10 37.51 36.57
6 23.45 36.24 39.19 38.72 37.76 36.37 36.87 35.90
7 23.45 36.36 38.75 37.97 37.07 35.75 36.29 35.29
8 23.45 37.20 38.19 37.40 36.50 35.22 35.73 34.77
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 5 6 7 8
1 23.42 46.82 70.16 79.74 99.81 119.81 139.66 159.14
2 23.43 46.70 70.23 79.92 99.67 119.51 139.46 159.59
3 23.28 46.87 69.89 79.86 99.75 119.48 139.19 158.96
4 23.44 46.77 70.18 79.88 99.89 119.77 139.84 159.70
5 23.43 46.84 70.31 84.92 106.13 127.40 148.59 169.63
6 23.45 46.88 70.32 88.65 110.75 132.94 155.00 177.10
7 23.45 46.89 70.32 91.51 114.30 137.18 160.06 182.57
8 23.45 46.89 70.33 93.75 117.15 140.60 163.95 187.25
Sum:
#nm 1 2 3 4 5 6 7 8
1 8.26 16.52 24.79 32.81 41.39 49.40 57.66 66.04
2 16.48 32.96 49.02 64.83 80.90 97.38 114.04 129.61
3 24.52 49.17 73.71 98.11 122.88 146.83 171.28 195.20
4 31.98 64.00 95.91 127.48 159.19 190.52 222.16 253.20
5 37.09 73.23 110.44 147.55 183.52 219.75 256.83 295.20
6 36.24 72.41 108.81 143.67 177.93 215.59 223.05 285.79
7 36.36 72.20 107.96 144.97 180.26 216.79 249.17 287.27
8 37.20 73.77 111.46 147.51 183.25 220.15 257.48 293.15
Copy
#nm 1 2 3 4 5 6 7 8
1 33.50 66.91 100.47 132.70 167.55 199.58 232.47 265.21
2 40.61 81.09 121.60 151.10 188.59 226.18 263.21 303.59
3 40.42 81.41 121.74 151.59 188.86 227.00 264.38 302.13
4 40.14 80.08 119.91 149.56 186.63 223.53 261.15 298.77
5 39.75 79.37 119.16 152.73 190.83 229.32 268.87 307.99
6 39.19 78.33 117.55 153.00 191.48 229.47 266.36 306.69
7 38.75 77.21 115.66 152.93 191.06 228.90 267.91 307.26
8 38.19 76.35 114.61 151.79 189.71 227.37 265.42 303.71
Update
#nm 1 2 3 4 5 6 7 8
1 32.72 65.49 98.35 131.18 163.67 196.33 225.93 262.11
2 41.74 83.74 125.37 156.84 194.72 233.94 272.10 314.24
3 39.59 80.00 119.68 153.02 191.57 228.93 267.51 306.65
4 39.22 78.42 117.84 150.22 188.30 226.15 263.14 301.78
5 38.96 78.64 119.09 154.84 194.48 234.79 275.15 317.17
6 38.72 77.89 117.87 154.75 194.60 234.80 276.06 317.99
7 37.97 76.54 115.76 153.58 193.61 232.97 274.98 318.20
8 37.40 75.17 114.01 152.87 192.59 233.01 274.53 317.29
Triad
#nm 1 2 3 4 5 6 7 8
1 36.35 73.22 108.15 144.66 178.66 214.39 249.71 284.94
2 41.24 82.67 123.52 163.07 202.38 242.89 282.42 324.82
3 39.49 79.06 118.59 157.37 195.88 234.94 274.72 314.13
4 38.80 77.56 116.18 153.36 191.68 230.00 267.73 306.05
5 38.42 76.66 114.90 152.56 190.77 228.27 267.16 305.33
6 37.76 75.52 113.23 149.97 187.59 225.28 261.59 300.73
7 37.07 73.81 110.76 147.61 184.33 221.30 256.69 294.89
8 36.50 72.87 109.66 145.62 181.58 217.66 254.16 290.76
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:
