Dev Builds » 20230223-1222

You are viewing an old NCM Stockfish dev build test. You may find the most recent dev build tests using Stockfish 15 as the baseline here.

Use this dev build

NCM plays each Stockfish dev build 20,000 times against Stockfish 14. This yields an approximate Elo difference and establishes confidence in the strength of the dev builds.

Summary

Host Duration Avg Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo
ncm-dbt-01 05:02:13 1108287 1690 721 171 798 +117.34 ± 7.73 0 29 259 535 22 +254.64 ± 21.21
ncm-dbt-02 04:59:57 1234041 1654 674 180 800 +107.03 ± 7.92 1 37 270 505 14 +231.01 ± 20.77
ncm-dbt-03 05:02:20 1237516 1660 650 168 842 +103.87 ± 7.66 0 32 297 488 13 +222.44 ± 19.77
ncm-dbt-04 05:01:35 1237071 1664 676 166 822 +110.02 ± 7.79 1 29 279 505 18 +236.81 ± 20.42
ncm-dbt-05 05:00:00 1222796 1666 687 159 820 +114.04 ± 7.89 0 33 260 519 21 +245.55 ± 21.17
ncm-dbt-06 05:01:56 1230462 1666 667 157 842 +109.88 ± 7.71 1 25 289 499 19 +235.78 ± 20.03
ncm-et-3 06:25:15 1299693 1668 684 178 806 +108.82 ± 7.8 1 28 289 496 20 +232.23 ± 20.05
ncm-et-4 06:27:18 1305708 1668 672 162 834 +109.74 ± 7.75 0 36 265 520 13 +238.62 ± 20.97
ncm-et-9 06:24:43 1296567 1670 685 166 819 +111.67 ± 7.51 0 25 281 514 15 +242.78 ± 20.32
ncm-et-10 06:28:37 1294768 1668 685 155 828 +114.35 ± 7.78 2 23 273 515 21 +247.81 ± 20.63
ncm-et-13 06:26:35 1309983 1664 676 161 827 +111.17 ± 7.97 0 38 260 515 19 +238.74 ± 21.16
ncm-et-15 06:28:24 1290193 1662 653 171 838 +103.74 ± 7.91 0 37 293 483 18 +219.04 ± 19.93
20000 8130 1994 9876 +110.14 ± 2.25 6 372 3315 6094 213 +236.99 ± 5.91

Test Detail

ID Host Base NPS Games WLD Standard Elo Ptnml(0-2) Gamepair Elo CLI PGN
189132 ncm-dbt-02 1229118 154 58 16 80 +97.21 ± 24.92 0 4 27 46 0 +212.59 ± 67.07
189131 ncm-dbt-03 1223095 160 54 19 87 +77.25 ± 25.21 0 6 33 41 0 +162.99 ± 60.19
189130 ncm-dbt-04 1237325 164 71 17 76 +118.82 ± 24.89 0 3 24 53 2 +259.99 ± 71.67
189129 ncm-dbt-06 1240991 166 66 15 85 +110.3 ± 27.12 1 3 25 52 2 +242.14 ± 69.92
189128 ncm-dbt-05 1228719 166 75 18 73 +124.34 ± 25.47 0 2 26 51 4 +262.57 ± 68.6
189127 ncm-dbt-01 1110149 190 83 18 89 +123.85 ± 25.26 0 4 27 59 5 +258.51 ± 67.23
189126 ncm-dbt-05 1236326 500 216 49 235 +120.67 ± 14.35 0 9 72 162 7 +263.42 ± 40.52
189125 ncm-dbt-03 1235699 500 201 52 247 +106.77 ± 13.06 0 6 91 151 2 +234.38 ± 35.72
189124 ncm-dbt-04 1244046 500 199 51 250 +106.01 ± 14.04 0 8 92 144 6 +223.94 ± 35.61
189123 ncm-dbt-02 1238203 500 208 44 248 +118.33 ± 14.36 0 8 78 156 8 +254.16 ± 38.89
189122 ncm-dbt-06 1214199 500 209 43 248 +119.89 ± 13.33 0 6 76 164 4 +268.17 ± 39.39
189121 ncm-dbt-01 1112342 500 211 42 247 +122.24 ± 14.33 0 8 73 161 8 +265.78 ± 40.25
189120 ncm-dbt-02 1225451 500 205 67 228 +98.44 ± 14.59 0 14 88 144 4 +207.95 ± 36.54
189119 ncm-dbt-05 1221567 500 202 48 250 +110.6 ± 14.53 0 11 80 153 6 +236.51 ± 38.38
189118 ncm-dbt-04 1228592 500 206 51 243 +111.37 ± 14.84 1 10 78 155 6 +240.82 ± 38.87
189117 ncm-dbt-03 1237879 500 190 46 264 +102.97 ± 14.79 0 11 92 139 8 +211.87 ± 35.7
189116 ncm-dbt-06 1254591 500 194 48 258 +104.49 ± 14.04 0 8 94 142 6 +219.87 ± 35.19
189115 ncm-dbt-01 1105132 500 206 58 236 +106.01 ± 13.39 0 8 88 152 2 +232.26 ± 36.48
189114 ncm-dbt-04 1238324 500 200 47 253 +109.83 ± 13.73 0 8 85 153 4 +238.66 ± 37.17
189113 ncm-dbt-02 1243393 500 203 53 244 +107.54 ± 14.37 1 11 77 159 2 +238.66 ± 39.1
189112 ncm-dbt-05 1204575 500 194 44 262 +107.54 ± 14.21 0 11 82 153 4 +232.26 ± 37.9
189111 ncm-dbt-06 1212070 500 198 51 251 +105.25 ± 14.2 0 8 94 141 7 +219.87 ± 35.19
189110 ncm-dbt-03 1253393 500 205 51 244 +110.6 ± 13.73 0 9 81 157 3 +243.0 ± 38.14
189109 ncm-dbt-01 1105528 500 221 53 226 +121.45 ± 14.34 0 9 71 163 7 +265.78 ± 40.81
167114 ncm-et-3 1290962 168 74 19 75 +118.08 ± 23.64 0 2 27 53 2 +258.14 ± 67.17
167113 ncm-et-15 1292236 162 62 13 87 +108.48 ± 25.13 0 4 25 51 1 +236.83 ± 69.91
167112 ncm-et-10 1289250 168 71 14 83 +122.74 ± 23.55 0 2 25 55 2 +272.25 ± 70.16
167111 ncm-et-4 1314438 168 65 19 84 +97.62 ± 25.91 0 5 30 47 2 +202.06 ± 63.42
167110 ncm-et-13 1306092 164 65 21 78 +95.55 ± 25.52 0 5 29 47 1 +202.34 ± 64.54
167109 ncm-et-9 1290644 170 65 13 92 +109.78 ± 23.51 0 2 31 50 2 +234.51 ± 62.07
167060 ncm-et-15 1268846 500 203 54 243 +106.77 ± 13.89 0 6 96 141 7 +223.94 ± 34.66
167059 ncm-et-10 1293337 500 194 47 259 +105.25 ± 13.05 0 7 90 152 1 +232.26 ± 35.99
167058 ncm-et-4 1309547 500 202 53 245 +106.77 ± 14.36 0 14 75 159 2 +234.38 ± 39.52
167057 ncm-et-13 1299765 500 193 44 263 +106.77 ± 14.52 0 11 85 148 6 +226.0 ± 37.21
167056 ncm-et-3 1300987 500 202 48 250 +110.6 ± 14.69 0 10 84 148 8 +232.26 ± 37.43
167055 ncm-et-9 1297423 500 204 48 248 +112.14 ± 14.06 0 10 78 158 4 +245.2 ± 38.89
167054 ncm-et-10 1303197 500 220 41 239 +130.14 ± 13.52 0 4 71 167 8 +290.66 ± 40.76
167053 ncm-et-4 1298273 500 193 42 265 +108.3 ± 14.37 0 11 82 152 5 +232.26 ± 37.9
167052 ncm-et-15 1296265 500 191 43 266 +106.01 ± 14.51 0 11 86 147 6 +223.94 ± 36.98
167051 ncm-et-13 1317587 500 207 47 246 +115.22 ± 13.72 0 8 78 160 4 +254.16 ± 38.89
167050 ncm-et-3 1310687 500 203 55 242 +106.01 ± 13.88 0 8 91 146 5 +226.0 ± 35.82
167049 ncm-et-9 1298338 500 201 50 249 +108.3 ± 13.23 0 6 90 151 3 +236.51 ± 35.94
167048 ncm-et-10 1293290 500 200 53 247 +105.25 ± 15.83 2 10 87 141 10 +217.85 ± 36.77
167047 ncm-et-9 1299866 500 215 55 230 +115.22 ± 13.88 0 7 82 155 6 +249.64 ± 37.86
167046 ncm-et-3 1296137 500 205 56 239 +106.78 ± 14.36 1 8 87 149 5 +230.16 ± 36.73
167045 ncm-et-4 1300576 500 212 48 240 +118.33 ± 13.35 0 6 78 162 4 +263.42 ± 38.85
167044 ncm-et-15 1303425 500 197 61 242 +96.94 ± 14.87 0 16 86 144 4 +204.07 ± 36.93
167043 ncm-et-13 1316491 500 211 49 240 +116.77 ± 15.31 0 14 68 160 8 +249.64 ± 41.39

Commit

Commit ID b4ad3a3c4b68f9c8736f444aeb3364f833247fdc
Author Sebastian Buchwald
Date 2023-02-23 12:22:03 UTC
Add support for ARM dot product instructions The sdot instruction computes (and accumulates) a signed dot product, which is quite handy for Stockfish's NNUE code. The instruction is optional for Armv8.2 and Armv8.3, and mandatory for Armv8.4 and above. The commit adds a new 'arm-dotprod' architecture with enabled dot product support. It also enables dot product support for the existing 'apple-silicon' architecture, which is at least Armv8.5. The following local speed test was performed on an Apple M1 with ARCH=apple-silicon. I had to remove CPU pinning from the benchmark script. However, the results were still consistent: Checking both binaries against themselves reported a speedup of +0.0000 and +0.0005, respectively. ``` Result of 100 runs ================== base (...ish.037ef3e1) = 1917997 +/- 7152 test (...fish.dotprod) = 2159682 +/- 9066 diff = +241684 +/- 2923 speedup = +0.1260 P(speedup > 0) = 1.0000 CPU: 10 x arm Hyperthreading: off ``` Fixes #4193 closes https://github.com/official-stockfish/Stockfish/pull/4400 No functional change
Copyright 2011–2024 Next Chess Move LLC