Can somebody compare the AMD Ryzen processors to the intel processors

[MikeB]

Are benches here with large pages? 10% speed-up for free with Intel. But I am not sure about AMD.

AMD also increases ~10%, the option is now automatic in Stockfish , but Large Pages still must be set by the user.. Gain varies by number of cores and hash size. Gains are greater with high number of cores and hash sizes that are set at 2048M or multiple thereof.

Current Stockfish, when running bench , does not inform the user if Large Pages are enabled. Also in current Stockifsh, it runs half the positions using classical mode and the half using nnue mode (every other position).

I have designed the Honey engines to run bench slightly different from Stockfish

First Honey engines run a bench of 96 positions, Second, when running bench, it will inform the user if large pages are set or not. Third, the engines can run bench in total NNUE mode or classical mode.

Example:
To run bench in nnue mode:
Harmon-v13-Eval-Threadripper.exe bench 2048 64 16 true > the true indicates nnue mode .
info string Found 510 tablebases
info string Hash Table: Windows Large Pages enabled: 256 Mb
info string Hash Table: Windows Large Pages enabled: 2048 Mb
info string Found 510 tablebases
Position: 1/96
FEN: rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1
info string NNUE evaluation using eval.bin enabled
info depth 1 seldepth 1 multipv 1 score cp 20 nodes 4773 nps 1591000 tbhits 0 time 3 pv c2c4
info depth 2 seldepth 2 multipv 1 score cp 17 nodes 19146 nps 6382000 tbhits 0 time 3 pv e2e4 c7c5
info depth 3 seldepth 3 multipv 1 score cp 9 nodes 43896 nps 14632000 tbhits 0 time 3 pv c2c4 e7e5 b1c3
info depth 4 seldepth 4 multipv 1 score cp 21 nodes 50134 nps 12533500 tbhits 0 time 4 pv e2e4 e7e5 g1f3
info depth 5 seldepth 5 multipv 1 score cp 26 nodes 55502 nps 13875500 tbhits 0 time 4 pv c2c4 g8f6 d2d4
info depth 6 seldepth 6 multipv 1 score cp 49 nodes 76185 nps 19046250 tbhits 0 time 4 pv d2d4 d7d5 g1f3 g8f6
info depth 7 seldepth 7 multipv 1 score cp 24 nodes 87037 nps 21759250 tbhits 0 time 4 pv d2d4 d7d5 g1f3 g8f6 c2c4 e7e6 b1c3
info depth 8 seldepth 8 multipv 1 score cp 23 nodes 98027 nps 24506750 tbhits 0 time 4 pv d2d4 d7d5 c2c4 e7e6 g1f3 g8f6 b1c3 a7a6
info depth 9 seldepth 12 multipv 1 score cp 21 nodes 136197 nps 27239400 tbhits 0 time 5 pv d2d4 d7d5 c2c4 d5c4 g1f3 b7b5 b1c3
info depth 10 seldepth 12 multipv 1 score cp 30 nodes 226862 nps 37810333 tbhits 0 time 6 pv d2d4 d7d5 g1f3 g8f6 c2c4 e7e6 b1c3 c7c5
info depth 11 seldepth 16 multipv 1 score cp 23 nodes 394295 nps 49286875 tbhits 0 time 8 pv d2d4 d7d5 c2c4 e7e6 b1c3 g8f6 g1f3 c7c5 c4d5 c5d4 d1d4 e6d5
info depth 12 seldepth 17 multipv 1 score cp 28 nodes 711947 nps 64722454 tbhits 0 time 11 pv e2e4 e7e5 g1f3 b8c6 d2d4 e5d4 f3d4 g8f6 d4c6 d7c6 d1d8 e8d8 f1d3
info depth 13 seldepth 20 multipv 1 score cp 28 nodes 923820 nps 65987142 tbhits 0 time 14 pv e2e4 e7e5 g1f3 b8c6 d2d4 e5d4 f3d4 g8f6 d4c6 b7c6 f1d3 d7d5 e4d5 c6d5 e1g1 c7c6 c2c4 f8c5 c4d5
info depth 14 seldepth 24 multipv 1 score cp 23 nodes 2021899 nps 74885148 tbhits 0 time 27 pv d2d4 g8f6 c2c4 e7e6 g1f3 d7d5 b1c3 f8b4 c4d5 e6d5 c1f4 e8g8 e2e3 f6e4 a1c1 c7c5 d4c5 b4c3 c1c3
info depth 15 seldepth 23 multipv 1 score cp 29 nodes 3725349 nps 77611437 tbhits 0 time 48 pv d2d4 g8f6 c2c4 e7e6 g1f3 d7d5 b1c3 c7c5 c1g5 d5c4 e2e3 a7a6 f1c4 b7b5 c4e2 c5d4
info depth 16 seldepth 27 multipv 1 score cp 25 nodes 5490902 nps 77336647 tbhits 0 time 71 pv d2d4 d7d5 c2c4 e7e6 b1c3 g8f6 g1f3 c7c5 c1g5 c5d4 f3d4 d5c4 e2e3 c8d7 f1c4 f8e7 e1g1 h7h6 g5h4 b8c6 d4b3
bestmove d2d4 ponder d7d5
Nodes/Second: 77385k
<snip>
Position: 96/96
FEN: 4K1k1/8/1p5p/1Pp3b1/8/1P3P2/P1B2P2/8 w - - 0 1
info string NNUE evaluation using eval.bin enabled
info depth 1 seldepth 1 multipv 1 score cp 94 nodes 34203 nps 11401000 tbhits 0 time 3 pv a2a4 g8g7
info depth 2 seldepth 2 multipv 1 score cp 104 nodes 45000 nps 15000000 tbhits 0 time 3 pv e8d7 g5d2
info depth 3 seldepth 3 multipv 1 score cp 104 nodes 47583 nps 15861000 tbhits 0 time 3 pv e8d7 g5d2 d7c6 d2a5
info depth 4 seldepth 4 multipv 1 score cp 101 nodes 51625 nps 17208333 tbhits 0 time 3 pv a2a4 g5d2 c2e4 g8g7
info depth 5 seldepth 5 multipv 1 score cp 98 nodes 55937 nps 18645666 tbhits 0 time 3 pv e8d7 g5d2 d7c6 d2a5 c6b7 g8f8
info depth 6 seldepth 6 multipv 1 score cp 98 nodes 61287 nps 20429000 tbhits 0 time 3 pv e8d7 g5d2 d7c6 d2a5 c6b7 g8f8
info depth 7 seldepth 7 multipv 1 score cp 98 nodes 67270 nps 22423333 tbhits 0 time 3 pv e8d7 g5d2 d7c6 d2a5 c6b7 g8f8 a2a4
info depth 8 seldepth 9 multipv 1 score cp 94 nodes 96922 nps 32307333 tbhits 0 time 3 pv e8d7 g5d2 d7c7 d2a5 c7b7 g8f8 a2a4
info depth 9 seldepth 13 multipv 1 score cp 104 nodes 238293 nps 59573250 tbhits 0 time 4 pv a2a4 g5d2 e8d7 d2a5 d7e7 g8g7 f3f4 a5c3
info depth 10 seldepth 16 multipv 1 score cp 103 nodes 425949 nps 70991500 tbhits 0 time 6 pv e8d7 g5d2 f3f4 g8g7 d7c7 d2a5 a2a4 h6h5 c2d3 h5h4
info depth 11 seldepth 13 multipv 1 score cp 105 nodes 477433 nps 79572166 tbhits 0 time 6 pv e8d7 g5d2 f3f4 g8g7 d7c7 d2a5 c2e4 h6h5 e4d3 h5h4 d3f5 g7f6 f5g4
info depth 12 seldepth 15 multipv 1 score cp 114 nodes 833745 nps 92638333 tbhits 57 time 9 pv e8d7 g5d2 a2a3 d2a5 d7c6 g8f7 b3b4 c5b4 a3b4 a5b4 c6b6 b4e1 b6b7 e1f2
info depth 13 seldepth 18 multipv 1 score cp 109 nodes 958978 nps 95897800 tbhits 166 time 10 pv e8d7 g5d2 d7c7 d2a5 a2a3 h6h5 b3b4 c5b4 a3b4 a5b4 c7b6 h5h4 c2f5 g8g7 b6a6 g7f6 f5c8
info depth 14 seldepth 22 multipv 1 score cp 101 nodes 1474037 nps 105288357 tbhits 270 time 14 pv e8d7 g5d2 f3f4 g8g7 d7c6 d2a5 a2a3 h6h5 c6b7 h5h4 c2f5 g7f7 f5g4 f7f6 b3b4 c5b4 a3b4 a5b4 b7b6
info depth 15 seldepth 20 multipv 1 score cp 99 nodes 2157266 nps 107863300 tbhits 520 time 20 pv e8d7 g5d2 d7c6 d2a5 f3f4 g8g7 a2a3 h6h5 c2f5 g7f6 f5h3 f6g7 c6c7 g7f7 h3c8 f7g6 f4f5 g6f7
info depth 16 seldepth 20 multipv 1 score cp 115 nodes 3064294 nps 117857461 tbhits 745 time 26 pv e8d7 g5d2 d7c7 d2a5 a2a3 g8g7 f3f4 g7f6 c2e4 h6h5 e4f3 f6g6 f3d1
bestmove e8d7 ponder g5d2
Nodes/Second: 118004k
NN evaluation using eval.bin enabled.

=================================
Total time (ms) : 8139
Nodes searched : 933394796

Nodes/second : 114681k

To run bench in classical mode:
Harmon-v13-Eval-Threadripper.exe bench 2048 64 16 false


Harmon v13 by The Stockfish Developers as noted in the AUTHORS file
info string Found 510 tablebases
info string Hash Table: Windows Large Pages enabled: 256 Mb
info string Hash Table: Windows Large Pages enabled: 2048 Mb
info string Found 510 tablebases

Position: 1/96
FEN: rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1
info string classical evaluation enabled
info depth 1 seldepth 1 multipv 1 score cp 88 nodes 6044 nps 2014666 tbhits 0 time 3 pv e2e3
info depth 2 seldepth 2 multipv 1 score cp 107 nodes 16996 nps 5665333 tbhits 0 time 3 pv e2e4 e7e6
info depth 3 seldepth 3 multipv 1 score cp 67 nodes 22161 nps 7387000 tbhits 0 time 3 pv d2d4 d7d5 b1c3
info depth 4 seldepth 4 multipv 1 score cp 71 nodes 25479 nps 8493000 tbhits 0 time 3 pv d2d4 d7d5 b1c3 g8f6
info depth 5 seldepth 5 multipv 1 score cp 61 nodes 53895 nps 17965000 tbhits 0 time 3 pv b1c3 d7d5 d2d4 b8c6
info depth 6 seldepth 6 multipv 1 score cp 58 nodes 65370 nps 21790000 tbhits 0 time 3 pv g1f3 d7d5 d2d4 c8f5 c1f4 e7e6
info depth 7 seldepth 7 multipv 1 score cp 54 nodes 85440 nps 21360000 tbhits 0 time 4 pv b1c3 b8c6 g1f3 d7d5 d2d4 c8f5 f3h4
info depth 8 seldepth 9 multipv 1 score cp 52 nodes 154743 nps 38685750 tbhits 0 time 4 pv e2e4 d7d5 e4d5 d8d5 b1c3 d5e6 d1e2 g8f6
info depth 9 seldepth 12 multipv 1 score cp 52 nodes 203385 nps 40677000 tbhits 0 time 5 pv d2d4 d7d5 e2e3 g8f6 g1f3 c7c6 b1d2 b8d7 c2c3
info depth 10 seldepth 16 multipv 1 score cp 61 nodes 302087 nps 50347833 tbhits 0 time 6 pv e2e4 e7e5 g1f3 b8c6 b1c3 f8c5 d1e2 d7d6
info depth 11 seldepth 15 multipv 1 score cp 59 nodes 583671 nps 64852333 tbhits 0 time 9 pv e2e4 e7e6 b1c3 d7d5 e4d5 e6d5 d1e2 c8e6 d2d4 b8c6
info depth 12 seldepth 17 multipv 1 score cp 58 nodes 1294357 nps 80897312 tbhits 0 time 16 pv e2e4 d7d5 e4d5 c7c6 d5c6 b8c6 g1f3 e7e5 b1c3 g8f6 f1b5 f8d6 d2d3 e8g8
info depth 13 seldepth 17 multipv 1 score cp 53 nodes 2649977 nps 82811781 tbhits 0 time 32 pv d2d4 g8f6 c2c4 e7e6 b1c3 d7d5 e2e3 b8c6 a2a3 h7h5 c4d5 e6d5 g1f3
info depth 14 seldepth 17 multipv 1 score cp 64 nodes 2968367 nps 82454638 tbhits 0 time 36 pv e2e4 e7e5 g1f3 g8f6 b1c3 b8c6 f1c4 f8c5 d2d3 d7d6 c1g5 a7a6 e1g1 b7b5 g5f6 d8f6
info depth 15 seldepth 29 multipv 1 score cp 54 nodes 5115813 nps 82513112 tbhits 0 time 62 pv e2e4 e7e5 b1c3 g8f6 g1f3 b8c6 f1c4 f8c5 e1g1 d7d6 d2d3 c6a5 b2b3 e8g8 c3a4 a5c4 b3c4
info depth 16 seldepth 21 multipv 1 score cp 58 nodes 5802102 nps 82887171 tbhits 0 time 70 pv e2e4 e7e5 g1f3 b8c6 f1b5 g8f6 e1g1 f8d6 d2d3 e8g8 b5c4 h7h6 b1c3 d6c5 c3d5 d7d6 c2c3 f6d5 c4d5
bestmove e2e4 ponder e7e5
Nodes/Second: 82943k
Classical evaluation enabled.
<snip>
Position: 96/96
FEN: 4K1k1/8/1p5p/1Pp3b1/8/1P3P2/P1B2P2/8 w - - 0 1
info string classical evaluation enabled
info depth 1 seldepth 1 multipv 1 score cp 104 nodes 24332 nps 8110666 tbhits 0 time 3 pv e8d7
info depth 2 seldepth 2 multipv 1 score cp 107 nodes 28155 nps 9385000 tbhits 0 time 3 pv a2a4 g5d2
info depth 3 seldepth 3 multipv 1 score cp 118 nodes 31171 nps 10390333 tbhits 0 time 3 pv e8d7 g5d2 d7c6
info depth 4 seldepth 4 multipv 1 score cp 107 nodes 35247 nps 11749000 tbhits 0 time 3 pv e8d7 g5d2 d7c6 d2a5
info depth 5 seldepth 5 multipv 1 score cp 104 nodes 44768 nps 14922666 tbhits 0 time 3 pv e8d7 g5d2 d7c6 d2a5
info depth 6 seldepth 6 multipv 1 score cp 104 nodes 49613 nps 16537666 tbhits 0 time 3 pv e8d7 g5d2 d7c6 d2a5 f3f4 h6h5
info depth 7 seldepth 7 multipv 1 score cp 110 nodes 55999 nps 18666333 tbhits 0 time 3 pv e8d7 g5d2 d7c6 d2a5 c2e4 h6h5
info depth 8 seldepth 9 multipv 1 score cp 91 nodes 138222 nps 34555500 tbhits 0 time 4 pv e8d7 g5d2 d7c6 d2a5 c2d3 h6h5 a2a4 h5h4
info depth 9 seldepth 12 multipv 1 score cp 83 nodes 197180 nps 49295000 tbhits 0 time 4 pv e8d7 g5d2 f3f4 h6h5 d7c7 d2a5 c2e4 h5h4 e4g2 g8g7 a2a4
info depth 10 seldepth 13 multipv 1 score cp 96 nodes 310538 nps 62107600 tbhits 7 time 5 pv e8d7 g5d2 d7c6 d2a5 a2a3 g8f8 f3f4 h6h5 b3b4 c5b4
info depth 11 seldepth 12 multipv 1 score cp 107 nodes 342365 nps 68473000 tbhits 22 time 5 pv e8d7 g5d2 d7c6 d2a5 a2a3 g8f8 b3b4 c5b4 a3b4
info depth 12 seldepth 16 multipv 1 score cp 101 nodes 624400 nps 78050000 tbhits 371 time 8 pv e8d7 g5d2 d7c7 d2a5 a2a3 h6h5 b3b4 c5b4 a3b4 a5b4 c7b6 b4e1 b6c6 e1f2 b5b6 f2g3
info depth 13 seldepth 19 multipv 1 score cp 91 nodes 3542132 nps 118071066 tbhits 1848 time 30 pv e8d7 g5d2 d7e6 g8g7 a2a4 h6h5 e6f5 g7h6 f3f4 d2c3 f5e6 h5h4 f4f5 h6g7 c2e4
info depth 14 seldepth 15 multipv 1 score cp 90 nodes 4050673 nps 119137441 tbhits 1853 time 34 pv e8d7 g5d2 d7e6 h6h5 e6f5 g8f7 a2a4 f7e7 c2e4 h5h4 f5g4 d2e1 f3f4 e1f2
info depth 15 seldepth 19 multipv 1 score cp 89 nodes 4672310 nps 122955526 tbhits 1854 time 38 pv e8d7 g5d2 d7e6 h6h5 e6f6 g8f8 a2a4 h5h4 c2f5 d2e1 f3f4 e1d2 f6g5 f8e7 f5g4 e7d6 g5h4 d2f4
info depth 16 seldepth 20 multipv 1 score cp 88 nodes 5695955 nps 123825108 tbhits 1873 time 46 pv e8d7 g5d2 d7e6 h6h5 e6f5 g8f7 a2a4 h5h4 f5g4 d2e1 f3f4 f7f6 c2d3 e1d2 g4h4 d2f4 h4g4 f4d2
bestmove e8d7 ponder g5d2
Nodes/Second: 123940k
Classical evaluation enabled.

=================================
Total time (ms) : 9380
Nodes searched : 1154656953

Nodes/second : 123097k

Download & Source

[Milos]

Seriously?

You're the only one here with a ratio far above 100% amongs a few users with an AMD processor. MikeB got 92%. I get 91%, Elpapa 90%, Modern Times 94%. You get 131% and you think that perhaps other's results are wrong?

I measured the right thing (to show, why the Ipman speed measurements with a non nnue-engine (asmFish) are outdated), but I wrote nonsense:

What I measured (and what has to be measured) was the speed gain from popcount binary to bmi2/avx2 binary. Not comparing avx2-speed of nnue on/off (that makes no sense at all). Sorry for that.

So, here is the real point, the speedups from popcount to bmi2/avx2:

Intel Haswell:
nnue off:
popcount -> bmi2 : +4.3%

nnue on:
popcount -> bmi2 : +22%


AMD Ryzen 3900:
nnue off:
popcount -> avx2 : +0.9%

nnue on:
popcount -> avx2 : +30%


So, nnue engines run measureable faster on AMD Ryzen/Threadripper, when the CPU is on a corresponding GHz-speed level than a Intel CPU, but slower, if the engine uses no nnue. Because the speedup from the base version (popcount) to the CPU-type optimized version (bmi2 for Intel, avx2 for AMD) is higher for AMD/avx2, if nnue is on and higher for Intel/bmi2, if nnue is off.
And because of this (and the fact, that nnue-engines have taken over), the Ipman-speed measurements using old asmFish are outdated.

So you discovered that NNUE that has handcrafted SIMD instructions implementation runs noticeably faster on CPUs that have support for those when the particular part of SIMD optimized code is enabled?!?!
Hm, congratulations, you've reinvented the wheel.

[Gian-Carlo Pascutto]

Table updated. Zen 1 (R7 1700) is ~1520000 but I didn't write down the exact numbers. I don't have anything from "modern" Intel though.

AMD Ryzen 7 5800X, GCC 10.2, Stockfish 13 BMI2
Total time (ms) : 1140
Nodes searched : 3766422
Nodes/second : 3303878

AMD Ryzen 9 3900X, GCC 10.2, Stockfish 13 AVX2
Total time (ms) : 1465
Nodes searched : 3766422
Nodes/second : 2570936

AMD Ryzen 7 2700X, GCC 10.2, Stockfish 13 AVX2
Total time (ms) : 1888
Nodes searched : 3766422
Nodes/second : 1994926

Intel Xeon E3-1240 v3 (Haswell), GCC 10.2, Stockfish 13 AVX2
Total time (ms) : 1838
Nodes searched : 3766422
Nodes/second : 2049195

Intel i7-3770K, GCC 10.2, Stockfish 13 Modern
Total time (ms) : 2114
Nodes searched : 3766422
Nodes/second : 1781656

Apple M1, Clang 12.0, Stockfish 13 NEON
Total time (ms) : 1512
Nodes searched : 3766422
Nodes/second : 2491019

[Chessqueen]

Table updated. Zen 1 (R7 1700) is ~1520000 but I didn't write down the exact numbers. I don't have anything from "modern" Intel though.

AMD Ryzen 7 5800X, GCC 10.2, Stockfish 13 BMI2
Total time (ms) : 1140
Nodes searched : 3766422
Nodes/second : 3303878

AMD Ryzen 9 3900X, GCC 10.2, Stockfish 13 AVX2
Total time (ms) : 1465
Nodes searched : 3766422
Nodes/second : 2570936

AMD Ryzen 7 2700X, GCC 10.2, Stockfish 13 AVX2
Total time (ms) : 1888
Nodes searched : 3766422
Nodes/second : 1994926

Intel Xeon E3-1240 v3 (Haswell), GCC 10.2, Stockfish 13 AVX2
Total time (ms) : 1838
Nodes searched : 3766422
Nodes/second : 2049195

Intel i7-3770K, GCC 10.2, Stockfish 13 Modern
Total time (ms) : 2114
Nodes searched : 3766422
Nodes/second : 1781656

Apple M1, Clang 12.0, Stockfish 13 NEON
Total time (ms) : 1512
Nodes searched : 3766422
Nodes/second : 2491019

Where do you get the latest benchmarck file to test these computers, where is the link

[cma6]

Mike B:
Mike, I was looking in my Engine Personality for SF 260221 in Aquarium running on dual Xeon system, and I did not see Large Pages options. Does that mean that Large Pages option would come up only on an AMD system?

[Dann Corbit]

Mike B:
Mike, I was looking in my Engine Personality for SF 260221 in Aquarium running on dual Xeon system, and I did not see Large Pages options. Does that mean that Large Pages option would come up only on an AMD system?

You have to enable large pages. There is information on this page about how you can enable large pages on Windows 10:
https://www.chaoticafractals.com/manual ... rt-windows

Or you can just google it yourself.

[Gian-Carlo Pascutto]

Where do you get the latest benchmarck file to test these computers, where is the link

You just run "bench" in the executable: Best to do it this way https://github.com/official-stockfish/S ... -775835158

These were Linux machines so I compiled from source, but using the Stockfish 13 from stockfishchess.org would also work.

[cma6]

Mike B:
Mike, I was looking in my Engine Personality for SF 260221 in Aquarium running on dual Xeon system, and I did not see Large Pages options. Does that mean that Large Pages option would come up only on an AMD system?

You have to enable large pages. There is information on this page about how you can enable large pages on Windows 10:
https://www.chaoticafractals.com/manual ... rt-windows

Thanks, Dan. I've enabled Large Pages as per instructions and am now running Aquarium in Admin mode. In SF parameters, there is no indication of Large Pages enabled, so do I take it on faith that LP are enabled for Aquarium/SF on a dual-Xeon (Intel) system with 36 cores? And if LPF are set, what difference should I see in running i.a.? I have been using 4096 MB hash for SF in Aquarium.

Mike B. said "Large Pages still must be set by the user. Gain varies by number of cores and hash size. Gains are greater with high number of cores and hash sizes that are set at 2048M or multiple thereof."

[towforce]

Can somebody compare the AMD Ryzen processors to the intel processors engine benchmarks, I do NOT want to see any other processors on the list.

For instance:
Ryzen 5 3600 and Next to it the intel i5
AMD Ryzen 7 5800X ‎and Next to it the intel i7
AMD Ryzen 9 5950X Next to it the intel i9

I'll do the first of those for you - "Ryzen 5 3600 and Next to it the intel i5".

I don't have either a Ryzen 5 3600 or an Intel I5 - but that's not going to stop me! You see, other people do have them, and they publish the GeekBench results!

Ryzen 5 3600 - link.

Single core: 1206
Multi core: 6689


Intel i5 - link. (there are many i5 models: I picked the 10600K)

Single core: 1311
Multi core: 6975

[Dann Corbit]

Mike B:
Mike, I was looking in my Engine Personality for SF 260221 in Aquarium running on dual Xeon system, and I did not see Large Pages options. Does that mean that Large Pages option would come up only on an AMD system?

You have to enable large pages. There is information on this page about how you can enable large pages on Windows 10:
https://www.chaoticafractals.com/manual ... rt-windows

Thanks, Dan. I've enabled Large Pages as per instructions and am now running Aquarium in Admin mode. In SF parameters, there is no indication of Large Pages enabled, so do I take it on faith that LP are enabled for Aquarium/SF on a dual-Xeon (Intel) system with 36 cores? And if LPF are set, what difference should I see in running i.a.? I have been using 4096 MB hash for SF in Aquarium.

Mike B. said "Large Pages still must be set by the user. Gain varies by number of cores and hash size. Gains are greater with high number of cores and hash sizes that are set at 2048M or multiple thereof."

The large pages will be an engine UCI parameter for engines that support them.
You will see when the engine starts up if there is a large pages failure, but it will still load with regular memory use.
Sometimes, large pages will fail simply because memory is fragmented or for some other reason the requested large pages cannot be allocated.
In these cases, the engine will fall back to ordinary allocation.