300,000,000 nps ?

[MikeB]

While adding code to display nodes per second for each bench position in the Honey family of engines:

As an example with current-dev-SF

Code: Select all

Stockfish-011220 b >/dev/null

Position: 1/46
Nodes/Second: 1888k

Position: 2/46
Nodes/Second: 2315k

Position: 3/46
Nodes/Second: 2850k

Position: 4/46
Nodes/Second: 2284k

Position: 5/46
Nodes/Second: 1964k

Position: 6/46
Nodes/Second: 1897k

Position: 7/46
Nodes/Second: 2004k

Position: 8/46
Nodes/Second: 2299k

Position: 9/46
Nodes/Second: 2049k

Position: 10/46
Nodes/Second: 2085k

Position: 11/46
Nodes/Second: 1996k

Position: 12/46
Nodes/Second: 2065k

Position: 13/46
Nodes/Second: 2025k

Position: 14/46
Nodes/Second: 2055k

Position: 15/46
Nodes/Second: 2334k

Position: 16/46
Nodes/Second: 2235k

Position: 17/46
Nodes/Second: 2638k

Position: 18/46
Nodes/Second: 2552k

Position: 19/46
Nodes/Second: 3207k

Position: 20/46
Nodes/Second: 3165k

Position: 21/46
Nodes/Second: 3544k

Position: 22/46
Nodes/Second: 3345k

Position: 23/46
Nodes/Second: 4649k

Position: 24/46
Nodes/Second: 2705k

Position: 25/46
Nodes/Second: 2876k

Position: 26/46
Nodes/Second: 2525k

Position: 27/46
Nodes/Second: 2577k

Position: 28/46
Nodes/Second: 2334k

Position: 29/46
Nodes/Second: 2416k

Position: 30/46
Nodes/Second: 2819k

Position: 31/46
Nodes/Second: 2308k

Position: 32/46
Nodes/Second: 2218k

Position: 33/46
Nodes/Second: 2073k

Position: 34/46
Nodes/Second: 1934k

Position: 35/46
Nodes/Second: 3766k

Position: 36/46
Nodes/Second: 3858k

Position: 37/46
Nodes/Second: 3588k

Position: 38/46
Nodes/Second: 3196k

Position: 39/46
Nodes/Second: 3226k

Position: 40/46
Nodes/Second: 4731k

Position: 41/46
Nodes/Second: 3043k

Position: 42/46
Nodes/Second: 1948k

Position: 43/46
Nodes/Second: 1550k

Position: 44/46
Nodes/Second: 0k

Position: 45/46
Nodes/Second: 0k

Position: 46/46
Nodes/Second: 1721k

===========================
Total time (ms) : 1839
Nodes searched  : 4586187
Nodes/second    : 2493848

I noticed this position due to the high nps count.
( Note: I know it is stupid position to include in benchmark - but that's not my call - it is SF benchmark position #37, thus the '?' in my subject heading, it's not a position you would ever be concerned about in a real game)
[d]8/3k4/8/8/8/4B3/4KB2/2B5 w - - 0 1

So I ran it with weakfish and ....

Code: Select all

michaelb7@Threadripper-32:~/Github/stockfish/src$ Weakfish-011320
Weakfish 011320 by M. Byrne and scores of others...
s threads 64 
Confirmation: threads set to 64
s hash 2048                                 
Confirmation: hash set to 2048
p f 8/3k4/8/8/8/4B3/4KB2/2B5 w - - 0 1
go depth 20
...
info depth 20 seldepth 20 multipv 1 score cp 664 nodes 17244699370 nps 318232470 hashfull 134 tbhits 0 time 54189 pv e2f3 d7c7 e3f4 c7d7 f3e3 d7e6 e3e4 e6f7 f4g5 f7g6 g5e3 g6h5 e3f4 h5g4 c1e3 g4h3 f4g5 h3g2 g5e7 g2h2
bestmove e2f3 ponder d7c7

17 billion in < than a minute...

[thunderstruck]

Your hardware is faster than Deep Blue.

[Jouni]

Try faster engine like Houdini 6! It has 50% higher NPS should be 450 000 000 easily . Also Crafty and Deep Fritz 12 have very high speed. SF is only average in node speed.