Engine-Speed for around 100 engines!

[Frank Quisinsky]

Hi there,

a small site to look a bit in Intel i9-10900K 4,9Ghz power for around 100 engines!
I got in the last months different questions to the speed of the machine.
Here the "all-in-one" answer!

Engine-Speed
https://www.amateurschach.de/main/_bench.html

Best
Frank

[Jouni]

Crafty is the speed king: 9.722.869 nps 1 core .

[Pi4Chess]

Nice test !

Crafty is the speed king: 9.722.869 nps 1 core .

So much nodes but do not find the solution ^^
I wonder what could do crafty if he could get NNUE evaluation with its speed (may be not much more since NNUE eval may limit its speed)

[mvanthoor]

Cool

I ran the position with my own fledgling engine, and though it doesn't find a mate (yet), it can still see Bd3xh7 as a tactical shot that gains it some advantage, after 39 seconds. Note that the engine's evaluation is still basically braindead and the search has no optimizations. Stockfish with white absolutely destroys black with a double bishop sacrifice, hitting depth 34 in a few seconds with an evaluation of +18.xx

Even so, I'm somewhat proud that it does find Bd3xh7 as a tactical shot, because I'm convinced that most casual players (and maybe some intermediate club players) wouldn't see it.

Code: Select all

position fen 3rr1k1/1pqb1ppp/p7/2bp2P1/3B1P2/P2B1Q2/1PP4P/4RR1K w - -
board

8   . . . r r . k .
7   . i q b . i i i
6   i . . . . . . .
5   . . b i . . I .
4   . . . B . I . .
3   I . . B . Q . .
2   . I I . . . . I
1   . . . . R R . K

    A B C D E F G H

Zobrist key:        ab4d01d3df86811e
Active Color:       White
Castling:           -
En Passant:         -
Half-move clock:    0
Full-move number:   1

go
info score cp 130 depth 1 seldepth 4 time 0 nodes 136 nps 0 pv f3d5 e8e1 f1e1 c5d4 d5d4
info score cp 130 depth 2 seldepth 3 time 0 nodes 558 nps 0 pv f3d5 e8e1 f1e1 c5d4 d5d4
info score cp 130 depth 3 seldepth 4 time 1 nodes 1917 nps 1917000 pv f3d5 e8e1 f1e1 c5d4 d5d4
info score cp 5 depth 4 seldepth 5 time 8 nodes 16055 nps 2006875 pv e1e8 d8e8 d4e5 c7c6 c2c4
info score cp 10 depth 5 seldepth 7 time 38 nodes 119875 nps 3154605 pv d4e5 c7c6 b2b4 c5b6 f4f5
info score cp 5 depth 6 seldepth 7 time 200 nodes 729050 nps 3645250 pv e1e8 d8e8 f3f2 c5d6 h1g1 d7h3 d4b6
info score cp 10 depth 7 seldepth 9 time 1006 nodes 3592516 nps 3571089 pv e1e8 d8e8 c2c3 c7b6 g5g6 c5d4 g6h7 g8h8 c3d4 b6d4
info score cp 10 depth 8 seldepth 9 time 6259 nodes 22883560 nps 3656105 pv e1e8 d8e8 f3f2 c5d4 f2d4 d7h3 f1c1 h3g4 h1g1
info score cp 75 depth 9 seldepth 11 time 39300 nodes 146894278 nps 3737768 pv d3h7 g8h7 f3h5 h7g8 g5g6 f7g6 h5d5 d7e6 d5c5 c7c5 d4c5 e6d5
info score cp 40 depth 10 seldepth 11 time 231804 nodes 877212039 nps 3784283 pv d3h7 g8h7 f3h5 h7g8 g5g6 f7g6 h5d5 g8h7 d5c5 d7c6 h1g1 e8e1 f1e1 c7f4

[Frank Quisinsky]

Hi there,

yes, such things can be really interesting with a collection of better positions.
And much more interesting if 200, 300 or more engines added!

I do that in this case only for looking in speed information.
Or to give an answer of the question which of the TOP engines support the analyze-mode.

It's really very bad that such great engines like Booot or Fizbo don't support ist.
Both are great for opening analyzes because both engines have a very strong king safty with many pieces on board and plays very aggressive computer chess.

The sources of Fizbo are free on GitHub, sources of Booot are in download file.
I hope the day will come all of the greatest "openings-analyzer" support the important analyze-mode.

Have fun with the small site!
I am happy that I have it in combination with my new configuration site.
So, all important information to all of the engines, using for the FCP Tourney-2020 are from my site open.

Best
Frank

[Pi4Chess]

Xiphos does very well in this position : reaches the most depth and finds the move in 1 sec.

Funny thing is the Rodent IV Tal personality finds the move in 1 sec while the standard Rodent IV does not find anything in 60 sec.

[Frank Quisinsky]

Hi,

yes, I am wonder too.
Or looking in AnMon 5.75 w32 ... a program with 2350 Elo is 6times faster as SF with 1100 Elo more.

OK, Open Tal, Rodent IV 0.30 and Rodent IV 0.30 with Personality Tal are inside the still running FCP Qualify Tourney-2020 in a group of 46 engines, still running in LIVE mode since 12 hours.

Maybe the aggressive personalities from Rodent are more interesting for solving tactical-positions test-suits and not for eng-eng game results? I am not a Rodent expert. Open Tal 1.0 was a secret-hint from TalkChess members. Now we can looking a bit.

Best
Frank

[sedicla]

Hi Frank,

Thanks for running this test. I fixed the issue with "movetime" for Tucano. It was exiting early, but now should use the specified time.
e.g.
uci
position fen 3rr1k1/1pqb1ppp/p7/2bp2P1/3B1P2/P2B1Q2/1PP4P/4RR1K w - -
go movetime 60000


I rerun the tests on my computer and found it in the last moment (using Hash=512). If you want to rerun on your computer let me know I can send you a compiled version for this test.

Code: Select all

info depth 29 score cp 13 time 25187 nodes 56323287 nps 2236204 tbhits 0 pv d4f6 d8c8 b2b4 d5d4 b4c5 d7c6 d3e4 c6e4 e1e4 e8e4 f3e4 g7f6 g5f6 c7c6 f1e1 h7h5 e4c6 c8c6 e1g1 g8f8 g1d1 c6c5 d1d4 f8g8 d4d7 c5c2 d7b7 c2c3 h2h4 c3a3 h1g2 a3a1 b7b8 g8h7 g2f3 h7g6
info depth 30 score cp 25 time 29800 nodes 66499015 nps 2231510 tbhits 498 pv d4f6 d8c8 b2b4 d5d4 b4c5 d7c6 d3e4 c6e4 e1e4 e8e4 f3e4 g7f6 g5f6 c7c6 f1e1 h7h6 e4c6 c8c6 e1g1 g8f8 g1b1 c6f6 b1b7 f6f4 h1g2 f4f5 c5c6 f5c5 c6c7 c5c2 g2f3 f8e8 f3e4 e8d7 e4d4 c2c7 b7c7 d7c7 d4e5 c7c6 e5f6 c6d5 f6f7
info depth 31 score cp 23 time 34062 nodes 76042598 nps 2232476 tbhits 507 pv d4f6 d8c8 b2b4 d5d4 b4c5 d7c6 d3e4 c6e4 e1e4 e8e4 f3e4 g7f6 g5f6 c7c6 f1e1 h7h6 e4c6 c8c6 e1g1 g8f8 g1d1 c6f6 d1d4 f6f5 d4b4 f5c5 b4b7 c5c2 b7b6 f8g7 b6a6 c2c4 f4f5 c4f4 h1g2 f4f5 a3a4 f5c5 a4a5 c5f5 a6a7
info depth 32 score cp 35 time 47795 nodes 107541089 nps 2250048 tbhits 549 pv d3h7
info depth 32 score cp 85 time 51167 nodes 115215429 nps 2251752 tbhits 582 pv d3h7
info depth 32 score cp 285 time 55742 nodes 126190345 nps 2263828 tbhits 606 pv d3h7
bestmove d3h7 ponder d8c8

[Frank Quisinsky]

Hi Alcides,

thats is very friendly and really great if the programmers try to solved such problems.
Tucano is much stronger as I await ... in my still running qualify tourney.

A very strong program from Brazil, pirarucu is strong too.

Alcides, if possible ...
I like more you add a version 9.01 or what you like for a version number on your GitHub,
because since many years I made nothing with private available versions of chess programs.

If available I will changed the test with your new version.
Or too many others asked ... where can I find the Tucano version you have,

THANK you ...
Again, very friendly!!

And good luck for Tucano for the still running tourney.
I think place 1-12 is possible ...

Best
Frank

[sedicla]

Hey Frank,

Tucano's last version can be found on my github.
The last official release is 9.00, but already made 3 updates including this one, so the last development version is 9.03.
I'm planning a new release when I have more improvements, probably it will take some time.

So maybe we can leave this test as it is for now and retest later. I was thinking only if you wanted to retest, but makes sense to test official releases.

Thanks! I will take a look at the tournament.

Alcides.