Maximum speed
Moderator: Ras
-
- Posts: 7251
- Joined: Mon May 27, 2013 10:31 am
Maximum speed
What's the maximum speed ever say measured relative to speed of Fairy max 4.8 on the same computer using one thread ? Requirement: the chess engine should play or look like it plays normal chess.
-
- Posts: 7251
- Joined: Mon May 27, 2013 10:31 am
Re: Maximum speed
If the question is too difficult to answer: what is at this moment the fastest engine measured in nodes / second.
Slowest is probably Skipper with speed four or five times less than Fairy Max although I improved it lately.
Slowest is probably Skipper with speed four or five times less than Fairy Max although I improved it lately.
-
- Posts: 2665
- Joined: Fri Nov 26, 2010 2:00 pm
- Location: Czech Republic
- Full name: Martin Sedlak
Re: Maximum speed
It depends, problem is many engines count nodes differently (some count moves made etc.)Henk wrote:What's the maximum speed ever say measured relative to speed of Fairy max 4.8 on the same computer using one thread ? Requirement: the chess engine should play or look like it plays normal chess.
From what I've seen, Crafty is very fast in terms of raw nps, so perhaps that's the engine you're looking for.
-
- Posts: 7251
- Joined: Mon May 27, 2013 10:31 am
Re: Maximum speed
Ok if Crafty is 5 times faster than Stockfish and looks like Stockfish is 1.5 or 2 times faster than Fairy max on my computer then Crafty is 7.5 to 10 times faster than Fairy max.
So maximum is 10 and minimum 0.2 multiplied by 10 gives speed range [2, 100].
Current speed of Skipper is at this moment two times speed of Fairy max = 20. But I removed almost all evaluation code. So it's artificial and temporary.
So maximum is 10 and minimum 0.2 multiplied by 10 gives speed range [2, 100].
Current speed of Skipper is at this moment two times speed of Fairy max = 20. But I removed almost all evaluation code. So it's artificial and temporary.
-
- Posts: 793
- Joined: Sun Aug 03, 2014 4:48 am
- Location: London, UK
Re: Maximum speed
I happen to have quite a few engines in my gauntlet, so I ran a search from start position on all of them. All single threaded.Henk wrote:Ok if Crafty is 5 times faster than Stockfish and looks like Stockfish is 1.5 or 2 times faster than Fairy max on my computer then Crafty is 7.5 to 10 times faster than Fairy max.
So maximum is 10 and minimum 0.2 multiplied by 10 gives speed range [2, 100].
Current speed of Skipper is at this moment two times speed of Fairy max = 20. But I removed almost all evaluation code. So it's artificial and temporary.
Arasan: 878 knps
Betsabe II: 1339 knps
Cheng: 1307 knps
Crafty: 2434 knps
Diablo: 966 knps
EXchess: 749 knps
Fairymax: 677 knps
GNU Chess 6: 1221 knps
GreKo: 1567 knps
Phalanx: 752 knps
RobboLito: 1551 knps
Scorpio: 1071 knps
Senpai: 717 knps
Sloppy: 1407 knps
Smash: 587 knps
Stockfish: 1110 knps
Sungorus: 3074 knps
Texel: 1016 knps
ZCT: 752 knsp
And of course my own engines:
Brainless: 1291 knps
Giraffe: 126 knps
I didn't realize my old engine is so fast. Maybe that's why it's already beaten by Giraffe by almost 100 Elo

Disclosure: I work for DeepMind on the AlphaZero project, but everything I say here is personal opinion and does not reflect the views of DeepMind / Alphabet.
-
- Posts: 7251
- Joined: Mon May 27, 2013 10:31 am
Re: Maximum speed
It's easier if divided by 100 knps and figures are rounded. So we get:matthewlai wrote:I happen to have quite a few engines in my gauntlet, so I ran a search from start position on all of them. All single threaded.Henk wrote:Ok if Crafty is 5 times faster than Stockfish and looks like Stockfish is 1.5 or 2 times faster than Fairy max on my computer then Crafty is 7.5 to 10 times faster than Fairy max.
So maximum is 10 and minimum 0.2 multiplied by 10 gives speed range [2, 100].
Current speed of Skipper is at this moment two times speed of Fairy max = 20. But I removed almost all evaluation code. So it's artificial and temporary.
Arasan: 878 knps
Betsabe II: 1339 knps
Cheng: 1307 knps
Crafty: 2434 knps
Diablo: 966 knps
EXchess: 749 knps
Fairymax: 677 knps
GNU Chess 6: 1221 knps
GreKo: 1567 knps
Phalanx: 752 knps
RobboLito: 1551 knps
Scorpio: 1071 knps
Senpai: 717 knps
Sloppy: 1407 knps
Smash: 587 knps
Stockfish: 1110 knps
Sungorus: 3074 knps
Texel: 1016 knps
ZCT: 752 knsp
And of course my own engines:
Brainless: 1291 knps
Giraffe: 126 knps
I didn't realize my old engine is so fast. Maybe that's why it's already beaten by Giraffe by almost 100 Elo.
Arasan: 9
Betsabe II: 13
Cheng: 13
Crafty: 24
Diablo: 10
EXchess: 7
Fairymax: 7
GNU Chess 6: 12
GreKo: 16
Phalanx: 8
RobboLito: 16
Scorpio: 11
Senpai: 7
Sloppy: 14
Smash: 6
Stockfish: 11
Sungorus: 31
Texel: 10
ZCT: 8
Brainless: 13
Giraffe: 1
Might be also interesting to know if they use pure bitboards or not. Perhaps if you use a hybrid design you may never reach top speed. By the way what was it twice the speed = 70 ELO ?
-
- Posts: 7251
- Joined: Mon May 27, 2013 10:31 am
Re: Maximum speed
And divide by 7 to get the stars:Henk wrote:It's easier if divided by 100 knps and figures are rounded. So we get:matthewlai wrote:I happen to have quite a few engines in my gauntlet, so I ran a search from start position on all of them. All single threaded.Henk wrote:Ok if Crafty is 5 times faster than Stockfish and looks like Stockfish is 1.5 or 2 times faster than Fairy max on my computer then Crafty is 7.5 to 10 times faster than Fairy max.
So maximum is 10 and minimum 0.2 multiplied by 10 gives speed range [2, 100].
Current speed of Skipper is at this moment two times speed of Fairy max = 20. But I removed almost all evaluation code. So it's artificial and temporary.
Arasan: 878 knps
Betsabe II: 1339 knps
Cheng: 1307 knps
Crafty: 2434 knps
Diablo: 966 knps
EXchess: 749 knps
Fairymax: 677 knps
GNU Chess 6: 1221 knps
GreKo: 1567 knps
Phalanx: 752 knps
RobboLito: 1551 knps
Scorpio: 1071 knps
Senpai: 717 knps
Sloppy: 1407 knps
Smash: 587 knps
Stockfish: 1110 knps
Sungorus: 3074 knps
Texel: 1016 knps
ZCT: 752 knsp
And of course my own engines:
Brainless: 1291 knps
Giraffe: 126 knps
I didn't realize my old engine is so fast. Maybe that's why it's already beaten by Giraffe by almost 100 Elo.
Arasan: 9
Betsabe II: 13
Cheng: 13
Crafty: 24
Diablo: 10
EXchess: 7
Fairymax: 7
GNU Chess 6: 12
GreKo: 16
Phalanx: 8
RobboLito: 16
Scorpio: 11
Senpai: 7
Sloppy: 14
Smash: 6
Stockfish: 11
Sungorus: 31
Texel: 10
ZCT: 8
Brainless: 13
Giraffe: 1
Might be also interesting to know if they use pure bitboards or not. Perhaps if you use a hybrid design you may never reach top speed. By the way what was it twice the speed = 70 ELO ?
Arasan: 1
Betsabe II: 2
Cheng: 2
Crafty: 3
Diablo: 1
EXchess: 1
Fairymax: 1
GNU Chess 6: 2
GreKo: 2
Phalanx: 1
RobboLito: 2
Scorpio: 2
Senpai: 1
Sloppy: 2
Smash: 1
Stockfish: 2
Sungorus: 4
Texel: 1
ZCT: 1
Brainless: 2
Giraffe: 0
-
- Posts: 276
- Joined: Fri Mar 17, 2006 8:01 am
- Location: Russia
- Full name: Vladimir Medvedev
Re: Maximum speed
This approach can be a little bit misleading if different engines count nodes in different ways.matthewlai wrote:I happen to have quite a few engines in my gauntlet, so I ran a search from start position on all of them. All single threaded.
-
- Posts: 859
- Joined: Mon Aug 10, 2009 10:05 pm
- Location: Italy
- Full name: Stefano Gemma
Re: Maximum speed
What CPU?
Author of Drago, Raffaela, Freccia, Satana, Sabrina.
http://www.linformatica.com
http://www.linformatica.com
-
- Posts: 793
- Joined: Sun Aug 03, 2014 4:48 am
- Location: London, UK
Re: Maximum speed
Most of them are hybrid designs I think. At least both Brainless and Giraffe are.Henk wrote:Might be also interesting to know if they use pure bitboards or not. Perhaps if you use a hybrid design you may never reach top speed. By the way what was it twice the speed = 70 ELO ?
They both use bitboards for all move generation, and arrays only for parts of eval. Eval terms use whatever representation is faster. Both representations are updated incrementally by make/unmake. I believe this is how all top engines do it.
I believe it's about 70 Elo on the low end, and close to 50 Elo on the higher end.
The difference between 200 knps and 100 knps is bigger than the difference between 2000 knps and 1000 knps.
Disclosure: I work for DeepMind on the AlphaZero project, but everything I say here is personal opinion and does not reflect the views of DeepMind / Alphabet.