Any information about a current Crafty version with NN?

[Eelco de Groot]

I wonder ...
After 10 months (TalkChess discussion) no reaction from Bob about it.

Vafra:
https://www.jurjevic.org.uk/chess/vafra ... nload.html

Best
Frank

He wrote last time he was here that he did not like all the borrowing stuff, but if that was what the majority wanted, he was okay with the democratic principle involved at least. But he was not so interested anymore.

A long time before that, he wrote that he had taken up fishing again. Rybkas, Stockfish, Bluefish, Redfish...

[Frank Quisinsky]

Hi Eelco,

yes, I remember on his message about fishing here.
For me Bob was one of the best supporter (V.I.P.) in computer chess.

He hold the computer chess flag brilliantly so many years for all of us.
No doubt about it.

Yesterday I am looking in bit in Vafra.
A lot of efforts for Crafty I can see here, over a longer time now.

But after all ...
I am not sure Prof. Hyatt will like that.

Normally each of such efforts should be good.
If anything is wrong it can be changed.

Thanks for your message!

Have a nice weekend ...
Unfortunately, in bad times we live.

Best
Frank

[Frank Quisinsky]

Hi there,

different test matches vs. 3100+ Elo engines played.

Vafra 3.3.1 NN (Crafty) is playing with personality, Nalimov (not syzybybases), NN file perfect as WB engine with WB2uci adapter under Shredder 12 GUI. But the Elo is not high enough for my list (I believe around 2.875 if Shredder 13 = 3125). Made no sense to test it for my KI-Tourneys.

Best
Frank

[Frank Quisinsky]

Ah, here the configuration I made a test:

Vafra.rc

Code: Select all

cpus 1

egtb
tbpath=c:\Chess\_nalimov-5
cache=64M

nnue .\net\nn-v0f000007014.nnue

hash=512M
hashp=32M

personality load .\vpf\vafra.cpf

ponder off

# (for use use with ponder on)
# mode=tournament

# (Allows Crafty to try to win drawn games (according to Endgame Tables))
swindle off

learn 0
 
book off

# (Makes crafty use a lot more time on the first 8 moves out of book)
# timebook 80 8

# show book
# book random 1
# bookw freq 0.7
# bookw ratio 0.8
# bookw eval 0.6
# bookw learn 1
# bookw cap 0.5
# book width 4

# resign 9 1

log=off
exit

WB2uci.eng
I am using the Winboard adapter (rename to Vafra.exe)
Oct. 08th, 2014, 245.760 bytes

Code: Select all

[ENGINE]
Name               = Vafra 3.3.1 NN (Crafty)
Author             = Prof. Robert M. (Bob) Hyatt, et al.
Filename           = Vafra.exe

[OPTIONS]
Program            = Vafra_331-NN.exe
;InitString        =
Ponder             = false
;Logfile           = false
;HashOnCommandline = false
Protocol           = 2
InitTime           = 5
LevelExtend        = none
Analyze            = true
MateScore          = 32768
ShowThinkingMove   = false
UseUndo            = false
WhiteScore         = true
TerminateHard      = true

Visible            = Extras,Logfile,Protocol,InitTime,ShowThinkingMove,LevelExtend

[EXTRAS]
Clear Hash         = clrhash|button

Shredder *.eng file:

Code: Select all

[ENGINE]
Name=Vafra 3.3.1 NN (Crafty)
Author=Prof. Robert M. (Bob) Hyatt, et al.
Filename=C:\Chess\Shredder_12\Engines\Vafra\Vafra.exe

The last WB adapter I am using can be found here:

Configuration site for FCP Tourney-2021
https://www.amateurschach.de/main/_conf ... n-2021.htm

If others need the information ...

Best
Frank

[Chessqueen]

Ah, here the configuration I made a test:

Vafra.rc

Code: Select all

cpus 1

egtb
tbpath=c:\Chess\_nalimov-5
cache=64M

nnue .\net\nn-v0f000007014.nnue

hash=512M
hashp=32M

personality load .\vpf\vafra.cpf

ponder off

# (for use use with ponder on)
# mode=tournament

# (Allows Crafty to try to win drawn games (according to Endgame Tables))
swindle off

learn 0
 
book off

# (Makes crafty use a lot more time on the first 8 moves out of book)
# timebook 80 8

# show book
# book random 1
# bookw freq 0.7
# bookw ratio 0.8
# bookw eval 0.6
# bookw learn 1
# bookw cap 0.5
# book width 4

# resign 9 1

log=off
exit

WB2uci.eng
I am using the Winboard adapter (rename to Vafra.exe)
Oct. 08th, 2014, 245.760 bytes

Code: Select all

[ENGINE]
Name               = Vafra 3.3.1 NN (Crafty)
Author             = Prof. Robert M. (Bob) Hyatt, et al.
Filename           = Vafra.exe

[OPTIONS]
Program            = Vafra_331-NN.exe
;InitString        =
Ponder             = false
;Logfile           = false
;HashOnCommandline = false
Protocol           = 2
InitTime           = 5
LevelExtend        = none
Analyze            = true
MateScore          = 32768
ShowThinkingMove   = false
UseUndo            = false
WhiteScore         = true
TerminateHard      = true

Visible            = Extras,Logfile,Protocol,InitTime,ShowThinkingMove,LevelExtend

[EXTRAS]
Clear Hash         = clrhash|button

Shredder *.eng file:

Code: Select all

[ENGINE]
Name=Vafra 3.3.1 NN (Crafty)
Author=Prof. Robert M. (Bob) Hyatt, et al.
Filename=C:\Chess\Shredder_12\Engines\Vafra\Vafra.exe

The last WB adapter I am using can be found here:

Configuration site for FCP Tourney-2021
https://www.amateurschach.de/main/_conf ... n-2021.htm

If others need the information ...

Best
Frank

I see you posted Intel core i9, how fast is it compared to other CPU, it is rank 2nd ?

[Frank Quisinsky]

Hi,

a small example:

From the site about condition / organization to my FCP Tourney-KI:
https://www.amateurschach.de/main/_fcp-tourney-ki.htm

Name of systems is maybe interesting for programmers ... *.pgn file, system game played (FQ-1, FQ-3, FQ-4).

Code: Select all

Processor
1. Intel® Core™ i9-10900K  = 10 cores, name: FQ-1
2. Intel® Core™ i5-11600K  = 6 cores, name: FQ-4
3. AMD Ryzen™ 9 5950X / RTX-3070 Ti = 16 cores, name: FQ-3

Ghz, RAM
4.4Ghz, DDR4 3200Mhz

Bench
Intel i9: time= 147.86 sec., nodes= 183738544, nps= 1242647
Intel i5: time= 128.69 sec., nodes= 183738544, nps= 1427729
AMD R9:   time= 123.76 sec., nodes= 183738544, nps= 1484649

Bench results by Wasp 5.20 NN with 1 core

The complete bench results (more as 1 core) can be found in Wasp 5.20 NN download file.
https://www.amateurschach.de/download/w ... mpiles.zip

The processor is interesting, 1-core can run with 5.1Ghz (with water cooling systems I am using) but need to many energy with 5.1Ghz. It is i9 from 10th generation. i9 from 11th generation have 8 cores (not 10). You can see that my absolutly favorite processor (Intel i5-11600k) is faster with 1 core. This processor can run with 5.3Ghz with water cooling systems. 4.4Ghz is perfect (temperature, energy).

The i9-10900k is interesting if you need many Intel cores only.
I am using the processor a long time, FCP Tourney-2020, FCP Tourney-2021.

For my KI-Ratings I am using 16 Intel and 16 AMD cores.
A good mix for better results.

You can see with "Wasp-Bench" the systems have around the same power (nps).
The difference isn't important for 80 minutes per game.

Best
Frank

If you need powerfull Intel cores ..
2x i5 11600k is more interesting as 1x i9-11900k or i9-10900k for chess.
AMD Ryzsen 9 5950x is faster as Intel processors in case of Wasp.

[Frank Quisinsky]

Forgot ...

AMD Ryzen 9 5950x can run max. with 4.6Ghz and water cooling (easy overclocking).
AMD Ryzen 9 5950x with 4.6 Ghz vs. Intel Core i5 11600k with 5.3Ghz is around the same (in case of Wasp-Bench, 1-core NPS).
I am testing chess programs allways with 1-core.

Today Intel i9 10900k is not the best processor if you like to buy a new system.
But 2020 and 2021 my FCP Tourneys runs with full power and 4.9 Ghz.

AMD is for testing more interesting.

But again, the price for Intel i5-11600k is great.
Very powerfull processor with AVX-512 support!!
I like the processor a lot.

Best
Frank