6 Man Syzygy vs 5 Man Nalimov

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

Would you like to see Crafty support 6 man Syzygy tablebases?

Poll ended at Tue Apr 26, 2016 3:24 pm

Yes
22
59%
No
1
3%
Does not matter to me
14
38%
 
Total votes: 37
User avatar
Nordlandia
Posts: 2822
Joined: Fri Sep 25, 2015 9:38 pm
Location: Sortland, Norway

Re: 6 Man Syzygy vs 5 Man Nalimov

Post by Nordlandia »

Will 5-Men catched in RAM outweight 6-Men on SSD or vise versa.
Top
User avatar
yurikvelo
Posts: 710
Joined: Sat Dec 06, 2014 1:53 pm

Re: 6 Man Syzygy vs 5 Man Nalimov

Post by yurikvelo »

Nordlandia wrote:Will 5-Men catched in RAM outweight 6-Men on SSD or vise versa.
What do you mean by "outweight"?

If talk about tournament - 5-men and 6-men do not provide any ELO.
At low computing resources (slow HW or short TC or both) by the time when TB probing start to influence eval and PV-selection - game outcome is already decided. Often game is even adjudicated (+4.5 ... +6.5 eval or consecutive -0.05...+0.05) before TBhits come into play.

At big D (TCEC-2015 like) nobody tested Syzygy-6, because no HW available to run statistically significant amount of games.

Also time management parameters of Stockfish, Komodo, Houdini and Gull are heavily tuned for non-EGTB scenario. TP Probing in late EG might save time, which could be spent to enhance midplay, but engines allocate enough time for EG to fight TB-less.


If talking about task solving, monte carlo, infinite analyze - nothing can compensate lack of Syzygy 6.
Try to solve Behting Study without Syzygy6.

Also Syzygy influence depends how mature is built-in EG knowledge. If author didn't spend enough efforts to implement all kind of EG knowledge - his engine will often miss win or draw possibility (depends on adjudication rules).
Implementing Syzygy code is easier than implementing EG knowledge.
Top
BBauer
Posts: 658
Joined: Wed Mar 08, 2006 8:58 pm

Re: 6 Man Syzygy vs 5 Man Nalimov

Post by BBauer »

It is simply an obvious observation. Data will sit in memory TWICE. Once in the RAM drive, once in normal filesystem cache. It is certainly true. How much the lost memory hurts is a different issue.
My point is that data may be swaped out by OS.
Then it will sit in memory only once. In RAM.
Now it may be reloaded faster.
Kind regards
Bernhard
Top
kgburcham
Posts: 2016
Joined: Sun Feb 17, 2008 4:19 pm

Re: 6 Man Syzygy vs 5 Man Nalimov

Post by kgburcham »

Yuri here are three tests using Crystal disk bench software.

First is SATA where my EGTB are stored.
Second is PCIE SSD.
Third test is RamDisk.

-----------------------------------------------------------------------
CrystalDiskMark 5.1.2 x64 (C) 2007-2016 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

Sequential Read (Q= 32,T= 1) : 542.822 MB/s
Sequential Write (Q= 32,T= 1) : 274.073 MB/s
Random Read 4KiB (Q= 32,T= 1) : 249.298 MB/s [ 60863.8 IOPS]
Random Write 4KiB (Q= 32,T= 1) : 179.867 MB/s [ 43912.8 IOPS]
Sequential Read (T= 1) : 494.711 MB/s
Sequential Write (T= 1) : 243.908 MB/s
Random Read 4KiB (Q= 1,T= 1) : 28.616 MB/s [ 6986.3 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 105.158 MB/s [ 25673.3 IOPS]

Test : 1024 MiB [E: 25.4% (56.6/223.2 GiB)] (x5) [Interval=5 sec]
Date : 2016/04/08 6:51:48
OS : Windows 8.1 Pro [6.3 Build 9600] (x64)


-----------------------------------------------------------------------
CrystalDiskMark 5.1.2 x64 (C) 2007-2016 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

Sequential Read (Q= 32,T= 1) : 1829.325 MB/s
Sequential Write (Q= 32,T= 1) : 447.114 MB/s
Random Read 4KiB (Q= 32,T= 1) : 273.191 MB/s [ 66697.0 IOPS]
Random Write 4KiB (Q= 32,T= 1) : 401.868 MB/s [ 98112.3 IOPS]
Sequential Read (T= 1) : 782.459 MB/s
Sequential Write (T= 1) : 383.784 MB/s
Random Read 4KiB (Q= 1,T= 1) : 25.541 MB/s [ 6235.6 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 97.861 MB/s [ 23891.8 IOPS]

Test : 1024 MiB [C: 15.7% (34.6/219.8 GiB)] (x5) [Interval=5 sec]
Date : 2016/04/08 6:38:58
OS : Windows 8.1 Pro [6.3 Build 9600] (x64)


-----------------------------------------------------------------------
CrystalDiskMark 5.1.2 x64 (C) 2007-2016 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

Sequential Read (Q= 32,T= 1) : 6313.399 MB/s
Sequential Write (Q= 32,T= 1) : 6145.733 MB/s
Random Read 4KiB (Q= 32,T= 1) : 845.159 MB/s [206337.6 IOPS]
Random Write 4KiB (Q= 32,T= 1) : 623.884 MB/s [152315.4 IOPS]
Sequential Read (T= 1) : 6870.329 MB/s
Sequential Write (T= 1) : 10227.936 MB/s
Random Read 4KiB (Q= 1,T= 1) : 851.547 MB/s [207897.2 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 886.764 MB/s [216495.1 IOPS]

Test : 1024 MiB [F: 0.0% (0.0/3992.2 MiB)] (x5) [Interval=5 sec]
Date : 2016/04/19 17:26:04
OS : Windows 8.1 Pro [6.3 Build 9600] (x64)
no chess program was born totally from one mind. all chess programs have many ideas from many minds.
Top
Dirt
Posts: 2851
Joined: Wed Mar 08, 2006 10:01 pm
Location: Irvine, CA, USA

Re: 6 Man Syzygy vs 5 Man Nalimov

Post by Dirt »

kgburcham wrote: Sequential Write (Q= 32,T= 1) : 447.114 MB/s
Why would you care what the write speed is?

A disk test program is going to make sure you can't read the data back out from the buffers. It has no relevance.

A faster disk will help, but not as much as using system RAM for a disk will hurt. External RAM would be great.
Deasil is the right way to go.
Top
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: 6 Man Syzygy vs 5 Man Nalimov

Post by bob »

BBauer wrote:
It is simply an obvious observation. Data will sit in memory TWICE. Once in the RAM drive, once in normal filesystem cache. It is certainly true. How much the lost memory hurts is a different issue.
My point is that data may be swaped out by OS.
Then it will sit in memory only once. In RAM.
Now it may be reloaded faster.
Kind regards
Bernhard
Not faster in RAM. Just duplicated. Unless you don't have enough memory. And then you are simply stealing memory for the RAM drive that can't be used for anything, whether the RAM drive is being accessed or not.

Guarantee you the operating system is FAR better at optimizing this stuff than you and I are.
Top
tpoppins
Posts: 919
Joined: Tue Nov 24, 2015 9:11 pm
Location: upstate

Re: 6 Man Syzygy vs 5 Man Nalimov

Post by tpoppins »

Dirt wrote:
kgburcham wrote: Sequential Write (Q= 32,T= 1) : 447.114 MB/s
Why would you care what the write speed is?
Or, indeed, what any of the Sequential Read numbers are.
Top
User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

Re: 6 Man Syzygy vs 5 Man Nalimov

Post by MikeB »

Current Crafty 25.1 with Syzygy

Single Core - Fine # 70, 6 MAN Syzygy TBs:
[d]8/k7/3p4/p2P1p2/P2P1P2/8/8/K7 w - -

Code: Select all

         52->   2.42/19.20   27.18   1. Kb1 Kb7 2. Kc1 Kc7 3. Kd1 Kd8 4. Kc2
                                     Kc8 5. Kd2 Kd8 6. Kc3 Kc7 7. Kd3 Kb6
                                     8. Ke2 Kc7 9. Kf2 Kd7 10. Kg3 Ke7 11. Kh4
                                     Kf6 12. Kh5 Kf7 13. Kg5 Kg7 14. Kxf5 Kf7
                                     15. Ke4 Ke7 16. Kd3 Kf6 17. f5 Ke7 18. f6+
                                     Kf7 19. Kc4 Ke8 20. f7+ Kf8 21. Kb5 Kg7
                                     22. f8=Q+ Kg6 23. Qxd6+ Kg5 24. Qe5+ Kg4
                                     25. d6 Kf3 26. d7 Kg2 27. d8=Q Kh1
         53     2.90/19.20   Mat48   1. Kb1 Kb7 2. Kc1 Kc7 3. Kd1 Kd8 4. Kc2
                                     Kc8 5. Kd2 Kd8 6. Kc3 Kc7 7. Kd3 Kb6
                                     8. Ke2 Kc7 9. Kf2 Kd7 10. Kg3 Ke7 11. Kh4
                                     Kf6 12. Kh5 Kf7 13. Kg5 Kg7 14. Kxf5 Kf7
                                     15. Ke4 Ke7 16. Kd3 Kf6 17. f5 Ke7 18. f6+
                                     Kf7 19. Kc4 Ke8 20. f7+ Kf8 21. Kb5 Kg7
                                     22. f8=Q+ Kg6 23. Qxd6+ Kg5 24. Kxa5
                                     <EGTB>
         53->   2.90/18.00   Mat48   1. Kb1 Kb7 2. Kc1 Kc7 3. Kd1 Kd8 4. Kc2
                                     Kc8 5. Kd2 Kd8 6. Kc3 Kc7 7. Kd3 Kb6
                                     8. Ke2 Kc7 9. Kf2 Kd7 10. Kg3 Ke7 11. Kh4
                                     Kf6 12. Kh5 Kf7 13. Kg5 Kg7 14. Kxf5 Kf7
                                     15. Ke4 Ke7 16. Kd3 Kf6 17. f5 Ke7 18. f6+
                                     Kf7 19. Kc4 Ke8 20. f7+ Kf8 21. Kb5 Kg7
                                     22. f8=Q+ Kg6 23. Qxd6+ Kg5 24. Kxa5
                                     <EGTB>
         54     3.12/18.00   Mat46   1. Kb1 Kb7 2. Kc1 Kc7 3. Kd1 Kd8 4. Kc2
                                     Kc8 5. Kd2 Kd8 6. Kc3 Kc7 7. Kd3 Kb6
                                     8. Ke2 Kc7 9. Kf2 Kd7 10. Kg3 Ke7 11. Kh4
                                     Kf6 12. Kh5 Kf7 13. Kg5 Kg7 14. Kxf5 Kf7
                                     15. Ke4 Ke7 16. Kd3 Kf6 17. f5 Ke7 18. f6+
                                     Kf7 19. Kc4 Ke8 20. f7+ Ke7 21. f8=Q+ Kxf8
                                     22. Kb5 Kg8 23. Kxa5 <EGTB>
         54->   3.12/18.00   Mat46   1. Kb1 Kb7 2. Kc1 Kc7 3. Kd1 Kd8 4. Kc2
                                     Kc8 5. Kd2 Kd8 6. Kc3 Kc7 7. Kd3 Kb6
                                     8. Ke2 Kc7 9. Kf2 Kd7 10. Kg3 Ke7 11. Kh4
                                     Kf6 12. Kh5 Kf7 13. Kg5 Kg7 14. Kxf5 Kf7
                                     15. Ke4 Ke7 16. Kd3 Kf6 17. f5 Ke7 18. f6+
                                     Kf7 19. Kc4 Ke8 20. f7+ Ke7 21. f8=Q+ Kxf8
                                     22. Kb5 Kg8 23. Kxa5 <EGTB>
         55     3.54/18.00   Mat44   1. Kb1 Kb7 2. Kc1 Kc7 3. Kd1 Kd8 4. Kc2
                                     Kc8 5. Kd2 Kd8 6. Kc3 Kc7 7. Kd3 Kb6
                                     8. Ke2 Kc7 9. Kf2 Kd7 10. Kg3 Ke7 11. Kh4
                                     Kf6 12. Kh5 Kf7 13. Kg5 Kg7 14. Kxf5 Kf7
                                     15. Ke4 Ke7 16. Kd3 Kf6 17. Kc4 Kf5
                                     18. Kb5 Ke4 19. Kxa5 Ke3 20. Kb6 Ke2
                                     21. Kc6 Kd1 22. Kxd6 <EGTB>
         55->   3.54/18.00   Mat44   1. Kb1 Kb7 2. Kc1 Kc7 3. Kd1 Kd8 4. Kc2
                                     Kc8 5. Kd2 Kd8 6. Kc3 Kc7 7. Kd3 Kb6
                                     8. Ke2 Kc7 9. Kf2 Kd7 10. Kg3 Ke7 11. Kh4
                                     Kf6 12. Kh5 Kf7 13. Kg5 Kg7 14. Kxf5 Kf7
                                     15. Ke4 Ke7 16. Kd3 Kf6 17. Kc4 Kf5
                                     18. Kb5 Ke4 19. Kxa5 Ke3 20. Kb6 Ke2
                                     21. Kc6 Kd1 22. Kxd6 <EGTB>
        time=3.54(100%)  nodes=29747441(29.7M)  fh1=96%  pred=0  nps=8.4M
        chk=1.8M  qchk=2.0M  fp=10.7M  mcp=79.6K  50move=1  egtb=34.7K
        LMReductions:  1/1.8M  2/1.2M  3/404.1K  4/30.8K  5/56
        null-move (R):  3/154.7K  4/209.3K  5/61.9K  6/15.4K  7/2.1K  8/14

mate in 44 moves.

White(1): Kb1
              time used:   3.54
Top
User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

Re: 6 Man Syzygy vs 5 Man Nalimov

Post by MikeB »

and how about this blast from the past with 12 cores:

White(1): [d]8/1p3pp1/7p/5P1P/2k3P1/8/2K2P2/8 w - - 0 1

Code: Select all

White(1): g
        time surplus   0.00  time limit 30.00 (2:30)
        depth     time       score   variation (1)
starting thread 1 2 3 4 5 6 7 8 9 10 11 <done>
         23     1.45/20.10   21.52   1. f6 gxf6 2. f4 f5 3. g5 hxg5 4. fxg5 Kd4
                                     5. g6 f6 6. g7 b5 7. g8=Q b4 8. h6 Ke5
                                     9. h7 b3+ 10. Qxb3 Kd6 11. h8=Q Ke7
                                     12. Qg7+ Kd6 13. Qxf6+ Kc5 14. Qxf5+
         23->   1.45/18.00   21.52   1. f6 gxf6 2. f4 f5 3. g5 hxg5 4. fxg5 Kd4
                                     5. g6 f6 6. g7 b5 7. g8=Q b4 8. h6 Ke5
                                     9. h7 b3+ 10. Qxb3 Kd6 11. h8=Q Ke7
                                     12. Qg7+ Kd6 13. Qxf6+ Kc5 14. Qxf5+
         24     2.33/18.00   Mat20   1. f6 gxf6 2. f4 f5 3. g5 hxg5 4. fxg5 f4
                                     5. g6 f6 6. g7 Kd4 7. g8=Q f3 8. Qg4+ Ke3
                                     9. Qe6+ Kf2 10. Qxf6
         24->   2.33/18.00   Mat20   1. f6 gxf6 2. f4 f5 3. g5 hxg5 4. fxg5 f4
                                     5. g6 f6 6. g7 Kd4 7. g8=Q f3 8. Qg4+ Ke3
                                     9. Qe6+ Kf2 10. Qxf6
         25     2.54/18.00   Mat18   1. f6 gxf6 2. f4 f5 3. g5 hxg5 4. fxg5 f4
                                     5. g6 f6 6. g7 Kd4 7. g8=Q f3 8. Qg7 Ke3
                                     9. Qxb7 <EGTB>
         25->   2.55/18.00   Mat18   1. f6 gxf6 2. f4 f5 3. g5 hxg5 4. fxg5 f4
                                     5. g6 f6 6. g7 Kd4 7. g8=Q f3 8. Qg7 Ke3
                                     9. Qxb7 <EGTB>
        time=2.55(80%)  nodes=175062349(175.1M)  fh1=91%  pred=0  nps=68.7M
        chk=11.5M  qchk=21.3M  fp=59.4M  mcp=602.7K  50move=0  egtb=18.2K
        LMReductions:  1/2.2M  2/1.1M  3/389.6K  4/116.0K  5/1.4K
        null-move (R):  3/604.9K  4/162.7K  5/9.9K  6/25
        splits=77.6K(48.2K)  aborts=4.1K  joins=130.8K  data=17%(20%)

mate in 18 moves.

White(1): f6
              time used:   2.55
Black(1):
Top
User avatar
yurikvelo
Posts: 710
Joined: Sat Dec 06, 2014 1:53 pm

Re: 6 Man Syzygy vs 5 Man Nalimov

Post by yurikvelo »

kgburcham wrote:Yuri here are three tests using Crystal disk bench software.
First is SATA where my EGTB are stored.
Second is PCIE SSD.
Third test is RamDisk.
Point is that if you have a set of <1Gb files and have at least 1Gb free RAM - access to this files through operating system API (not through direct access layer to controller, as Benchmark software do) will be identical - no matter if it is ancient HDD or RAMDisk.

If you store 5 Syzygy on ancient HDD - you consume only 1GB RAM to provide fast latency-free access to these files.

If you store 5 Syzygy on RAMDisk - you consume twice - about 2 GB RAM to provide the same fast latency-free access to these files.

Operating system do not disable its internal filesystem RAM-buffers if drive is of RAMDisk type.
Top

Return to “Computer Chess Club: General Topics”