Komodo 13 released

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

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
mjlef
Posts: 1428
Joined: Thu Mar 30, 2006 12:08 pm
Contact:

Re: Komodo 13 released

Post by mjlef » Wed May 08, 2019 2:06 am

cma6 wrote:
Wed May 08, 2019 1:18 am
mjlef wrote:
Tue May 07, 2019 9:19 pm
cma6 wrote:
Tue May 07, 2019 9:06 pm


Using Aquarium GUI, after 1 minute in a typical endgame position with 6-man Syzygy installed, SF shows 45,680 kN/s and 260K TB access; Komodo shows 40,000 kN/s and 827K TB access; Komodo MCTS does not show any speed nor any TB access.
Do you think that this is an Aquarium issue or perhaps Komodo MCTS does not give TB access?
Since I may not be able to get any MCTS speed reading, what would your best guess for a good MCTS hash to use other than the default 128K?
K MCTS does use Syzygy (as long as you have it on and the SyzygyPath set). I would estimate 40 nps per thread, so maybe 1600 nps on your machine. I would then estimate about 1600*40*60 or about about 4 megs per minute of search, so maybe 230 megs per hour of search. I do not know why Aquarium is not showing you the nps, but maybe below some value it does not display. You could double click the Komodo executable, then type these, substituting whatever number of Thread you use for the n below:

setoption name Use MCTS value true
setoption name Threads value n
go movetime 30000

note the nps on the last line it shows after the 30 second search. This does not use Syzygy but the nps difference would be small with that anyway. Take that nps and multiply it by 2400 then divide by 1000000 to get an estimate of million Monte Carlo nodes per minute. then adjust for the total search time you want to run.
Mark
I get 2.92 mm Monte Carlo nodes/minute. But what does that tell me about MCTS hash size, e.g., if my average search is 10 minutes, how large should my MCTS hash be?
Perhaps you should email me directly to not fill the forum with one specific machine info. I am not sure what 2.92 mm Monte Carlo nodes/min means. What is mm? just take the Monte Carlo reported nps, multiply it by 2400 to get the nodes per minute. If you want to know how much memory for 10 minutes, then multiply that number by 10. That is the number of bytes you need. The Komodo MCTS Hash option is in megabytes, so divide that number by 1 million. It is very likely the default value will be fine for 10 minute searches even on a big machine. You really only need to set it higher for large machines with very long analysis.

User avatar
Nordlandia
Posts: 2477
Joined: Fri Sep 25, 2015 7:38 pm
Location: Sortland, Norway

Re: Komodo 13 released

Post by Nordlandia » Wed May 08, 2019 4:08 am

What is optimal table memory settings for my system? i7-5960X 4.5GHz, 16 GB ram, M.2 SSD for 6-men.

I've often used 256 MB for table memory. Opinions please.

peter
Posts: 1786
Joined: Sat Feb 16, 2008 6:38 am
Full name: Peter Martan

Re: Komodo 13 released

Post by peter » Wed May 08, 2019 6:15 am

mjlef wrote:
Tue May 07, 2019 7:40 pm
Each MCTS tree entry is about 40 bytes, and an hour has 3600 seconds, so 1400 * 40 * 3600 is about 200 megs an hour.
Sorry, Mark, but knowing this calculation already from Readme, isn't there (at Readme) one 0 missing? There's about the same example given for 8 cores and 5 hours and factor is 1000 instetad of 1400, yet only 576 mebabytes result, I guess that should read 5760, shouldn't it?
Each node is 40 bytes. So it you wanted to analyze with say 8 cores for 5 hours, you should set MCTS Hash to 1000 x 40 x 8 x 3600 x 5 (secs in an hour)/ 1000000 or about 576 megs.
Peter.

Andrew
Posts: 119
Joined: Wed Mar 08, 2006 11:51 pm
Location: Australia

Re: Komodo 13 released

Post by Andrew » Wed May 08, 2019 7:28 am

cma6 wrote:
Tue May 07, 2019 7:19 pm
"We think the KRPPvsKRP and KPPPvs KPP are the most useful 7 piece Syzygy files to use first."
Andrew, may I ask from where are you getting these 7-man Syzygy TB files?
Thanks, CMA

Andrew
We think the KRPPvsKRP and KPPPvs KPP are the most useful 7 piece Syzygy files to use first. Note that it is quite possible to promote a pawn to a queen or other piece, and this can make the engine think it is bad if that specific table is missing, since to the engine, it looks like it has done from a win to a position where it is just vary ahead. I wanted to mention that since it can look odd when you analyze.
[/quote]
[/quote]

Hi I haven't downloaded and 7 piece files (yet!) WIll wait a short while until I have at least a 512Gb USB drive

The Komodo readme file includes this link: http://tablebase.lichess.ovh/tables/standard/ I'm pretty sure there are other sites as well

Andrew

Uri Blass
Posts: 8609
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: Komodo 13 released

Post by Uri Blass » Wed May 08, 2019 9:45 am

I wonder why CCRL do not include Komodo MCTS as a version of Komodo in their comparison

https://ccrl.chessdom.com/ccrl/4040/cgi ... 4-bit_4CPU

There are 61 versions of Komodo but no version is Komodo MCTS


Looking at the full list I find

2 Komodo 11.3.1 64-bit 4CPU 3398
6 Komodo 12.3 MCTS 64-bit 4CPU 3292

I believe that both Komodo and Komodo MCTS share the same code and it is not fair to have them as seperate engines.
I believe Komodo MCTS is simply a version of Komodo and there is no justification to treat it differently.

If Houdini tactical is considered to be a version of Houdini then by the same logic Komodo MCTS should be considered as a version of Komodo.

User avatar
Ozymandias
Posts: 1137
Joined: Sun Oct 25, 2009 12:30 am

Re: Komodo 13 released

Post by Ozymandias » Wed May 08, 2019 11:23 am

Uri Blass wrote:
Wed May 08, 2019 9:45 am
I believe that both Komodo and Komodo MCTS share the same code and it is not fair to have them as seperate engines.
I believe Komodo MCTS is simply a version of Komodo and there is no justification to treat it differently.

If Houdini tactical is considered to be a version of Houdini then by the same logic Komodo MCTS should be considered as a version of Komodo.
Fair or not, they do consider Komodo MCTS as a separate engine and include all versions in a separate group. They're between their rights to do so.

Playing devil's advocate, if you want a better example, exchange Houdini tactical for ShashChess in your analogy.

Modern Times
Posts: 2420
Joined: Thu Jun 07, 2012 9:02 pm

Re: Komodo 13 released

Post by Modern Times » Wed May 08, 2019 11:37 am

You can't please everyone. Some people want it one way, and other people the other way.
.

Opinions expressed here are my own, and not necessarily those of the CCRL Group.

mjlef
Posts: 1428
Joined: Thu Mar 30, 2006 12:08 pm
Contact:

Re: Komodo 13 released

Post by mjlef » Wed May 08, 2019 11:50 am

peter wrote:
Wed May 08, 2019 6:15 am
mjlef wrote:
Tue May 07, 2019 7:40 pm
Each MCTS tree entry is about 40 bytes, and an hour has 3600 seconds, so 1400 * 40 * 3600 is about 200 megs an hour.
Sorry, Mark, but knowing this calculation already from Readme, isn't there (at Readme) one 0 missing? There's about the same example given for 8 cores and 5 hours and factor is 1000 instetad of 1400, yet only 576 mebabytes result, I guess that should read 5760, shouldn't it?
Each node is 40 bytes. So it you wanted to analyze with say 8 cores for 5 hours, you should set MCTS Hash to 1000 x 40 x 8 x 3600 x 5 (secs in an hour)/ 1000000 or about 576 megs.
Yes, there was an extra 0 in the readme. We will update it soon. Sorry about that.

mjlef
Posts: 1428
Joined: Thu Mar 30, 2006 12:08 pm
Contact:

Re: Komodo 13 released

Post by mjlef » Wed May 08, 2019 11:59 am

Nordlandia wrote:
Wed May 08, 2019 4:08 am
What is optimal table memory settings for my system? i7-5960X 4.5GHz, 16 GB ram, M.2 SSD for 6-men.

I've often used 256 MB for table memory. Opinions please.
The default should be fine for Table Memory for most any game you would play. We made this an option because people doing long analysis (like over night or several days) might benefit by using a higher value. In that case users can just use whatever setting gives the highest nps for them over whatever time period they expect to use. Note in most programs, Table Memory is just a fixed amount per thread. If you add up the similar tables in Stockfish I believe it comes ut to 3-4 megs per thread.

For very fast games (like game in 30 seconds) a smaller number might be faster, which would make the program a bit more cache friendly. This kind of thing is going to vary by how big the cache is and how fast main memory is so I cannot give a definite answer without machine access. Just try some and go with whatever is fastest in ns.

Mark

User avatar
Ovyron
Posts: 2829
Joined: Tue Jul 03, 2007 2:30 am

Re: Komodo 13 released

Post by Ovyron » Wed May 08, 2019 8:44 pm

Uri Blass wrote:
Wed May 08, 2019 9:45 am
I believe that both Komodo and Komodo MCTS share the same code and it is not fair to have them as seperate engines.
This is a common thing for the CCRL, as Fritz 16 is +90% Rybka 4, and they have them as separate engines.
Great spirits have always encountered violent opposition from mediocre minds.

Post Reply