Komodo 8 Announcement

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.
Post Reply
lkaufman
Posts: 3761
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Komodo 8 Announcement

Post by lkaufman » Thu Aug 28, 2014 3:23 pm

Komodo 8 will be released within the next two weeks, both by ChessBase and on komodochess.com. They will have the same engine, but the ChessBase version will include their own GUI and their own opening book, plus PlayChess membership. Our release will include Windows 64 bit and 32 bit versions plus Mac and Linux versions. We anticipate that our price will remain the same as for Komodo 7, and that Komodo 7 purchasers will get the same discount as Komodo 6 purchasers received when buying Komodo 7. Our exact release date is not yet decided but we will post an announcement here when it is.
Komodo 8 uses the same code as development version 1288.00, but due to a better choice of compiler it runs 4% faster. As has been reported here by Graham Banks, version 1288.00 (without the 4% speedup) defeated Houdini 4 by 16 wins to 4 (with 30 draws), and as reported here by Ingo Bauer it finished 9 elo ahead of Stockfish 5 in his 3000+ game rating test, without either the speedup or the use of Syzygy tablebases!
In my own testing at 2' + 1" Komodo 8 defeated Houdini 4 by 30 to 43 elo points on one, two, and four threads (each match a thousand games), despite Houdini's well-known strength at blitz. Against Stockfish 5 at this same time limit, results were very close, seven elo plus on one thread, seven elo minus on 2, 4, and 8 threads (combined results). Based on the results of outside testers, it appears that Komodo gains much more from more thinking time than Houdini, and perhaps slightly more than Stockfish. There is little doubt that Komodo 8 will be the top rated commercial engine; whether it will outrate Stockfish will depend on test conditions and on whether or not the latest dev. version of Stockfish is used.
How much is the elo gain from Komodo 7? Based on results against other engines on one thread it appears to be about 40 points, both according to Ingo (IPON) and to my own 2' + 1" tests. Based on a direct match on one thread Komodo 8 beat Komodo 7a by 50 elo in two thousand games at that time control; on two threads it won by 57 elo, and on four it leads by 502 to 348 (+63 elo) at this writing. We have improved MP performance in Komodo 8 (beyond the general improvements), and these results confirm that this is the case. Bear in mind that direct matches of related engines tend to overstated rating differences by 20-25%, so I think the most accurate estimate for elo gain against other engines is about forty on one thread and about fifty on four, at time limits in the five minute ballpark.

Werewolf
Posts: 1198
Joined: Thu Sep 18, 2008 8:24 pm

Re: Komodo 8 Announcement

Post by Werewolf » Thu Aug 28, 2014 3:39 pm

So presumably a user who ran Komodo 8 on many cores (16-32) would see even greater gains..?

Komodo 7a ran nicely on 24 cores but had real issues on 32. Do you think K8 will be better here?

User avatar
reflectionofpower
Posts: 1456
Joined: Fri Mar 01, 2013 4:28 pm
Location: USA

Re: Komodo 8 Announcement

Post by reflectionofpower » Thu Aug 28, 2014 4:10 pm

Werewolf wrote:So presumably a user who ran Komodo 8 on many cores (16-32) would see even greater gains..?

Komodo 7a ran nicely on 24 cores but had real issues on 32. Do you think K8 will be better here?
WHat happened when it ran on 32 cores?
"hodie mihi, cras tibi"

Lonnie

Werewolf
Posts: 1198
Joined: Thu Sep 18, 2008 8:24 pm

Re: Komodo 8 Announcement

Post by Werewolf » Thu Aug 28, 2014 4:21 pm

reflectionofpower wrote:
Werewolf wrote:So presumably a user who ran Komodo 8 on many cores (16-32) would see even greater gains..?

Komodo 7a ran nicely on 24 cores but had real issues on 32. Do you think K8 will be better here?
WHat happened when it ran on 32 cores?
When I tested it (admittedly only once) the task manager wasn't at 100% for some reason and the NPS were lower than expected. I asked someone who has a 32 core machine to test this and he reported K7 was running slower on 32 cores than 24. I specifically asked if HT was off and he confirmed this to be the case.

lkaufman
Posts: 3761
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Komodo 8 Announcement

Post by lkaufman » Thu Aug 28, 2014 4:21 pm

Werewolf wrote:So presumably a user who ran Komodo 8 on many cores (16-32) would see even greater gains..?

Komodo 7a ran nicely on 24 cores but had real issues on 32. Do you think K8 will be better here?
The elo gain from Komodo 7 should continue to rise with more cores than 8, but I would expect the added gain to be just a couple elo points. We didn't change anything directly related to number of cores/threads, we just made MP work better. If you had some problem on 32 cores I would not expect it to go away with Komodo 8; my best machine is 20 cores and the most any of our regular testers has is 24, so I have no way to check out what happens on 32. I think a 32 core machine is probably AMD rather than Intel; perhaps this is where the problem was. We only develop on Intel.

Werewolf
Posts: 1198
Joined: Thu Sep 18, 2008 8:24 pm

Re: Komodo 8 Announcement

Post by Werewolf » Thu Aug 28, 2014 4:26 pm

lkaufman wrote:
Werewolf wrote:So presumably a user who ran Komodo 8 on many cores (16-32) would see even greater gains..?

Komodo 7a ran nicely on 24 cores but had real issues on 32. Do you think K8 will be better here?
The elo gain from Komodo 7 should continue to rise with more cores than 8, but I would expect the added gain to be just a couple elo points. We didn't change anything directly related to number of cores/threads, we just made MP work better. If you had some problem on 32 cores I would not expect it to go away with Komodo 8; my best machine is 20 cores and the most any of our regular testers has is 24, so I have no way to check out what happens on 32. I think a 32 core machine is probably AMD rather than Intel; perhaps this is where the problem was. We only develop on Intel.
Thanks for the answer.

It was an Intel machine, but several engines were having trouble to be fair.

Frank Quisinsky
Posts: 4852
Joined: Wed Nov 18, 2009 6:16 pm
Location: Trier, Germany
Contact:

Re: Komodo 8 Announcement

Post by Frank Quisinsky » Thu Aug 28, 2014 4:46 pm

Hello,

a good decision that Komodo will be available on the webpage by the programmers. Thinking today to replaced Komodo with an other engine if Komodo is available from Chessbase only.

I dislike the buggy software from ChessBae without to have support in the past. Latest release versions by ChessBase are IPP-Clones if I read here.

No interest to support a bad call with my money.

If Komodo is available on the webpage by the programmers all is OK for myself and I am very happy to see the new version in my new list.

Good luck with your new version!

BUT
***

Very fast new Komodo updates are available without to get 1x a better version as customer for free. That is not a good service for customers in my opinion. But in the past are Doch and Komodo free available. Today Komodo life from that Bonus but with time that Bonus ended. I hope to get a better service as customer of Komodo in the future?

Sorry, not a nice message by myself ...

I don't like ...

1. To give persons free versions to make publicity. We have a small community and I can't see that we have privileged classes of users. A good publicity for a new version is very easy to do in self work.

2. To give information how strong is the engine and to compare Komodo with Clones or with the great Stockfish developement.

3. With Abandon ... means the development ... in my opinion with such things ... that is history!

Sorry!

Best
Frank
I like computer chess!

Frank Quisinsky
Posts: 4852
Joined: Wed Nov 18, 2009 6:16 pm
Location: Trier, Germany
Contact:

Re: And to the address of the SF webmaster ...

Post by Frank Quisinsky » Thu Aug 28, 2014 5:00 pm

Hello again,

and to the Stockfish Team address ...

To write ...
Stockfish
The strongest chess engine in the world.

Is bullshit too!
Sorry, but it isn't clear so far, means all the time!

I am not very happy if I am looking what is happen in computer chess. Normely we should a bit learning from msitakes in the past.

A bit more sensible please!
And all is OK!

Best
Frank
I like computer chess!

User avatar
reflectionofpower
Posts: 1456
Joined: Fri Mar 01, 2013 4:28 pm
Location: USA

Re: And to the address of the SF webmaster ...

Post by reflectionofpower » Thu Aug 28, 2014 5:13 pm

Frank Quisinsky wrote:Hello again,

and to the Stockfish Team address ...

To write ...
Stockfish
The strongest chess engine in the world.

Is bullshit too!
Sorry, but it isn't clear so far, means all the time!

I am not very happy if I am looking what is happen in computer chess. Normely we should a bit learning from msitakes in the past.

A bit more sensible please!
And all is OK!

Best
Frank
In the big three, Stockfish 5,Houdini 4, & Komodo 7a there is not much separating them from one of them being the strongest. Gull 3 is not far behind either. This is the golden age for computer chess software users but not the ones trying to make a few bucks off their hard work.
"hodie mihi, cras tibi"

Lonnie

Frank Quisinsky
Posts: 4852
Joined: Wed Nov 18, 2009 6:16 pm
Location: Trier, Germany
Contact:

Re: And to the address of the SF webmaster ...

Post by Frank Quisinsky » Thu Aug 28, 2014 5:25 pm

Hi,

sorry, big three are Stockfish, Komodo and GullChess.
Houdini is a copy of Robbolite and have nothing to search in the list of TOP-3.

Have a look in SmarThink development.
Around 200 ELO since engine is available again.

200 ELo more reach in an own development. That is a programmer I like. It is not a big wonder what the Rybka or the Houdini programmer do with sources of others. A good programmer can do the same, but the most really good programmers are working in here own code.

But you have right ...
The developments of Komodo and Stockfish are great for us. The development of other engines are great too ... have a look what the programmer of Vajolet do, or the programmer of Nirvanachess, Exchess, Arasan over all the years.

What the community or users have in here own brain is that what the commercial world pumps in our brains. But we have enough people with are able to looking more in detail.

Computer chess isn't go with a good example in front to others. We have really a very very bad call and the reason is ...

What the community do with cloens!
In Computer Chess we celebrate clones!

Ugh, yuck!

Best
Frank

And a Commercial programmer gave the Information ... we are better as a clone in Elo ... normaly we should no speak about it. We give more and more clones a platform. A helper / programmer "Larry Kaufmann" with gave so many Information what he do for Rybka celebrate the copy of it own work with a comment ... we are better as Houdini. With other word, Larry is better as Larry? Perhaps Larry is now a Partner of ChessBase and of Houdini? Times changed!

What for a Nonsens in my opinion ... but perhaps I don't understand the World. Perhaps a Person can explain me the situation? I am new in computerchess without any Information about it ... please explain me the situation in easy English!
I like computer chess!

Post Reply