WCCC 2009 - Limit of 8 cores

[Sean Evans]

http://hiarcs.net/forums/viewtopic.php?t=2008

Times They Are A Changing

David Levy

During the first three decades in the history of computer chess tournaments there developed a debate over the question: what hardware should the contestants be allowed to use? For a while there were even suggestions, prompted by Tony Marsland, of handicapping programs according to the power of their hardware. I was never in favour of the idea of handicapping because I have always felt that it would be impossible to achieve a fair way of managing the process. Furthermore, in 1981, at the ACM Computer Chess Championship in Los Angeles, a program called Philidor, developed by my team in London and running on a lowly Z80 processor in an Osborne 1 computer, finished in equal 5th-6th place (with Duchess), well ahead of Tony Marsland’s program AWIT, which ran on an all powerful Amdahl mainframe and which we defeated in our individual encounter. After that particular tournament the idea of handicapping lost whatever lustre it might have had.

When microcomputers emerged as suitable vehicles for chess programs, it became clear the very best microcomputer chess programs were absolutely no match for the strongest programs running on mainframes, with the result that separate microcomputer chess tournaments became popular, with the first World Microcomputer Chess Championship taking place in 1980. The top programs running on micros also often competed alongside their mainframe colleagues in “open” tournaments, and with good reason – they were usually able to give a reasonable account of themselves. Thus the world of computer chess witnessed two strands of regular competitions – those for microcomputers alone and those for everyone. There have also been tournaments organized for “uniform platforms”, taking the hardware element out of the competition altogether. For several years Don Beal hosted and organized such tournaments at Queen Mary College in London, but they did not meet with the same level of enthusiasm from the chess programming community.

It was not long after the defeat of Garry Kasparov by IBM’s Deep Blue that the strongest PC programs were winning just about all of the open computer chess tournaments. And so in 2002 the ICGA changed the rules for the World Computer Chess Championships, enveloping all programs in one tournament. From then on PC programs usually took the top honours in our tournaments. Some programmers have been content to enter our championships on single processor machines, others have strived to use dual or quad machines or even more cores. And we are now seeing PC programs that run on systems with tens of cores – the more the merrier.

The ICGA feels that the time has come to take stock of this trend of hunting for astronomic numbers of cores, and to bring matters back to Earth. We see no point in organizing an event that can be won by a simple weight of processing power, when just about all the competing programs are able to use computers with only a handful of processors. By allowing 20 cores, or 40, or 80, at the present time, we would be saying to the vast majority of chess programmers that, if they want their program to be able to give of its best, they must first acquire the use of an expensive computer system with a very large number of processors. That is not what we believe the World Computer Chess Championship should be about. One should not be able to buy the title in this way.

The ICGA has therefore decided to change the rules of the World Computer Chess Championship, starting with our 2009 event. For 2009 no entry will be permitted to run on a system with more than 8 cores. We have chosen the number 8 because such systems are readily available in the retail computer market. We will review this number every year in accordance with whatever we feel is appropriate, given the retail market at the time. But for 2009 the message is – no more than 8 cores will be allowed for any program competing in the World Computer Chess Championship.

There will, however, still be some scope for more powerful hardware systems to show off their capabilities. Alongside the World Computer Chess Championship the ICGA organizes the annual Computer Olympiad, and when there are two or more contestants we will organize, as part of the Computer Olympiad, an open chess tournament (or match if there are just two entries) in which hardware systems with more than 8 cores may compete. This will NOT be given the status of a world championship event – it will be the Computer Olympiad Open Chess Championship.

The ICGA is not making the same 8-core rule applicable to the Computer Olympiads, for those games in which the strongest programs have not yet reached the level of world champion human players. The reason for the different rules for different games is that our competitive goal in ICGA events is to encourage and assist the development of programs capable of defeating human world champions. This target has already been achieved in Chess, hence our competitive goal for Chess is now principally to discover which program is the best. The corresponding target has not yet been achieved in Go and several of the other games contested at the Computer Olympiad, hence the ICGA’s competitive goal for those games is partly to encourage the fastest possible progress towards that target.

We believe that the above decision is in keeping with the wishes of the majority of games programmers who compete at the World Computer Chess Championships.

[Sean Evans]

Hi, I brought this topic in another post and I agree with the decision of having uniform hardware, but I would have just stated one-cpu and let the users decide on the one cpu, instead of core number. I believe this decision has a lot to do with Rybka having 40 cores certainly a ridiculous situation!

Cordially,

Sean

[Yar]

Hi,

This is redicilous from my point of view:
1) in 2009 you will be able to build 32 core system in one motherboard and 64-128 in 3-4 years
2) why it was not a problem when Zappa used 512 CPUs in world champ in Turino 2006 ?
3) Other engines also can use clusters whats the problem ? Problem probably is because other engines don't support clusters.
4) rybka on 8 core system would not be dramatically weaker then on running 40 core cluster.

Limiting book size is a good idea i think.

With best regards,
Yar

[Steve B]

Hi, I brought this topic in another post and I agree with the decision of having uniform hardware, but I would have just stated one-cpu and let the users decide on the one cpu, instead of core number. I believe this decision has a lot to do with Rybka having 40 cores certainly a ridiculous situation!

Cordially,

Sean

Hi Sean
there has been a fairly intensive discussion here about this issue
as you can see here Levy is perhaps reconsidering his position on this:

http://www.talkchess.com/forum/viewtopi ... 11&t=25458

Regards
Steve

[swami]

Hi,

This is redicilous from my point of view:
1) in 2009 you will be able to build 32 core system in one motherboard and 64-128 in 3-4 years
2) why it was not a problem when Zappa used 512 CPUs in world champ in Turino 2006 ?
3) Other engines also can use clusters whats the problem ? Problem probably is because other engines don't support clusters.
4) rybka on 8 core system would not be dramatically weaker then on running 40 core cluster.

Limiting book size is a good idea i think.

With best regards,
Yar

I agree that limiting the hardware is no brainer. But I disagree with the idea of limitations for the book size.

No limits. You can bring in the best out of book, TB's, hardware or whatever.

[Peter Skinner]

All this decision has done is make tournaments like the ACCA and CCT more important and might as well call them the _true_ World Championships.

Image telling telling any athlete that they have to use the same equipment as everyone else, or they can't compete. Why penalize a participant because he/she/team can muster up better hardware than anyone else?

I trust due to the new rules, the ICGA will no longer be deciding who is professional, semi-pro, and amateur, and the results fees will be removed?

If everyone has to be _equal_, then why state one is better than the other and collect higher fees due to programming ability or sponsorship?

Why sell your program if not to be able to afford the very best hardware to run it off of? Being commercial shouldn't mean you have to pay more in the way of fees. Hell get rid of them all together.

Oh and btw... the CCT is FREE to enter. Have fun in March.

[BubbaTough]

We believe that the above decision is in keeping with the wishes of the majority of games programmers who compete at the World Computer Chess Championships.

I have not talked to a single "games programmer" that agrees with this. They may be out there, but if so they constitute a significant minority. There may be a number who like the idea, including on this site. But not the programmers.

-Sam

[rebel777]

Any WC is about the best to win, in computerchess that means the program that plays chess best. Current rules are just fine. But nothing can stop David Levy to organize a second and simultaneous WC of limited and equal single cpu hardware provided by the organizer.

Ed

[Sean Evans]

Again, getting back to my original position, I suggest a standard for hardware. I think the best would be either one-cpu or two-cpu's with a predetermined set of RAM, say 6 Gigs...

Let the owners of the software decide on the type of cpu, ram and mobo they want to use, so if the best cpu is an Intel with 6-cores, then I presume most participants will use it, but they still have the flexibility.

What is the point of a match if Rebel program is playing with 40-cores and 50-gigs of RAM against Fritz with 4-cores and 2-gigs of RAM.

Can somebody on this thread tell me why this match of Rebel vs. Fritz is of any interest whatsoever? What does it prove or disprove?

Cordially,

Sean

[Yar]

Again, getting back to my original position, I suggest a standard for hardware. I think the best would be either one-cpu or two-cpu's with a predetermined set of RAM, say 6 Gigs...

Let the owners of the software decide on the type of cpu, ram and mobo they want to use, so if the best cpu is an Intel with 6-cores, then I presume most participants will use it, but they still have the flexibility.

What is the point of a match if Rebel program is playing with 40-cores and 50-gigs of RAM against Fritz with 4-cores and 2-gigs of RAM.

Can somebody on this thread tell me why this match of Rebel vs. Fritz is of any interest whatsoever? What does it prove or disprove?

Cordially,

Sean

512 CPUs Zappa didn't win wch in Turino in 2006. Hydra had superior hardware but had bad time vs rybka on playchess. Of course hardware is important but rybka probably on one CPU will win match vs Fritz running on 4 CPUs.