Computer Draughts 2012

[BertTuyt]

First of all the best 2012 wishes to all of you, and everyone near to you.

Last year I posted below text in this forum

To DO:
* Improvements in the search efficiency (as I still believe Kingsrow out searches Damage in critical positions)
* Improvements in the Eval functions (as a result of match play)
* Improvements in the Endgame DB, I guess I to much approach the DB SSD in situations where it is not really needed.
* More matches against KingsRow (if Ed agrees )
* Communicate often in this forum

Don't DO
* I will not start with the generation of the 8p DB (see next item).
* I will not buy a new computer. I will wait for 2012, and then i hope to get my 16 core ( 2 Sandy Bridge Processors) 48 GByte + 500 GB SSD computer system
* I will most likely not participate in any tournament, as I don't expect to have a significant better program as compared with 2010 (which for me always was a reason to compete)

Next to that I hope competition will also improve , in 2011 every program really should have 7p DB's and parallel search ................

In terms of the To DO list I made some progress with the Endgame DB (where I found a bug), and at least I communicated often in this forum.
I played several matches against KingsRow, which is now relatively easy as I have a KingsRow version with 6p DB now (Ed , thanks ).
Im working during this XMas holiday on the search efficiency , both in terms of detecting bad moves and expanding the search in critical positions.
As usual the first implementation shows worse game play compared with the base version ...

So herewith my To DO and Dont Do list for 2012:

To DO:
* I will keep trying to improve the search, without some major improvements KingsRow will be out of sight for some time.
* For the Eval function I will write a new breakthrough routine.
* More matches against KingsRow. My record score in a 158 games match was 16 games lost, I really want to get into the single digit loose scores (to start with) ....
* I will participate in the Dutch Open 2012
* Communicate often in this forum

Dont DO:
* I will (still) not start with the generation of the 8p DB.
* I will not buy a new computer, as we are going through a huge reconstruction of the house, and it seems that "we" have other priorities.

Im also very glad that we have additional competition trough Maximus (thanks JJ), hope the other programs/programmers will catch up...

Keep you all posted, and hope to see some of you in 2012


Bert

[TAILLE]

I wish you an happy new year 2012 and of lot of success in your projects (except if you want to win against Damy )

2011 has been the year of a major restructuration of Damy and 2012 will be the year for the stabilization of all the work done in all domains: new search algorithm, eval function, breakthrough detection, attack defense of an outpost, bad move detection and corresponding reduction/pruning, extension of critical situations, loop handling, GUI etc. etc.
In order to debug and stabilize my new version I planed to play a lot of thematic matches against Kingsrow (I mean the commercial version).

BTW Bert you wrote a sentence that looks very unclear for me:
“I will keep trying to improve the search, without some major improvements KingsRow will be out of sight for some time”
We all know perfectly that the great majority of positions seen during a search are completely uninteresting because they are in variants in which one or more bad moves have been played previously. I guess we all built a search in which bad moves leads to a reduction or a pruning and I suspect the main differences between our search algorithms is the detection of these bad moves and the corresponding tuning for reduction and/or pruning.
What do you meant by the sentence above? Do you think you have a problem with the search itself or do you have a problem with the detection and the handling of bad moves, or is it something else?

[BertTuyt]

Gerard, thanks for your post.

What i meant is that my handling of bad moves and the extensions of interesting (forced) lines is not as effective as the KingsRow search implementation.
At least in specific positions i see the search-depth of Kingsrow explode , next to its score, and the program is locked on the target, which unfortunately is Damage.
So i need to be more clever here...

Bert

[BertTuyt]

For the real computer draughts (power hungry) programmers, Dell introduced a new workstation, the T7600.

A dual processor system, 2 * Intel E5-2600 ( so 2 * 8 cores).
In total 16 DIMM slots ( 8 per processor), so using ""normal" 8 GByte DIMM --> 128 GByte.
For fast endgame DB access a 512 GBYte SSD would be welcome (for a 8P DB).

Think however I have an issue to get in through the local financial approval committee, maybe Ed has some creative ideas

Bert

[Rein Halbersma]

For the real computer draughts (power hungry) programmers, Dell introduced a new workstation, the T7600.

A dual processor system, 2 * Intel E5-2600 ( so 2 * 8 cores).
In total 16 DIMM slots ( 8 per processor), so using ""normal" 8 GByte DIMM --> 128 GByte.
For fast endgame DB access a 512 GBYte SSD would be welcome (for a 8P DB).

Think however I have an issue to get in through the local financial approval committee, maybe Ed has some creative ideas

Bert

do you have a link with the price tag?

[BertTuyt]

Rein, think this machine will be introduced in May, so no prices available yet?

What do you think about this setup??
Think that also Gerard has a dual-processor Dell but i'm not 100% sure...

Bert

[BertTuyt]

And here a video of the T7600 ....

http://www.dell.com/content/topics/topi ... =04&~ck=mn

Bert

[TAILLE]

Hi Bert

Rein, think this machine will be introduced in May, so no prices available yet?

What do you think about this setup??
Think that also Gerard has a dual-processor Dell but i'm not 100% sure...

Bert

Yes Bert I have a Dell T7500 2*6cores, 3,33GHz 24Gb RAM

[BertTuyt]

Gerard,

just for my curiosity.
How long does it take (on your machine) to generate the full 8P DB?
I assume you used a parallel implementation for the dB-generation?
Does this DB-process scale well with the number of cores (or are there other bottlenecks....).

Could you make a prediction of the total time needed when you would use a 16-core machine with 128GByte - 256 GByte + a 512GByte SSD (to store all DB's) ?

Bert

[TAILLE]

Hi Bert,

Gerard,

just for my curiosity.
How long does it take (on your machine) to generate the full 8P DB?
I assume you used a parallel implementation for the dB-generation?
Does this DB-process scale well with the number of cores (or are there other bottlenecks....).

Could you make a prediction of the total time needed when you would use a 16-core machine with 128GByte - 256 GByte + a 512GByte SSD (to store all DB's) ?

Bert

I am not able to answer your question because because this machine is new for me and I used a far less powerfull machine to generate my 8 pieces egdb. Yes I used a parallel implementation for the db-generation based on (only?) 8 cores and 12Gb RAM. I guess it will take less than 2 months to generate the full 8 egdb with my new machine but it is only a feeling.
In any case it is not really significant to reach 1 month rather than 3 months because the more significant delay is the program development isn't it?
For the egdb generation process I do not think that SSD could help. In Damy process I use the disk only for backup in order to save intermediate results (every 3-4 hours) to recover after a possible crash. You may even run this backup task as a backgroung process with almost no impact on your main calculation process.
I consider that SSD is certainly more interesting for a real game than for the egdb generation process.

[BertTuyt]

At least in Holland you now can order the Dell T7600.

A system with 2 * the 2.7 GHz E5 2680 ( 8 core Xeon) and 128 GByte RDIMM cost around 11K Euro (including tax and delivery).

The bad news , dont think the local financial system will approve.

Maybe that ordering a base system and then getting the 2nd processor and the bulk of memory yourself is much cheaper.
Anyway , toys for boys...

Bert

[BertTuyt]

For the experts in self-assembly.
A base T7600 with an E2620 2.0 GHZ ( 6 core) and 8 GByte RAM is around 2,5K Euro (incl tax and delivery).
So all can do the math further.....

Bert

[Rein Halbersma]

For the experts in self-assembly.
A base T7600 with an E2620 2.0 GHZ ( 6 core) and 8 GByte RAM is around 2,5K Euro (incl tax and delivery).
So all can do the math further.....

Bert

Buying such high-end servers from Dell might not be the best idea. E.g they charge > EUR 800 for 4x8 Gb of 1600Mhz DDR3 ECC RAM, which is available on Newegg for < $300; which is 3 times cheaper! Unfortunately, Newegg doesn't ship outside the US, but memory is being overcharged by Dell.

[BertTuyt]

Rein, you are right.

A base system with 2 * E5 2680 ( 2.7 GHz 8 core) + 8 GByte RAM is around 6.85 K.
So with 16 * 8 GByte DDR3 ECC RAM added (just replace the 8 GByte of the base system), you might end up around 8K.
Nevertheless still outside the budget approval system...

For Intel these Xeon chips are still the Jackpot.
Maybe you can find them a little less expensive than via Dell, but even then i dont expect you get this equivalent power for under 7K.
So waiting might help.
On the other hand in 1 year (or later) we all want those Xeon 16-core chips, ....

Bert

[BertTuyt]

All quiet on the western front?

It seems that all programmers are enjoying a long summer shutdown/vacation.
Hope to hear some great news from you all
Anyway I was working on a new architecture for the Engine, implemented/coded some new ideas for the search, en changed some evaluation routines.

As already posted, the new search is quit fast, but so far instable at times. I found some bugs which i removed, so hope to provide some updates in the near future.

Bert