Dream Engine Matches @ 4.5 GHz

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

Moderator: Ras

LiquidNitrogenOverclocker

Re: Dream Engine Matches @ 4.5 GHz

Post by LiquidNitrogenOverclocker »

mhalstern wrote:Yes, this is Kilo Knodes Per Second.

When I run your program, will it create any registry keys or put any dll's anywhere.
No Marc,

Nothing like that. It will save some files to your disk, over 800 MB worth, so you might want to install the program in a folder on your desktop before you begin.

The important file it creates is "report.txt" which will show the time taken to run on the last line of the file. It also will have ASCII drawings of the longest wins.
LiquidNitrogenOverclocker

Re: Dream Engine Matches @ 4.5 GHz

Post by LiquidNitrogenOverclocker »

Whoever sent me an email from this thread, please send it again. I tried to reply but your email address bounced.
LiquidNitrogenOverclocker

Re: Dream Engine Matches @ 4.5 GHz

Post by LiquidNitrogenOverclocker »

Robert Flesher wrote:Interesting, I have the same setup. However, I had no idea it could be overclocked to such extremes.
Well, it's not rocket science by any means, but it does require a great deal of careful testing. The uncooled CPU @ 4.5 GHz will reach 90 degrees Celcius (194 Fahrenheit) and it is constantly putting out that much heat.

You need some radical cooling design to deal with it, and this is not easy. Trying fitting it all into the same case, and you see why the original design looked like this:

Image
mhalstern
Posts: 484
Joined: Wed Nov 18, 2009 1:09 am

Re: Dream Engine Matches @ 4.5 GHz

Post by mhalstern »

Ok,

I'l run it tomorrow morning, before I leave for work. I'll report the results.

I have 6 gig of DDR3 Ram, if that is important for the result.

What Checker Programs can use your database?

Before you tie up the pc for 6 -7 months, get a copy of Rybka 3, and download a good opening book and purchase a playchess.com account. Play 500 Blitz Games and see if you can make #1. If you don't have the time and don't mind spending a little money, play a few games to see how good your hardware is.
LiquidNitrogenOverclocker

Re: Dream Engine Matches @ 4.5 GHz

Post by LiquidNitrogenOverclocker »

mhalstern wrote:Ok,

I'l run it tomorrow morning, before I leave for work. I'll report the results.
Cool, thanks.
mhalstern wrote: I have 6 gig of DDR3 Ram, if that is important for the result.
DDR3 RAM is awesome. The program uses just under 1 GB to load all of its data in RAM, so you will be fine.
mhalstern wrote: What Checker Programs can use your database?
I've been chatting with the programmer of GUI Checkers and sent him my access code and loading routines for the 6-piece database you are computing. He's already sent me a revised version that can use these databases. I made sure that the lookup time to retrieve a result is faster than any move generator can generate a single move, so the beauty of this is the program is NEVER slowed down by a database probe! It plays a wicked game.

I'm also writing my own program, called Only Perfect Checkers. The name is kind of a pun. How can something be "only" perfect? :) But the other meaning of the name is that is "only" probes the perfect-play databases (what the chess community would call Distance To Win Tablebases) and it will use the database hits to guide its search. There will be a "material only" evaluation function, but no positional terms at all.
mhalstern wrote: Before you tie up the pc for 6 -7 months, get a copy of Rybka 3, and download a good opening book and purchase a playchess.com account. Play 500 Blitz Games and see if you can make #1. If you don't have the time and don't mind spending a little money, play a few games to see how good your hardware is.
I'll probably do something like that. I'm still a little ways away from starting the computation run. I'm still writing the parallel building manager, which will execute the next portion of the build based on the previously solved data. The big difference is I will have 4 computers x 8 cores each doing work, saving data to both SSD drives and 10,000 RPM drives. I need to put the portions that will most likely need the fastest lookup times from disk on the SSD drives, but some of these occur later in the build. I don't want to have these drives idle for too long, so I need a mixed approach if I want to save months of computation.

I am still staging this and doing tests.
mhalstern
Posts: 484
Joined: Wed Nov 18, 2009 1:09 am

Re: Dream Engine Matches @ 4.5 GHz

Post by mhalstern »

It is running now. It is only using one core, as the task manager reports 25% cpu usage.
LiquidNitrogenOverclocker

Re: Dream Engine Matches @ 4.5 GHz

Post by LiquidNitrogenOverclocker »

mhalstern wrote:It is running now. It is only using one core, as the task manager reports 25% cpu usage.
That is correct, this benchmark does no parallelizations.
Mark
Posts: 216
Joined: Thu Mar 09, 2006 9:54 pm

Re: Dream Engine Matches @ 4.5 GHz

Post by Mark »

Ed- Any plans to write a strong checkers program for the iPhone??
LiquidNitrogenOverclocker

Re: Dream Engine Matches @ 4.5 GHz

Post by LiquidNitrogenOverclocker »

Mark wrote:Ed- Any plans to write a strong checkers program for the iPhone??
Funny you should mention that. I did play against one over the Thanksgiving weekend over a friend's house. It was not too strong, it did maybe 9 ply searches. The GUI on it was incredible though, I was impressed.

I don't think the current generation of iPhones would support loading 854 MB of databases (the size of my 6-piece Distance To Win) and if I drop the database size down too far (to 4 pieces) I get the feeling that games vs. my checkers iPhone app would be shown on too many pages of checkers programmers as having crushed it.

By the time I figure out how to create an iPhone application, I'm betting it could load the 6-piece databases.

:)
mhalstern
Posts: 484
Joined: Wed Nov 18, 2009 1:09 am

Re: Dream Engine Matches @ 4.5 GHz

Post by mhalstern »

This is the end of report.txt:

Thursday, March 04, 2010 @ 13:04:46
This slice of the database was solved in 0 hours, 0 minutes, 52 seconds.

Thursday, March 04, 2010 @ 13:04:46
#################### This entire database was solved in 5 hours, 10 minutes, 10 seconds. ####################

Thursday, March 04, 2010 @ 13:04:46
Database solving run completed in 5 hours, 36 minutes, 6 seconds.

If you want me to paste the entire file to this thread of ftp it somewhere, let me know.

Later