Crafty on Cray Blitz vs. Rybka3

[bob]

Hi, if Dr. Hyatt had played his Crafty program by Internet hook up on the Cray Blitz system running at his University, would it be strong enough to beat Rybka3 on the hardware it was using by long distance?

Thanks

Sean

We have never had a Cray here. The fastest machine we ever used was the Cray T932, where CB could search about 7M nodes per second. I regularly search 20M on an 8-core intel box today. Which means Cray Blitz would have a tough time beating Crafty today with a 3x speed handicap caused by the 1994-era hardware.

Cray Blitz is also way old and we don't have the best version available either. In 1994 we were using singular extensions and some other things that the released source does not have (the source is 1991-1992). It would not be very competitive today due to software improvements discovered over the past 15+ years, and due to the hardware improvements that would make the old T90 look slow by today's standards.

[bob]

Crafty has no chance even if get 30:1 hardware advantage
Rybka is very very strong.

The elo gap between rybka and Crafty is clearly more than 400 elo.


Uri

Or not...

[BubbaTough]

Or not...

Hehe, cannot argue with that. Trying to translate speed into ELO is a pretty funny business.

-Sam

[bob]

Or not...

Hehe, cannot argue with that. Trying to translate speed into ELO is a pretty funny business.

-Sam

My point was a bit subtle for most. A cluster is _not_ the best platform to run on. I have run on machines with 64 cores and shared memory. That is _far_ faster than a cluster with 64 cores and message passing, even if using something like Infiniband. Testers don't have a "good book" for Crafty. Too many variables to take a rating list and use that to estimate actual Elo differences in a real tournament.

[Uri Blass]

Or not...

Hehe, cannot argue with that. Trying to translate speed into ELO is a pretty funny business.

-Sam

It is clearly possible to compare ssdf ratings
Here is some data:

ssdf used the following hardwares
same programs were tested in different hardware

hardware that was used:

1)Q6600 2,4 GHz
2)Athlon 1200 MHz
3)K6-2 450 MHz
4)P200 MMX
5)Pentium 90 MHz
6)486/50-66 MHz

The difference between 6 and 5 for the same program is
83,77,109.46,72,79,56,38 elo(average near 70 elo)
The difference between 5 and 4 is
96,62,86,86,70 (average 80 elo)
The difference between 4 and 3 is
82,105,37,62,59 (average near 70 elo)

The difference between 3 and 2 is
121,133,111,106.99,53,103,62,104,75,68,61,100,49,78(average near 90 elo)

The difference between 2 and 1 is
135,123,101

ssdf has the following

7 Rybka 2.3.1 Arena 256MB Athlon 1200 MHz 2925 26 -25 775 59% 2857
71 Crafty 18.12/CB 256MB Athlon 1200 MH 2608 21 -21 1141 39% 2685

Both rybka and Crafty improved since that time but I believe that rybka improved more than Crafty.

The ssdf leader is
1 Deep Rybka 3 2GB Q6600 2,4 GHz 3238 53 -43 372 88% 2899

300 elo came both from better hardware better software and better book(the non hardware factor is probably near 200 elo).

Uri

[Werewolf]

[/quote]
We have never had a Cray here. The fastest machine we ever used was the Cray T932, where CB could search about 7M nodes per second. I regularly search 20M on an 8-core intel box today. Which means Cray Blitz would have a tough time beating Crafty today with a 3x speed handicap caused by the 1994-era hardware.[/quote]

I get about 2M nodes per second on my Core 2 duo 2.33. It's running Crafty 22.01 32 bit version. Is the 64 bit version a lot faster? If so where can I get it please?

[bob]

We have never had a Cray here. The fastest machine we ever used was the Cray T932, where CB could search about 7M nodes per second. I regularly search 20M on an 8-core intel box today. Which means Cray Blitz would have a tough time beating Crafty today with a 3x speed handicap caused by the 1994-era hardware.[/quote]

I get about 2M nodes per second on my Core 2 duo 2.33. It's running Crafty 22.01 32 bit version. Is the 64 bit version a lot faster? If so where can I get it please?[/quote]

my core-2 duo runs about 3M nodes per second using two threads. The position can affect this somewhat as endgame positions are faster. But here is an example:

time=22.86 mat=0 n=69677811 fh=90% nps=3.0M

So yes, 64 bit is faster. I don't provide executables since there are so many different architectures, operating system versions, library versions, etc. But many sites provide executables for most any platform.

[Werewolf]

So I'm looking for 64 Bit 22.01 Crafty? And there are lots of sites I can download this from, and it would be about 50% faster?

[Dann Corbit]

So I'm looking for 64 Bit 22.01 Crafty? And there are lots of sites I can download this from, and it would be about 50% faster?

Jim Ablett's builds are generally the fastest.
http://homepages.tesco.net/henry.ablett/jims.html

[Eizenhammer]

Or not...

Hehe, cannot argue with that. Trying to translate speed into ELO is a pretty funny business.

-Sam

My point was a bit subtle for most. A cluster is _not_ the best platform to run on. I have run on machines with 64 cores and shared memory. That is _far_ faster than a cluster with 64 cores and message passing, even if using something like Infiniband. Testers don't have a "good book" for Crafty. Too many variables to take a rating list and use that to estimate actual Elo differences in a real tournament.

I dont see how all this guessing could change the fact that rybka is about as strong as Uri suggests. Everybody who has both programs can see it, i ran a match on equal hardware, old hardware, ok, and in 50 games crafty got a draw, or maybe two, nothing more: Now, by using long books and effective learning one can obfuscate this difference, of course, but it does hardly change the basic facts. I would really be surprised if Crafty could win a long match, using the Silver suite or similar neutral suites, on all its 64 cores against a decent 64Bit single version of rybka ...
sorry, but sometimes one has to say the truth, even if it hurts.

Peter