Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

ACM Tournament - Specs

11 views
Skip to first unread message

Jim Bumgardner

unread,
Jun 28, 1994, 3:51:24 AM6/28/94
to
Here are the specs on the machines which completed in the tournament.
Hopefully Hal or Steve will post the results for rounds 3-5 soon.

Note that the nodes per second vary greatly from position to position
- they are provided to give you a sense of the order of magnitude.


Cray Blitz: Fortran + Cray assembly, 4 proc. C94 (1000 mip). 150k for
executable code + space for hash tables. Uses dynamic tree splitting,
a variation of PVS which minimizes idle time and sync wait times by
using a sort divide and conquer algorithm. Book 300K+. 750k nodes
per second.

Deep Thought II: C + microassembly, IBM RS/6000 580, Memory space:
100 mb; 100 mips Runs on 12-16 processors. Endgame DB from Thompson's
CD-ROMS. 4000k nodes per second. Note: this is not the new "Deep
Blue" hardware, as some hoped to see, but the old hardware running
new software.

Evaluator: C, IBM PC 486 DX4 100 Mhz, 8mb, 131k for executable code.
Book 15K positions. 7K nodes per second.

Innovation II: C, Mac Power PC (some games were played in emulation mode due
to crashes, others native), 20mb Executable code: 435k, Book: 6500 entries,
Endgame DB: KPK, KQK, KRK, 4k nodes per second.

M-Chess Pro: C + Intel asm, Pentium 90 mhz, 11mb, 200kb for executable
code, Book: 300k moves. 9k nodes per second.

Now: Pascal, PC Clone, Pentium 90 mhz, Book: 7200 positions, hash table 256k
positions. KPK perfect endgame knowledge (rules, rather than table look up).

Spector: C, PC Clone 486 66Mhz with 256kb level two cache, 11 mips, executable
code 200k, 3 meg for data, book 200k positions, 3k nodes per second.

Star-Socrates: Threaded-C, Connection Machine CM-5, 512 processors, 16 gb,
Uses a parallelization of scout (pvs) search called Jamboree search.
Book 13k positions, KPK endgame database. 1000k nodes per second.

WChess: C, 90 Mhz Pentium, 16 mb, 200kb for executable code, Book: 105k moves.
37k nodes per second.

Zarkov: C, HP735, 87 mips, 175k executable code, book 17k positions; KPK
end game database. 256k pos hash table, 15k nodes per second.


--
--
-- Jim Bumgardner | jb...@netcom.com
-- Time Warner Interactive
-- GSC/MUd--p(-)c+++!lu+e-m*s/+n++(---)h--f!gw+t+ry++

Hal Bogner

unread,
Jun 28, 1994, 10:41:35 AM6/28/94
to
Thanks for posting the specs, Jim. Being at one of these tournaments is a bit
hectic, but here's the round by round results for rds 3-5, and the final
standings. Steve Edwards will probably have the games up within two days, if
not sooner. (I don't have them at hand.)

Rd 3:
The two leaders (both 2-0) played, and Star-Socrates beat M-Chess Pro
Deep Thought II beat WChess
Zarkov beat Now
Cray Blitz beat Innovation II
Evaluator beat Spector
As in human tournamnets, sometimes you have to play the fellow you are rooming
with; Zarkov and Now were doing so, as were Spector and Evaluator

leader: Star Socrates (3), DT, MChess, Zarkov (2)...

Rd 4:
Deep Thought II beat Star-Socrates in a brilliant game to grab a share of the
lead,
Zarkov and M-Chess Pro drew a long R&P ending
WChess beat Cray Blitz
Innovation II beat Spector in a wacky game with lots of hanging pieces
Now beat Evaluator

scores:
3-1 DT, Star-Soc
2.5 MChess Pro, Zarkov
2 Innovation II, Now, WChess
1.5 Cray Blitz
1 Evaluator
.5 Spector

Rd 5:
Deep Thought II beat MChess Pro handily to guarantee at least a share of first
Zarkov beat Star-Socrates in another long R&P ending, leaving DT II clear
first, and taking clear second,
Now beat Innovation II
Cray Blitz clobbered Spector
and Evaluator recovered from an apparent typo in its opening book (f3 in a
double-KP opening!?) to upset WChess

Final scores:
4-1 Deep Thought II
3.5 Zarkov
3 Star-Socrates, Now
2.5 Cray Blitz, MChess Pro
2 Wchess, Evaluator, Innovation II
.5 Evaluator

DT II was 4-0 in played games, though it was probably losing to Zarkov in the
middlegame in Rd 1; it's "loss" was a forfeit when the IBM facility lost all
power Saturday night prior to the 7 PM start time. The scheduled game against
MChess stayed on hold until 10:45 PM, when Hsu finally gave up hopes of
recovery. MChess Pro gave DT II a second chance in the last round....

Zarkov beat Star-Soc's monster machine, Cray's super-computer, and gave DT II
a run for its money; all on a machine about 1.5x faster than the Pentiums.
For me, this was very satisfying, having worked some with John Stanback since
1990. John does this in his spare time (he's an IC engineer at HP). He got
some help from IM Marc Leski over the last six months, and it looks like it
was worthwhile!

Innovation scored its first points after going 0-5 last year (the first time
it played in the ACM tournament)

Evaluator and Spector played in this event for the first time.

The games will be forthcoming soon!

-hal bogner
(usually unbiased...)

Feng-Hsiung Hsu

unread,
Jun 29, 1994, 8:30:57 AM6/29/94
to
In article <hmbCs4...@netcom.com> h...@netcom.com (Hal Bogner) writes:
>DT II was 4-0 in played games, though it was probably losing to Zarkov in the
>middlegame in Rd 1.

Maybe, maybe not. It was a case of bad opening preparation on our part. The
book ended too early, and it misplaced the king. Zarkov played energetically
for a while to exploit the king misplacement, but could not convert the
pressure to permanent advantages. White had to play very accurately to avoid
disasters nonetheless. Once white's queen rook got activated, it was all
downhill for black. The sudden collapse of black's position was kind of
surprising. But things can change rapidly in King's Gambit games.

I already said it to John himself, but congratulations once again to him for
an excellent performance. Couldn't happen to a nicer guy.

0 new messages