Author: Robert Hyatt
Date: 20:27:47 07/21/00
Go up one level in this thread
On July 21, 2000 at 22:40:16, blass uri wrote:
>[D]2k1rr2/pp1n2p1/2p1n1p1/3pP3/5qPP/P4NR1/1PP1QP2/1K1NR3 w - - 0 1
>This is one of the positions from deeper blue-kasparov(game 4 before 23.Nc3)
>
>Deep blue claims that they could search 15 plies in 15 seconds that is
>200,000,000*15 nodes=3,000,000,000 nodes
>
>I am interesting to know if you can search 15 plies with no pruning and no
>extensions with less than 3,000,000,000 nodes.
Here is my numbers, on a quad xeon/550mhz machine.
9 plies took 2M nodes
10 plies took 5M nodes (this is 5M total from plies 1-10)
11 plies took 9M nodes
12 plies took 100M nodes
13 plies took 500M nodes
14 plies took 700M nodes
15 plies took 1300M nodes
If I could average 200M nodes per second, I could do that search in probably
under 5 seconds, given enough memory. If I could peak at 1B, I could do that
search between 1 and 5 seconds somewhere, depending on how the peak went...
Note that his 30% efficiency figure is an average as is my 3.2X faster on a
quad. I have many positions where I run 4x faster. I have a couple where
I run 1/10th as fast as one cpu...
For me, these numbers should be reduced by at least 25%, which is my search
overhead (extra nodes searched that a sequential search would not examine).
Hsu's 200M figure already had his overhead factored out...
I am not sure what this proves, when you factor in parallel search. Odd
things happen. Some searches go way fast. Others go way slow. Trying to
compare searches by comparing depths is not so useful. In some positions
I might extend way too much. In other positions they might do the same.
In other positions we might extend pretty equally. How to know and compare?
I could probably search this tree in less than 1/2 the nodes if I had a decent
sized hash table. This grossly overruns anything I can use on this machine
tonight...
Btw Crafty liked b4 here beyond depth=13... Prior to that it liked
Nc3 as did DB.
In any case, if you compare nodes, their number seems reasonable. Here is
my output:
10 5.21 1.23 1. Nc3 Ndc5 2. b4 Nd7 3. Qd3 Qf7 4.
b5 Ndc5 5. Qd2 Kb8 6. bxc6 bxc6
10-> 7.13 1.23 1. Nc3 Ndc5 2. b4 Nd7 3. Qd3 Qf7 4.
b5 Ndc5 5. Qd2 Kb8 6. bxc6 bxc6
11 9.21 1.23 1. Nc3 Ndc5 2. b4 Nd7 3. Qd3 Qf7 4.
b5 Ndc5 5. Qd2 Kb8 6. bxc6 bxc6
11-> 11.71 1.23 1. Nc3 Ndc5 2. b4 Nd7 3. Qd3 Qf7 4.
b5 Ndc5 5. Qd2 Kb8 6. bxc6 bxc6
12 52.70 1.01 1. Nc3 Nc7 2. Qe3 Qxe3 3. Rxe3 Re7
4. Ne2 Rfe8 5. Nf4 Nxe5 6. Nxe5 Rxe5
7. Rxe5 Rxe5 8. Nxg6
12-> 1:51 1.01 1. Nc3 Nc7 2. Qe3 Qxe3 3. Rxe3 Re7
4. Ne2 Rfe8 5. Nf4 Nxe5 6. Nxe5 Rxe5
7. Rxe5 Rxe5 8. Nxg6
13 2:34 0.66 1. Nc3 Nc7 2. Ka2 Re7 3. Qe3 Qxe3 4.
Rxe3 Rfe8 5. h5 gxh5 6. gxh5 Nxe5 7.
Nxe5 d4 8. Ng6 Rxe3 9. fxe3 dxc3 10.
bxc3
13 7:59 1.00 1. b4 Nc7 2. Qe3 Qxe3 3. Rxe3 Rf4 4.
Nb2 Re4 5. Rxe4 dxe4 6. Ng5 Nxe5 7.
Nxe4 Kd7 8. Nc5+ Kc8
(4) 13-> 9:08 1.00 1. b4 Nc7 2. Qe3 Qxe3 3. Rxe3 Rf4 4.
Nb2 Re4 5. Rxe4 dxe4 6. Ng5 Nxe5 7.
Nxe4 Kd7 8. Nc5+ Kc8
(3) 14 10:59 0.93 1. b4 a5 2. Nb2 axb4 3. Nd3 Qc4 4.
Nd2 Qb5 5. axb4 Kc7 6. Nf3 Ra8 7. Qd2
Kd8
14-> 12:40 0.93 1. b4 a5 2. Nb2 axb4 3. Nd3 Qc4 4.
Nd2 Qb5 5. axb4 Kc7 6. Nf3 Ra8 7. Qd2
Kd8
15 18:55 0.86 1. b4 a5 2. c3 axb4 3. axb4 Nc7 4.
Qe3 Re6 5. Kc1 Qxe3+ 6. Rxe3 Rfe8 7.
h5 gxh5 8. gxh5 R6e7 9. Kc2 Nxe5 10.
Nxe5 Rxe5 11. Rxg7 Rxh5 12. Rxe8+ Nxe8
15-> 22:21 0.86 1. b4 a5 2. c3 axb4 3. axb4 Nc7 4.
Qe3 Re6 5. Kc1 Qxe3+ 6. Rxe3 Rfe8 7.
h5 gxh5 8. gxh5 R6e7 9. Kc2 Nxe5 10.
Nxe5 Rxe5 11. Rxg7 Rxh5 12. Rxe8+ Nxe8
time=22:21 cpu=399% mat=1 n=1301818161 fh=15% nps=970702
>
>You can use evaluation of 0 for all the moves so you will always have a perfect
>order of moves.
>
>You can use hash tables.
>please tell me how many nodes you need to see
>1 ply,2 plies,....15 plies and how many plies do you need to fill the hash
>tables.
>
>repeat the experiment with smaller hash tables in order to guess if bigger hash
>tables can help significantly.
>
>Here is Deep blue logfile in the relevant position
>
>3(4)[Nc3](30) 30^ T=0
>nd1c3 Qf4g3r pf2g3Q Rf8f3n
> 3(5) 61^ T=0
>qe2e3 Qf4e3q re1e3Q Rf8f4 ph4h5 Pg6h5p pg4h5P
> 3(5) 68 T=0
>qe2e3 Qf4e3q re1e3Q Rf8f4 ph4h5 Pg6h5p pg4h5P
> 4(5) 78 T=0
>qe2d3 Ne6c5 qd3g6P Qf4g3r
> 5(5)[Qd3](74)[Qe3](82)[Nc3](86) 86 T=0
>nd1c3 Ne6d4 nf3d4N Qf4g3r pf2g3Q Nd7e5p
> 6(5)[Nc3](75) 75 T=1
>nd1c3 Pg6g5 ph4g5P Qf4g3r pf2g3Q
> 7(5) #[Nc3](78)############################# 78 T=2
>nd1c3 Nd7c5 qe2e3 Qf4e3q pf2e3Q
> 8(6) #[Nc3](61)############################# 61 T=5
>nd1c3 Nd7c5 pb2b4 Nc5d7 pb4b5
> 9(6) #[Nc3](74)############################# 74 T=15
>nd1c3 Nd7c5 pb2b4 Nc5d7 pb4b5 Ne6d4 nf3d4N Qf4d4n pb5c6P
>10(6) #[Nc3](65)############################# 65 T=102
>nd1c3 Nd7c5 pb2b4 Nc5a6 pb4b5 Na6c5 pb5c6P Pb7c6p kb1a2 Rf8f7
>11(6)[TIMEOUT] 65 T=159
>nd1c3
>
>Uri
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.