Author: Robert Hyatt
Date: 10:41:25 06/25/98
Go up one level in this thread
On June 25, 1998 at 12:25:32, Ernst A. Heinz wrote: >On June 25, 1998 at 12:06:14, Robert Hyatt wrote: > >>On June 25, 1998 at 06:35:40, Vincent Diepeveen wrote: >> >>> >>>On June 25, 1998 at 04:54:02, Roberto Waldteufel wrote: >>> >>>> >>>>Here's another question about Hsu's chess chip. I seem to recall reading some >>>>time ago that Hsu was considering a commercial release of his chip. Does anyone >>>>know anything more about this? If the chip were to become available, how could I >>>>use it in conjunction with a PC? would the fixed depth not be "out of sync" for >>>>the speed of, eg a Pentium 333Mhz if it was designed to work with a >>>>supercomputer, or can the fixed depth be adjusted to redo the balancing act in >>>>the new environment? If it were possible, I would be very interested in >>>>experimenting with this sort of hardware coupling. I assume that it would extend >>>>the depth to which a program could search by something like 4 extra plies within >>>>the same time. This would surely improve the strength of the PC ches programs >>>>quite a lot! >>>> >>>>Roberto >>> >>>You see it wrong. If you do 4 ply searches without hash etc, then >>>2.5 million drops quickly to say 300k nodes a second. >>> >>>So in fact you're playing against a kind of fritz5, which DOES search >>>all leafs fullwidth, which gives you some extra tactics, so commercial against >>>programs which are only tested at the same hardware and are only >>>busy with outbooking and trying to finish the game by means of tactics, >>>you beat with big numbers then, but it will play horrible. >>> >>>Vincent >> >> >>what are you talking about here? their chess processor most definitely has a >>hash table, and it most definitely has a good evaluation. And a 4 ply search >>without hash *does not* cut the speed by a factor of 10, except perhaps in a >>simple ending. In the middlegame it is not a factor of 2. 4 ply searches don't >>have many transpositions anyway. > >Bob, > >I am really not sure whether the DB-II chess processors have their own >local transposition tables (maybe the 4? on each chip share a small one) ... > >As far as I know, Hsu intentionally omitted them in earlier versions of >his chess processors due to space limitations and because the added >complexity does not really pay off with shallow searches. For, the chess >processors receive leaf nodes from many different and probably unrelated >parts of the search tree such that local transposition tables should not >be able to graft much information from one search to the next anyway ... > >=Ernst= The last discussion I had with him was a while back in Cape May. I think that he said that each "board" had a hash table that was accessible by all the processors on that board, but it was not shared between boards. I also base this on the fact that he "cloned" belle onto a chip, and belle also had a workable hash table, although without the "best move" feature as it was difficult to poke a specific move into his one-at-a-time move generator. But, at least for DB-2, they had a press release that had some tech specs, and I *think* it said XX MB per board. I just don't recall whether XX was 16 or 64. Part of his tree search, however, is "intelligent distribution" of nodes. So that on iteration N+1, you can try to give the same move to the same processor again, since it has searched it once already. This is a common idea on programs based on distributed memory, although I totally ignore it since I am a "shared memory" fan. :)
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.