Searching for Deep Blue

[Reinfeld]

Questions:

1. Was this engine/beast (or its predecessors) ever imported to any machine or software?

2. If not, could it be? (The marketing opportunities, though dimmed by time, are interesting)

3. How would this program fare against 21st-century counterparts?


- R.

[IanO]

1. Was this engine/beast (or its predecessors) ever imported to any machine or software?

No. Feng-hsiung Hsu had some ambitions of doing a private run of his VLSI chess chips for an independent, single-chip product after the final match, but I don't think he ever raised the funding required to do the manufacturing run. Details are in his book "Behind Deep Blue".

2. If not, could it be? (The marketing opportunities, though dimmed by time, are interesting)

Probably not. It was an incredibly complex parallel hardware/software system, with plenty of bugs still to be worked out. Also don't forget that the secret sauce were those custom chips, which were only used at the tips of the search tree for the final half-dozen ply of search and evaluation.

The closest we've had to something similar was Hydra, the Saudi-backed attempt at a chess supercomputer from Chrilly Donninger. Its chess chips were implemented in FPGAs instead of custom fabbed chips. Funding for that project ran out in 2009, with it only winning the IPCCC against computers in 2004 and 2005, ahead of Fritz and Shredder.

3. How would this program fare against 21st-century counterparts?

Thus begins a thousand post thread! Objectively, we will never know. But honestly, processors and computer chess have advanced significantly since 1997. I doubt even with its speed, massive parallelism, and singular-search extensions that Deep Blue would even get into the top twenty of modern rating lists. Deep Blue averaged depths of about 12 ply, Hydra got to about 18 ply, and modern engines are getting over 25 ply at tournament TCs.

[spacious_mind]

1. Was this engine/beast (or its predecessors) ever imported to any machine or software?

No. Feng-hsiung Hsu had some ambitions of doing a private run of his VLSI chess chips for an independent, single-chip product after the final match, but I don't think he ever raised the funding required to do the manufacturing run. Details are in his book "Behind Deep Blue".

2. If not, could it be? (The marketing opportunities, though dimmed by time, are interesting)

Probably not. It was an incredibly complex parallel hardware/software system, with plenty of bugs still to be worked out. Also don't forget that the secret sauce were those custom chips, which were only used at the tips of the search tree for the final half-dozen ply of search and evaluation.

The closest we've had to something similar was Hydra, the Saudi-backed attempt at a chess supercomputer from Chrilly Donninger. Its chess chips were implemented in FPGAs instead of custom fabbed chips. Funding for that project ran out in 2009, with it only winning the IPCCC against computers in 2004 and 2005, ahead of Fritz and Shredder.

3. How would this program fare against 21st-century counterparts?

Thus begins a thousand post thread! Objectively, we will never know. But honestly, processors and computer chess have advanced significantly since 1997. I doubt even with its speed, massive parallelism, and singular-search extensions that Deep Blue would even get into the top twenty of modern rating lists. Deep Blue averaged depths of about 12 ply, Hydra got to about 18 ply, and modern engines are getting over 25 ply at tournament TCs.

Hi Ian,
A question and I might totally off base here. I thought I remember reading that Hydra was a Crafty clone?

As I said I might be totally wrong but this somehow sticks in my memory,

Regards
Nick

[spacious_mind]

I knew I read it somewhere. Not Crafty but Fruit. I wonder if this is the same Hydra?

http://computer-chess.org/doku.php?id=c ... ngine_list

Regards
Nick

[IanO]

Hi Ian,
A question and I might totally off base here. I thought I remember reading that Hydra was a Crafty clone?

As I said I might be totally wrong but this somehow sticks in my memory,

Regards
Nick

Oh, definitely not! Hydra was a continuation of Chrilly's own Brutus engine. In fact, the project started from a comment by Ken Thompson of Belle fame that FPGAs were becoming fast and large enough to implement custom chess logic as was done in Belle, HiTech, Deep Thought, etc.

More history of the project here:

https://chessprogramming.wikispaces.com/Brutus

One nice thing about custom chess hardware: the effort is so great and so unique that it completely eliminates the possibility of cloning!

[IanO]

I knew I read it somewhere. Not Crafty but Fruit. I wonder if this is the same Hydra?

http://computer-chess.org/doku.php?id=c ... ngine_list

Regards
Nick

No link was provided, but it can't be. The precursor program Brutus was started in 2000, several years before Fruit was a twinkle in Fabien's eye!

There may be knock-off UCI engines called both Brutus and Hydra which have no relation to Chrilly's work.

[spacious_mind]

I knew I read it somewhere. Not Crafty but Fruit. I wonder if this is the same Hydra?

http://computer-chess.org/doku.php?id=c ... ngine_list

Regards
Nick

No link was provided, but it can't be. The precursor program Brutus was started in 2000, several years before Fruit was a twinkle in Fabien's eye!

There may be knock-off UCI engines called both Brutus and Hydra which have no relation to Chrilly's work.

Well I am glad to hear that it is not. It put a damper on me when I saw it on a clone list.

Regards,

[RadioSmall]

Questions:

1. Was this engine/beast (or its predecessors) ever imported to any machine or software?

2. If not, could it be? (The marketing opportunities, though dimmed by time, are interesting)

3. How would this program fare against 21st-century counterparts?


- R.

Your third question has been debated for many years on the old talkchess forum ........ Deep Blue calculated at 126 Million positions per second , but still got on average 12 plys in middle games at tournament time controls....Today with a Cell phone running Droidfish for example we easily reach 23 plys at tournament time controls . Deep Blue's search was very brute force while todays programs prune heavily to get to the high depth ...It is highly improbable , however that a brute force 12 ply search can withstand a 23 ply heavily pruned seach .......the difference in search depth is just too much........