Has anyone ever gotten their moves to depth 99? What kind of computer should I build to reach higher depth analysis faster?
depth of 99 is more than enough, some GMs say even 24 or 25 is enough. You don't need 99!!!
The computer can sometimes get to depth 99 when there is a forced mating combination, after it finds the mate the depth often scrolls up quickly to the limit.
In middlegame positions getting to around 30 usually takes a few minutes, in the endgame there are fewer possibilities to analyse so it can go somewhat deeper, to somewhere around 50.
If you want a computer that analyses faster, what you need is fast RAM access, chess engines store codes that identify positions with evaluations, depth an associated data in a file called the "hash table", accessing this data is usually a major bottleneck in the computation.
In addition, I would suggest not using the engine in a browser, download the latest Stockfish ( http://abrok.eu/stockfish/ has good Linux versions, the Windows versions at http://chess.ultimaiq.net/stockfish.html are slightly faster) and run it in Arena ( http://www.playwitharena.com ) or another GUI (the Fritz GUI is very good but rather expensive, many people buy Fritz just for the GUI, as the engine is not nearly as strong as Stockfish).
You will often see three versions labelled "BMI", "modern", and "standard", I listed them from fastest to slowest (functionally, they are equivalent), use the fastest your hardware can run, BMI can be run only by Haswell processors, while "standard" can run on essentially any 64 bit system.
Most engines can run on multiple processors at once, Stockfish scales fairly well using LazySMP, but you will never get N times the speed from N times the cores, for this reason you should avoid multithreading, as that reduces efficiency.
Finally, some additions can really add power to an engine:
An opening book is a collection of pre-analysed opening positions, this can give you deep analysis of many opening positions very quickly, as the engine only has to look them up in a file, Cerebellum (used by BrainFish) is a very good one.
An endgame tablebase is a table of exactly solved endgame positions, you can download 5-man syzygy (around 20 GB, I believe) or 6-man syzygy (around 120 GB, I believe), "6-man" means it has all positions with up to 6 total "men" (including kings and pawns) included. A tablebase is looked up every time the engine reaches a position with that number of pieces, and can drastically improve endgame performance.
In middlegame positions getting to around 30 usually takes a few minutes, in the endgame there are fewer possibilities to analyse so it can go somewhat deeper, to somewhere around 50.
If you want a computer that analyses faster, what you need is fast RAM access, chess engines store codes that identify positions with evaluations, depth an associated data in a file called the "hash table", accessing this data is usually a major bottleneck in the computation.
In addition, I would suggest not using the engine in a browser, download the latest Stockfish ( http://abrok.eu/stockfish/ has good Linux versions, the Windows versions at http://chess.ultimaiq.net/stockfish.html are slightly faster) and run it in Arena ( http://www.playwitharena.com ) or another GUI (the Fritz GUI is very good but rather expensive, many people buy Fritz just for the GUI, as the engine is not nearly as strong as Stockfish).
You will often see three versions labelled "BMI", "modern", and "standard", I listed them from fastest to slowest (functionally, they are equivalent), use the fastest your hardware can run, BMI can be run only by Haswell processors, while "standard" can run on essentially any 64 bit system.
Most engines can run on multiple processors at once, Stockfish scales fairly well using LazySMP, but you will never get N times the speed from N times the cores, for this reason you should avoid multithreading, as that reduces efficiency.
Finally, some additions can really add power to an engine:
An opening book is a collection of pre-analysed opening positions, this can give you deep analysis of many opening positions very quickly, as the engine only has to look them up in a file, Cerebellum (used by BrainFish) is a very good one.
An endgame tablebase is a table of exactly solved endgame positions, you can download 5-man syzygy (around 20 GB, I believe) or 6-man syzygy (around 120 GB, I believe), "6-man" means it has all positions with up to 6 total "men" (including kings and pawns) included. A tablebase is looked up every time the engine reaches a position with that number of pieces, and can drastically improve endgame performance.
Thanks Allonautilus, you summed that up perfectly.
This topic has been archived and can no longer be replied to.