Author: Kolss
Date: 07:35:13 06/07/05
Go up one level in this thread
On June 07, 2005 at 09:17:25, Torstein Hall wrote: >On June 07, 2005 at 06:56:31, Uri Blass wrote: > >>On June 07, 2005 at 04:23:07, Kolss wrote: >> >>>On June 07, 2005 at 02:09:56, Uri Blass wrote: >>> >>>>2)I do not believe that hydra's hardware give hydra advantage that is equivalent >>>>to being more than 10 times faster. >>> >>> >>>Hello Uri, >>> >>>At the last IPCCC (Feb. 2005) I had the (dubious?!) pleasure to play Hydra (the >>>16-way version) with Ikarus. Ikarus was running on a Centrino 2.0 GHz which is >>>approximately equivalent in terms of speed to an "Athlon 3800+". We (i.e. >>>Chrilly, Christopher (Lutz) and I) discovered that Hydra plies seemed to be more >>>or less equal to Ikarus plies with respect to what is seen (i.e. tactical >>>horizon). Hydra was reaching three, sometimes four more plies than Ikarus: >>>Ikarus would play a move that Hydra had not expected, Hydra's evaluation would >>>jump up, and Ikarus would need one more move to see that it had been bad; or >>>Hydra would expect a move with a good score (for itself, of course), Ikarus >>>would "plan" something different with OK evaluation, then after two or three >>>minutes it would fail low, see what was wrong and eventually change to the move >>>expected by Hydra; in gereral, Ikarus would more or less show the evaluation >>>that Hydra had had two moves earlier in the game. >>> >>>Three to four plies (very) roughly translates into a factor of 20 (+- perhaps 5) >>>with an average effective braching factor. >> >>Thanks for the information but it does not prove nothing because Hydra may have >>better software than Ikarus. >> >>Uri > >Or perhaps Ikarus has better software than Hydra.... > >Torstein Hi, Thanks for trying to defend Ikarus... :-) In reality, I would guess that Hydra's software is not significantly inferior to that of a two-year-old amateur program - there is a whole professional team working on it full-time. On the other hand, Hydra's search is probably relatively "standard" (whatever that means) as compared to say Shredder. And then again, it is more or less meaningless anyway to speak about software vs. hardware in Hydra; they are so intrinsically intertwined in their approach that considering them completely separately is IMHO not possible any more. You could try and pit Hydra just using one processor (and of course one FPGA) against Ikarus (or any other "software" program) and see what happens - in fact, I have no idea what the outcome would be, though I would rather put my money on Hydra -, but still it would *not* be comparing software with software. Best regards - Munjong.
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.