Author: Zappa
Date: 09:37:12 12/27/05
Go up one level in this thread
On December 27, 2005 at 10:29:20, Djordje Vidanovic wrote: >On December 27, 2005 at 10:22:28, Uri Blass wrote: > >>On December 27, 2005 at 10:12:57, Djordje Vidanovic wrote: >> >>>On December 27, 2005 at 09:05:43, Greg Simpson wrote: >>> >>>>On December 27, 2005 at 08:44:18, Uri Blass wrote: >>>> >>>>>On December 27, 2005 at 08:36:24, JNoomen wrote: >>>>> >>>>>>Hi all, >>>>>> >>>>>>It's quite possible that the computerchess lists show something like this at the >>>>>>end of the year: >>>>>> >>>>>>1. Rybka 32-bit >>>>>>2. Fruit 2.2.1 >>>>>>3. Fritz 9 >>>>>>4. Shredder 9 >>>>>>5. Hiarcs 10 >>>>>> >>>>>>Now suppose we enter Hydra and Zappa in these lists. Condition: they both play >>>>>>on the SAME hardware as the above 5. So no program has a hardware advantage. All >>>>>>programs use their own book. >>>>>> >>>>>>Now comes the poll question: where do you think that Hydra and Zappa will be in >>>>>>the list given above? >>>>> >>>>>hydra cannot run on the same hardware and this was explained in the past also by >>>>>chrilly. >>>>> >>>>>He said that there are things that are cheap for hydra but very expensive if he >>>>>tries to do it on the same hardware that most use and that there are things that >>>>>are expensive for software on Normal PC so he even did not try to do it in Nimzo >>>>>but cheap for hydra. >>>>> >>>>>Uri >>>> >>>>Of course it should be possible to emulate Hydra's FPGAs in software, and it >>>>wouldn't surprise me if the Hydra team already has that for testing. I would >>>>expect Hydra to be by far the worst program listed in that situation. >>> >>> >>>I would expect Chrilly Donninger to have improved his "basic" program much >>>further and it is surely way above the old Nimzo. Thus I would expect it to be >>>on a par, if not even better, with the leading pack. >>> >>>Naturally, we are only speculating. >>> >>> >>>Djordje >> >>The question is not if Chrilly could improve nimzo but what is the strength of >>hydra if it is emulated in software. >> >>It is possible that in that case hydra could search only 1000 nodes per second. >> >>Uri > > > >That's exactly what I had in mind. I believe that a "simple" PC version (FPGA >emulation) of Hydra is probably equal in strength to the top pack (first 3 or 4 >programs). Just my "gut feeling", but a strong gut feeling. > >Djordje I really doubt it. Hydra on some sort of FPGA emulator would search about 4 nodes per second. I used to run simplescalar which simulates an out-of-order core. That slowed things down by a factor of 1000. Simulating all the gates and wires in the hydra HDL would be far worse. anthony
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.