Author: vladan
Date: 06:02:24 10/27/05
Go up one level in this thread
On October 27, 2005 at 06:34:50, Dagh Nielsen wrote: >Dear Mr. Vuckovic. > >Thank you very much for a very thorough explanation! I am impressed with the >logic of Axon and its intent. > >I am not so impressed with the general increase of hardware speed, though! Is it >just me, or do the chip manufacturors seem to have reached the wall in processor >speed advances? > >I know that they are now focusing more on going multicore and embedding >sophisticated features in the chips, but that really don't help our chess >engines that much, does it? :-) I mean, sadly two processors don't mean twice as >fast chess algorithm calculations, due to the looped nature where constantly >renewed feedback values are needed for maximum effectivity. > >I wonder what we can expect in the future regarding hardware advances for chess >engines. > >All the best, >Dagh Nielsen Dear Mr. Nielsen, Yes, you are right, we are approaching the technology limits. There are many researching efforts in direction of changing basic materials or principles (quant, optical, bio technology) to increase the speed of information processing. But, the fastest technology could not overcome light speed. Geometry of processor is also finite, so there will be problems for sure. The problems in thermal overheat and nanomener capacity are also very important. The other way is to change the processor/machine architecture. In this area situation is very unclear – the problems of distribution and synchronization in parallel machines became dominant. Currently, as the part of my PhD theses, I work on 16 processor LAN connected chess machine. Each node encapsulates Axon v2.0 MMX 32-bit machine coded chess engine. The engine is very fast null-mover, using about 20 modern pruning techniques with 256 Mb hash memory. Preliminary tests are very solid, today I overcome 9 million positions per second with 14 processors (I have problems with 2 machines). I except to play first complete game with the machine till the end of November. It is obvious that new parallel techniques will be mainstream in compute chess, in the very next future. best wishes, Vladan
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.