Author: Eugene Nalimov
Date: 14:37:03 02/25/00
Go up one level in this thread
On February 25, 2000 at 16:48:46, Roy Brunjes wrote: >My apologies for asking this if it was previously discussed ... > >As computers head toward the 100 TB disk storage range, and not too far away are >systems with 1 PB (Peta Byte) of storage online (1,000 TB = 1 PB), how long >before we will see really large tablebases being built? [8, 9, 10 pieces] > >And what is the formula used to compute the size required to store all n piece >tablebases? Of course, if it were possible (and I doubt it is for the >forseeable future), how large would a 32 piece tablebase set be? At that point, >chess is no longer a mystery for the machine of course. > >Anyone know the formula [obviously an exponential, but what is the formula]? I >suspect Bob Hyatt knows -- Bob are you reading this? > >Thanks, > >Roy Each (N+1)-man TB is roughly 60 times larger than N-man TB. Plus, there are more (N+1)-m different TBs than there are N-man TBs - you can find exact formula in archieves. Assuming that average 5-man TB is ~100Mb, average 7-man TB would be 360Gb after compression. And there would be *a lot of* them - I suspect that 100Tb is not enough. Please notice that you need a lot of RAM to produce TB, too - ~1/8 of TB size (before compression) using the best algorithm I know of (or you'll have to do a lot of disk I/O). So to produce the average 7 man TB you'll have to use a machine with ~150Gb of RAM. And I suspect that machine would have to run for several months to produce one TB out of hundreds. For now practical limit is 6-man TBs, with 7-man TBs *barely reachable* in several years. IMHO, of course. Eugene
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.