Author: Robert Hyatt
Date: 19:44:24 02/15/03
Go up one level in this thread
On February 15, 2003 at 21:52:40, Tom Kerrigan wrote: >On February 13, 2003 at 18:12:05, Robert Hyatt wrote: > >>If you had ever followed discussions on Cray Blitz, you would know that we did a >>lot of >>things in it that I can't do in crafty due to the cost. For example, >>"qualitative mobility". >>Not just the number of squares a piece can move to, but the number of squares it >>can move >>to safely _and_ how useful the square is (ie e4 is much better than a8). Or > >Bah. This discussion is worthless. You say that Cray processors are blazingly >fast and your only evidence is that some arbitrary program runs at an arbitrary >speed. > >The 1MHz Z80 is faster than a Cray because it runs lightcycles at 10 kilopixels >per second. Can't argue with that. > >-Tom The discussion is only worthless because that is what you want. I gave the reasons for why a cray is so much faster than any PC ever made. Once again: it can read four words and write two words, per clock cycle, per CPU. And since it is a fully SMP machine, not NUMA, the aggregate bandwidth actually means something. But for a single CPU, at 500mhz, that turns into 500 million cycles per second, times 48 bytes per cycle, which turns into a single-cpu bandwidth of 24 _gigabytes_ per second. What PC are you going to propose to come within a factor of 10 of that number? And then there are 32 cpus so that turns into 768 gigabytes of bandwidth per second. What PC approaches that? And finally, it can do at _least_ eight floating point operations per cycle, which is 4 gigaflops per processor. PC that delivers that??? So yes, the Cray _is_ fast, for those interested in how it works and in real number-crunching ability. BTW those numbers can be found all over the net, it isn't a secret. There are supercomputers that are faster. I just have not ever run a chess program on them enough to get good benchmark numbers. The SX machines, for example... Supercomputers are a different animal from the microcomputer, even though some lists (top-500 for example) includes huge clusters. Several of the top 500 are simply a roomfull of PCs that could not touch a single cpu Cray for serious floating point calculations...
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.