Author: Dann Corbit
Date: 15:30:12 01/27/04
Go up one level in this thread
On January 27, 2004 at 17:06:19, Slater Wold wrote: >On January 27, 2004 at 15:37:58, Dann Corbit wrote: > >>On January 27, 2004 at 15:18:31, Jorge Pichard wrote: >> >>>http://asia.cnet.com/newstech/systems/0,39001153,39154800,00.htm >>>http://www.dvhardware.net/article.php?sid=2016 >>>http://www.hoise.com/primeur/03/articles/monthly/AE-PR-12-03-85.html >>>http://news.zdnet.co.uk/hardware/chips/0,39020354,39117150,00.htm >>>http://www.gridtoday.com/03/1020/102148.html >>>http://www.idg.com.sg/idgwww.nsf/0/52F6C8B4A87C437A48256DC000351F9D?OpenDocument >> >>"New Scientist reports that a single chip will cost around US$16,500" >>25 GFlops can be had cheaper. >> >>I doubt if it can do chess calculations faster than an 8-chip Brutus system. > >I don't. The chip is rated at 25 Gflops. It will take a lot of instructions to perform a single chess move. 25 GFlops sounds like a lot, but this document: http://www.intel.com/software/products/mkl/techtopics/wp_mkl_benchmark.pdf shows that a single 3.4 GHz 32 bit Intel chip can get 4 Gflops performing DGEMM. So it is only 6 times faster than a standard CPU. 16500/6 = $2750 per chip for break even. Price performance is a total yawner. >>And probably still costs a lot more. > >8 FPGA cards are a lot more than $16,500. Not including the PCs they have to go >into. I don't know what they cost. But a fairly strong PC can be had for $1000, so that leaves $1062.50 per card.
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.