Author: Dieter Buerssner
Date: 14:21:54 04/20/05
Go up one level in this thread
One data point from my memory. I bought a 386SX 116 MHz PC 1989 or 1990 (I remember rather sure, that the BIOS said (C) 1989). An ancestor of Yace could do 1000 nodes/s on that computer (when the position was rather easy). Since then, I did not work on the engine for many years; I speeded up many things, and made other things slower (for example more complicated search and eval, much bigger memory footprint, no assembler anymore, ...). The speed should be quite comparable to the current engine. At the end of 2003, I bought a computer with P4, 2.53 GHz. It could do 1e6 nodes/s in the positions where it showed fastest speed (exluding some not very interesting cases, where practically no eval is needed because of forced mate). So, a factor of 2^10 in 13 years. Just recently I bought a notebook with 1,7 GHz Centrino. It was more expensive (of course, because it is a notebook ...). On average, it may have a factor of 1500 compared to that old 386SX. 1986, I bought an Atari ST (Motorola 68000). It was not much slower, than the 386SX. So, doubling of speed was clearly not the case in that time - but faster than doubling every 2 years. My impression is, that the gradient of the curve speed vs. time flatted down in the last few years. If I would buy an Intel P4 now for a similar prize, that I paid 2003 for the P4 2.53 GHz, I would hardly get the double speed. Disk size and memory size would have doubled. Regards, Dieter
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.