From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!think.com!samsung!uunet!idtg!dow Mon Dec  9 10:47:35 EST 1991
Article 1816 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!think.com!samsung!uunet!idtg!dow
>From: dow@idtg.UUCP (Keith Dow)
Newsgroups: comp.ai.philosophy
Subject: Re: Physical limits when programming neurons and minds
Message-ID: <341@idtg.UUCP>
Date: 2 Dec 91 22:57:44 GMT
References: <57850@netnews.upenn.edu> <445@trwacs.UUCP> <44011@mimsy.umd.edu>
Organization: Integrated Device Technology, Santa Clara
Lines: 23

stuff deleted

>	I recall that 17 years ago, or therabouts (to show how out-of-date
>I am) - it took hundreds of hours of computing by the current state-of-
>the-art approximate calculation on a CDC Cyber (now extinct species with
>long words) for solution of Schroedinger's equation for even the simplest 
>stripped atoms. This was the first "successful" calculation, and nobody 
>was sure whether it was correct.
>	Times have surely changed if "drug companies make millions"
>out of this sort of thing now. (Do their computers suffer side-effects of
>these calculations?)




Seventeen years ago, in the world of computers, is two years after
they invented dirt. :-)  Molecular modeling on workstations runs at about
one third the speed it does on the fastest supercomputers.  The reason
is the problem is tightly coupled and cannot be vectorized.
There is a good article on it in the EE Times of a few weeks back.

Workstations speed up about a factor of two every year.  So you will
see a lot more power coming up.  


