X-Message-Number: 26914
From: 
Date: Fri, 2 Sep 2005 10:00:35 EDT
Subject: Re: CryoNet #26904 Sequential and parallel processors

From: Daniel Crevier _ (mailto:) :

I  believe the discussion between Yvan Bozzonetti and Thomas Donaldson  would
be clarified if they took into account the notion of cycle time. A  parallel
computer cannot react faster than the cycle time of its  processors. Call
this cycle time tau. If there are N processors, then it  should be possible
to simulate them in real time with a single processor  having a shorter cycle
time. The cycle time of this single processor simply  has to be short enough
that it can figure out, in time  tau/N, what  one of the parallel processors
will do in time tau.


This is perfectly true. It is why I said that parallel processors are  faster 
than single one * at equal technology level*, that is, at equal clock  speed. 
If the sequential, single processor run far faster than the parallel  device 
it can compete with it or even overrun it. At most, the biological system  is 
at one kHz clock speed, the electronics microprocessor is a 3 - 4 GHz. The  
last Intel biprocessor at 3.2 GHz is not as fast as the 3.8 GHz Intel  P4...
 
Yvan Bozzonetti.


 Content-Type: text/html; charset="US-ASCII"

[ AUTOMATICALLY SKIPPING HTML ENCODING! ] 

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=26914