Computer Chess Club Archives




Subject: Re: Introducing "No-Moore's Law"

Author: Robert Hyatt

Date: 17:00:21 03/01/03

Go up one level in this thread

On March 01, 2003 at 01:57:32, Jeremiah Penery wrote:

>On March 01, 2003 at 00:20:22, Robert Hyatt wrote:
>>>There are plenty of other options for high-performance computing that don't
>>>exist today.  I wouldn't be surprised to see asynchronous chips being seriously
>>>considered soon.  The clock-based approach is beginning to cause lots of
>>>problems as speed is agressively increased.
>>IE data-flow?  (Circa 1970 or so)??  :)
>Yes, they've been considered in the past, but always rejected because the
>clock-based approach was entrenched in the market, and continued to perform and
>scale well.  I've read that something like 50% of the power draw of current CPUs
>is related to the clock synchronization effort.  Not to mention other problems
>caused by the necessity of clock synchronization...
>I think there's a good chance that an asynchronous approach will be tried again
>in the near future.

Actually, to the best of my knowledge, no "asynchronous" data flow machine has
ever been built.  TI built some dataflow chips years ago, but they were really
bizarre interpretations that were very similar to the P6 reorder buffer concept
but using a ring buffer instead.

But the idea was interesting, and the programming ideas were unique, as in
"single-assignment languages".

This page took 0.01 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.