Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!swrinde!emory!hubcap!kbreinho%ug
From: kbreinho%ug@cs.utah.edu (Keith Breinholt)
Newsgroups: comp.parallel
Subject: Re: scalability of n-cubes, meshes (was: IPSC Communications)
Summary: At some point of scale any mesh will cook.
Message-ID: <7384@hubcap.clemson.edu>
Date: 8 Dec 89 19:39:20 GMT
Sender: fpst@hubcap.clemson.edu
Lines: 21
Approved: parallel@hubcap.clemson.edu

Before I get anymore fan mail (read "flame mail") let me post an apology to
Gene Itkish and all of the others who sought to enlighten me on 3d meshes.

In my posting I was assuming a cube connected 3d mesh.  Thus the reference
to 6 sides which can be represented in a 2d hexagonal mesh. However, as 
Gene pointed out, this is not a 3d mesh (for you purists out there) although
I have the same signal paths and possibly signal lengths.

And as more signal paths are added in a 3d mesh, it becomes impractical to
mimic them in 2d.

The fact remains that at some huge scale 3d meshes cannot be properly cooled.
(And I think we need to add the words, "without some special method of
cooling.")  I say this because even 2d or 1d curcuits at some point of scale
will "cook" themselves.

Now, did I leave something out?

Keith L. Breinholt             kbreinho@ug.utah.edu or usiklb@vm.usi.utah.edu
Utah Supercomputing Institute
(801) 581-4439