[ydl-gen] Teraflop computing with the PS3 - correction

[Derick Centeno]

On Sun, 07 Oct 2007 23:56:11 -0400
"Thomas A. McGonagle" <tom at dataero.com> wrote:
Hi Thomas:

Essentially our collective means of evaluating or determining
processing power of CPUs needs to become more precise as we abandon
single core for multicore systems.  How that is defined or if such a
determination can be made at all depends on what processing is actually
happening versus what we believe to be happening; this is not as
straightforward as it was with single core systems.  

Here's a nice table to flesh the problem out more clearly.  The first
is the early reported output for the Folding at Home project - a
reasonably intensive test for any system.  It takes the PS3 8 hours to
process one work unit for the project, however note the early
(March 2007) results here:

http://www.imminst.org/forum/index.php?act=ST&f=217&t=15169&s=

However if we consider what Stanford is reporting currently (Oct 2007)
regarding their results the PS3 is approaching 1000 Teraflops.  The
unprecedented availability of such processing power is such that the
project is hoping that if 50,000 PS3 participate the processing power
involved can approach Petaflop scales.  Currently competing systems or
rather other participating computers are not even close to approaching
PS3 processing rates.  Here's the current figures:

http://fah-web.stanford.edu/cgi-bin/main.py?qtype=osstats

We are in a really fascinating time in the sense that the capacity of
the variety of multicore systems are so new that there are no
measurements which are commonly familiar that are meaningful or
really well understood, yet. In the meantime, the numbers Stanford
reports are stunning in that one can clearly see the force of all the
various systems pitching in an doing their bit.  

My perspective is that will all the enormous presence of Intel and
other systems out there I would have thought that the scale of
processing would have been quite different.  However, Stanford's
figures are compelling in that the admittedly tiny -- miniscule --
amount of PS3's contributing to the project are doing an enormous
amount of work.  Of course, programming the PS3 to do something as
intense for one's own interests is not a trivial weekend pursuit ...
however the potential that is there to be used for personal research
and learning and professional level contributions is unprecedented.

> Hello All,
>  Earlier this year Wired Magazine published a one pager stating the
> computational power of various computers. If I remember correctly (I
> can't find the 1 pager on wired.com), they said that:
>    the PS3 could operate at 1.5 Teraflops
>    the XBox360 could operate at 1 Teraflops
>    430 Pentium 4 computers could operate at 1 Teraflop.
> 
>  Since reading the "article" I have repeated this to anyone who would
> listen.
> 
>  Upon visiting the Terrasoft website, I read that 1TFlop can
> theoretically be reached by an 8 PS3 cluster.
> 
>  Would it be correct to say that one PS3 is .125(1T / 8 PS3s)  TFlops?
> 
>  Also does anyone know what was wrong with that "article", or does
> anyone have it handy?
> 
>  Thanks a lot.
> -Tom


===============

 "If I were not a physicist, I would probably be a musician. I often
think in music. I live my daydreams in music. I see my life in terms of
music. ... I get most joy in life out of music."  

"What Life Means to Einstein: An Interview by George Sylvester
Viereck," for the October 26, 1929 issue of The Saturday Evening Post.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.terrasoftsolutions.com/pipermail/yellowdog-general/attachments/20071008/cfa7f7fb/attachment.bin