Discussion:
Cell Processor performance - upto 100x Intel Xeon in certain operations
(too old to reply)
2005-06-13 17:17:02 UTC
Permalink
These are a lot of slides, but this is a normalized comparision between an
Intel Xeon processor (pumped up Pentium 4) to Cell in what's known as a
Fourier Fast Transform. At a 256 point FFT each SPE is slightly more
powerful than an entire Xeon making Cell as a whole greater than 7x faster
than the Intel chip. Under the circumstances of the 16 million point FFT,
there's no competition. 100x faster. Xeon ran out of bandwidth and
computational resources. Cell kept on going.


Loading Image...
Loading Image...
Loading Image...
Loading Image...
Loading Image...
Loading Image...
Loading Image...
Loading Image...
Loading Image...
Loading Image...
Maynard Handley
2005-06-13 21:48:50 UTC
Permalink
In article <krmdnRsF8vsaIjDfRVn-***@comcast.com>, <Rage6c> wrote:

NEWSFLASH: Chip optimized for task X performs better at task X than chip
optimized for task Y.

Idiot.
Post by
These are a lot of slides, but this is a normalized comparision between an
Intel Xeon processor (pumped up Pentium 4) to Cell in what's known as a
Fourier Fast Transform. At a 256 point FFT each SPE is slightly more
powerful than an entire Xeon making Cell as a whole greater than 7x faster
than the Intel chip. Under the circumstances of the 16 million point FFT,
there's no competition. 100x faster. Xeon ran out of bandwidth and
computational resources. Cell kept on going.
http://i21.photobucket.com/albums/b275/tacitblue73/BarcelonaCellProgrammingsli
des.gif
http://i21.photobucket.com/albums/b275/tacitblue73/BarcelonaCellProgrammingsli
des1.gif
http://i21.photobucket.com/albums/b275/tacitblue73/BarcelonaCellProgrammingsli
des2.gif
http://i21.photobucket.com/albums/b275/tacitblue73/BarcelonaCellProgrammingsli
des3.gif
http://i21.photobucket.com/albums/b275/tacitblue73/BarcelonaCellProgrammingsli
des4.gif
http://i21.photobucket.com/albums/b275/tacitblue73/BarcelonaCellProgrammingsli
des5.gif
http://i21.photobucket.com/albums/b275/tacitblue73/BarcelonaCellProgrammingsli
des6.gif
http://i21.photobucket.com/albums/b275/tacitblue73/BarcelonaCellProgrammingsli
des7.gif
http://i21.photobucket.com/albums/b275/tacitblue73/BarcelonaCellProgrammingsli
des8.gif
http://i21.photobucket.com/albums/b275/tacitblue73/BarcelonaCellProgrammingsli
des9.gif
Klaus Fehrle
2005-06-13 23:27:28 UTC
Permalink
Post by Maynard Handley
NEWSFLASH: Chip optimized for task X performs better at task X than chip
optimized for task Y.
Idiot.
While your _Newsflash_-statement stands undisputed, there is no need to
go ad hominem. At least not as long TOP500 is published on a
Linpack-score. :-)

KF
Yousuf Khan
2005-06-15 01:36:27 UTC
Permalink
Post by Klaus Fehrle
While your _Newsflash_-statement stands undisputed, there is no need to
go ad hominem. At least not as long TOP500 is published on a
Linpack-score. :-)
Hmm, I wonder if Cell could keep up against an Opteron running TPC-C?

Yousuf Khan

Loading...