Pentium IV Xeon memory bandwidth. Any experience?

Thomas Guignon guignon at
Mon Jun 25 05:19:15 PDT 2001

Le Lundi 25 Juin 2001 11:59, Jean-Marie Teuler a écrit :
> Dear Beowulfers,
> We are studying various architectures for dong efficient Fast Fourier
> Transforms (FFTs). As our algorithms can be easily split across many
> processors, we will build a Beowulf system with as many nodes as needed to
> reach our performace target.
> So the only question left is deciding which is the best suited (i.e. with
> the best performane/price ration) node for this type of calculations.
> One of the problems of FFTs is that their computational efficiency
> decreases dramatically as the length of the transform increases, due to
> "out of cache" effects: it is a "momory-bounded" problem.
> This is why we have been very interested by the announcement of the Pentium
> IV Xeon processor, given that it is credited of a 800 MHz memory bandwidth.
> Have some of you already have had experience with these processors (on
> mono- or bi-processors boards) that would confirm this very attractive
> figure?
> Jean-Marie Teuler

Hello Jean-Marie,
Before spending money in expensive processors and expensive motheboards
Try to have a look at Athlon processors with DDR memory
We have tested and 1.2 Ghz with PC2100 DDR with Level 1 Blas and results are
quite nice:
-dnrm2: (one vector read)
1450 10^6 B/s
-ddot: (two vector read)
1040 10^6 B/s
-daxpy (one vector read and one vector read/write)
1150 10^6 B/s
- copy  (one vector read and one vector write)
990 10^6 B/s

To obtain this kind of performance you have to use
the Athlon hardware prefetch.

Guignon Thomas, guignon at
Tel: 01 69 18 95 95
Fax: 01 69 18 95 96
69, rue de Paris
91400 Orsay

More information about the Beowulf mailing list