From: Charles Grassl (
Date: Wed Jul 26 1995 - 14:58:09 CDT


For inclusion in your STREAM tables, below are T3D results for 512 PEs.

As with the previous T3D listings, there are two sets of results: one
with assembly language coded SUM and TRIAD and one with straight
Fortran versions.

The TRIAD which I used is not SAXPY from libsci. The used TRIAD does
the correct handling of C <- A + s*B.

 *** STREAM benchmark ***
 Number of PEs: 512
 Number of iterations: 2
 Size of Arrays: 501 Kwords
 Offset: 2048 words
                              Bandwidth (Mbyte/s)
 Machine Copy Scale Sum Triad
 ----------------- -------- -------- -------- --------
 CRAY T3D 512 PEs 161479.7 168193.5 91775.6 95304.2

 #CRAY T3D 512 PEs 169677.7 166578.1 114976.8 112126.4

* Use assembly language SUM and TRIAD

I believe that the 512 PE CRAY T3D has the highest bandwidth of any
computer listed in the STREAM table.

We are getting excited about our next MPP system, the CRAY T3E. It
should have more than three times the bandwidth per PE at approximately
the same price per PE.

Charles Grassl
Cray Research, Inc.

This archive was generated by hypermail 2b29 : Tue Apr 18 2000 - 05:23:05 CDT