Much better T932 stream results

From: Jeff Brooks (jpb@cray.com)
Date: Thu Nov 21 1996 - 16:44:10 CST


Hi John-

You can put these numbers up officially. We found the problem.
The 1 CPU number is just a bit slower due to the longer latency
of the big chassis.

-jeff

>
>
> T932 Stream benchmark results
>
>
> Stream benchmark performance for the T932/321024-3E. This system
> has IEEE cpus and CM03 memory modules.
>
> cpus benchmark test results, Mbytes/s
> assignment scaling summing saxpying
> 1 10,653 10,221 13,014 13,682
> 2 20,326 19,622 25,459 26,117
> 4 40,457 39,017 50,203 50,718
> 8 80,929 78,131 98,701 99,343
> 16 160,263 154,880 193,335 194,562
> 24 237,415 230,100 281,038 282,300
> 32 310,721 302,182 359,841 359,270
>
>
> cpus benchmark test results, 64 bit words per clock
> assignment scaling summing saxpying
> 1 3.0 2.9 3.7 3.9
> 2 5.8 5.6 7.2 7.4
> 4 11.5 11.1 14.3 14.4
> 8 23.0 22.2 28.0 28.2
> 16 45.5 44.0 54.9 55.3
> 24 67.4 65.4 79.8 80.2
> 32 88.3 85.8 102.2 102.1
>
>
>
>
>
>
>
>
>
>
>
>
>
> **** Cray Research, Silicon Graphics Company -- 1, Sandness,21-Nov-96
>
>
>

-- 
================================================================================
Jeff Brooks                                           Email: jpb@cray.com
Benchmarking Group                                    Phone: 612-683-3410 
Cray Research, Inc.                                     Fax: 612-683-5599
655F Lone Oak Drive
Eagan, MN  55121
================================================================================



This archive was generated by hypermail 2b29 : Tue Apr 18 2000 - 05:23:06 CDT