Take any two: motherboard performance, compatibility, value

[Lechner, David]

I would like to especially thank Paralogic Corp. (Doug and Adam) for making
this posting - VERY interesting indeed - 

Do I have it right in that the STREAM functions used in the benchmark are
really better suited for a single processor, and thus when the OS tries to
allocate and divide the work then extra work has to be done coordinating the
work (looks like a 10% hit beyond linear division of the bandwidth by 2) ? 

Do I interpret the timing results to see that the single processor function
is much faster than the dual processor functionfor the same work?
Is this because of the nature of the data set (such as very large matrix
manipulation) that the second processor is not only unable to help at all
but actually makes things twice as worse by getting involved?

Or is it because 2 versions are running in parallel, getting into trouble,
and then taking a little longer than 2x as long to do 2x as much work?

(Adam/Doug - Were you using an auto-parallelizing compiler for the dual
processor version? Or the hack suggested to run 2 jobs at once?)

W/Regards/ & Thanks In Advance/ 
 Dave Lechner


-----Original Message-----
From: Adam Lazur [mailto:alazur at plogic.com]
Sent: Friday, June 30, 2000 10:36 AM
To: Beowulf mailing list
Cc: Josip Loncaric; Douglas Eadline
Subject: Re: Take any two: motherboard performance, compatibility, value


Doug Eadline (deadline at plogic.com) said:
> We have Linux working on Serverworks systems. We are still testing, but
> our dual PIII Cu-733 with 1GB PC-133 SDRAM(ecc) gave pretty good stream
> numbers.  

For the benefit of the list, here are the stream numbers from a dual PIII
Cu-733 w/PC-133 SDRAM(ecc):

ServerWorks ServerSet III LE chipset board

[uniproc stream_wall test]
Function      Rate (MB/s)   RMS time     Min time     Max time
Copy:         376.7285       0.0935       0.0934       0.0940
Scale:        373.5860       0.0942       0.0942       0.0943
Add:          483.6362       0.1092       0.1092       0.1092
Triad:        341.0235       0.1548       0.1548       0.1549

[dualproc stream_wall test]
Function      Rate (MB/s)   RMS time     Min time     Max time
Copy:         161.1795       0.2216       0.2184       0.2244
Scale:        173.3435       0.2090       0.2031       0.2118
Add:          191.2891       0.2805       0.2760       0.2839
Triad:        185.7363       0.2953       0.2843       0.3018

Tyan S1834 Tiger 133 - VIA Apollo Pro 133A Chipset board

[uniproc stream_wall test]
Function      Rate (MB/s)   RMS time     Min time     Max time
Copy:         313.4796       0.1124       0.1123       0.1131
Scale:        312.1564       0.1128       0.1128       0.1128
Add:          411.2952       0.1284       0.1284       0.1287
Triad:        312.8129       0.1688       0.1688       0.1689

[dualproc stream_wall test]
Function      Rate (MB/s)   RMS time     Min time     Max time
Copy:         146.1533       0.2432       0.2408       0.2469
Scale:        147.7545       0.2402       0.2382       0.2420
Add:          165.5157       0.3218       0.3190       0.3237
Triad:        158.7726       0.3360       0.3326       0.3385


The results from the ServerWorks chipset are quite interesting. I also
experienced issues with not being able to set ECC (the menu option is
*missing*) on the Tyan S1834 mobo.

.adam

-- 
[                  Adam Lazur <alazur at plogic.com>                     ]
[      Paralogic Inc. - www.plogic.com - www.xtreme-machines.com      ]

_______________________________________________
Beowulf mailing list
Beowulf at beowulf.org
http://www.beowulf.org/mailman/listinfo/beowulf