Re: Submission of Standard Stream Result AMD Istanbul 8435 ( 2.6GHz )

Dear Dr.McCalpin



Thanks again, and best wishes.


Mark Han


PS:
Below, for reference, is the result of probe disabled Istanbul, I =
cleaned it up a bit trying not to create too much trouble. :-)

System Description / Configuration:
Motherboard: Supermicro BHQME ( blade )

CPU: AMD Opteron Istanbul 8435

CPU Speed: 2.6GHz

CPU(s): 4 processors, 6 cores/processor

L3 Cache: 4 x 6MB ( per processor )

Memory: 64GB DDR2-800MHz ( 4GB per DIMM, 4 socket x 4 DIMMs x 4GB , =
2DPC, Dual Channels).



HT-Assist: disabled ( probe filter disabled, by BIOS )



Operation system: Windows 2008 64bit R2


Result: ( Probe filter disabled )

.
.
.
.
C:\Stream_64bit>REM : 4-Ways Ps - 24 cores : =
0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23
 
C:\Stream_64bit>REM 2.6 GHz x 4 Opteron 8435 / 4 x 4 x 4GB ( 64GB ) =
DDR2-800MHz
 
C:\Stream_64bit>set MP_BIND=yes
 
C:\Stream_64bit>rem Multithreaded version Stream5.8_omp-64.exe Windows =
2008 64bit R2
 
C:\Stream_64bit>ECHO ON

C:\Stream_64bit>REM - HT Assist / Probe filters Disabled ( BIOS )

C:\Stream_64bit>set OMP_NUM_THREADS=24

C:\Stream_64bit>set =
MP_BLIST=0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23 =


C:\Stream_64bit>start /b /WAIT /HIGH stream5.8_omp-64.exe
-------------------------------------------------------------
STREAM version $Revision: 5.8 $
-------------------------------------------------------------
This system uses 8 bytes per DOUBLE PRECISION word.
-------------------------------------------------------------
Array size = 40000000, Offset = 0
Total memory required = 915.5 MB.
Each test is run 40 times, but only
the *best* time for each is used.
-------------------------------------------------------------
Number of Threads requested = 24
-------------------------------------------------------------
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
Printing one line per active thread....
-------------------------------------------------------------
Your clock granularity/precision appears to be 1 microseconds.
Each test below will take on the order of 296310 microseconds.
   (= 296310 clock ticks)
Increase the size of the arrays if this shows that
you are not getting at least 20 clock ticks per test.
-------------------------------------------------------------
WARNING -- The above is only a rough guideline.
For best results, please be sure you know the
precision of your system timer.
-------------------------------------------------------------
Function Rate (MB/s) Avg time Min time Max time
Copy: 28058.2750 0.0244 0.0228 0.0433
Scale: 28026.9815 0.0246 0.0228 0.0390
Add: 28072.9026 0.0356 0.0342 0.0510
Triad: 28095.4364 0.0374 0.0342 0.0516
-------------------------------------------------------------
Solution Validates
-------------------------------------------------------------



  ----- Original Message -----
  From: John McCalpin
  To: Mark_Digicor
  Sent: Wednesday, October 07, 2009 11:39 PM
  Subject: Re: Submission of Standard Stream Result AMD Istanbul 8435 ( =
2.6GHz )


  Mark_Digicor wrote:
>
> Dear Dr. McCalpin,
>
> I would like to submit the standard STREAM result of a Supermicro =
AMD
> Istanbul 4-Way system.
>
>
> Below is the system description and STREAM test Result.
>
> Hope to see it being posted soon.
>
>
>
> Best Regards.
>
>
>
> Mark Han
>

  The results are up on the STREAM web site now. I had a little =
trouble
  at first because I accidentally reversed two fields -- my scripts did
  not know what to do with an 8-thread result using 24-Byte precision! =
:-)

  This is a very nice result for a Windows system running an older
  binary. The memory controller improvements that they made in =
Shanghai
  definitely seem to be visible here once the probe filters are enabled.

  We have a small set of two-socket Shanghai and Istanbul boxes at the
  office -- I should get those results published too. (Our big machine
  has 3,936 4-socket Barcelona nodes, but those STREAM results are not
  very exciting -- except in aggregate where it comes out to over 65 =
TB/s.)

  Thanks for the results!

  john

  --
  John D. McCalpin, Ph.D. "Dr. Bandwidth"
  john@mccalpin.com http://www.streambench.org/
          http://www.cs.virginia.edu/stream/




Received on Wed Oct 07 21:13:46 2009