RE: Results for Unisys Corporation ES7000 Orion 130 (8 p rocessors)

From: Nowatzki, Thomas L
Date: Fri May 09 2003 - 09:59:32 CDT

  • Next message: Frank Johnston: "STREAM results on a 1.5GHz 16-CPU IBM eServer pSeries p670"
  • Next message: Frank Johnston: "(no subject)"

    Hi John,

    The style that I listed in my submission (Style ES7005263-GS) is for
    an 8X configured with 2 processors per bus on a total of 4 busses.
    The ES7000 is orderable with this style and it is what we expect to
    market to our 8X HPC customers.

    I understand your concerns, that is why I verified and documented the
    style number in my submission.

    Please let me know what you think.

    Regards, Tom.
     

    -----Original Message-----
    From: John D McCalpin [mailto:jmccalpin@austin.rr.com]
    Sent: Friday, May 09, 2003 8:33 AM
    To: Nowatzki, Thomas L
    Subject: Re: Results for Unisys Corporation ES7000 Orion 130 (8
    processors)

    On Wednesday 07 May 2003 05:48 am, you wrote:
    > These STREAM results have been approved for publication.
    >
    > They were obtained on a Unisys Corporation ES7000 Orion 130 Server
    > running SCO Linux Server 4.0 for Itanium Processor Family. The system was
    > configured with 64 GB RAM, consisting of four 16 GB memory features. The
    > ES7000 Orion 130 Server contained eight (8) 900 MHz Itanium 2 processors
    > each with 1.5 MB L3 cache. These are submitted as "standard" results.
    > The arrays were sized to each be at least 4x the size of the L3 cache in
    > each system. In this case we set N=12,000,000.

    Hi Thomas,

    Thanks for the submission.

    These look like "partially depopulated" results, which I do not usually
    publish in the "standard" tables.

    I could be misinterpreting this, of course, but it looks like you have
    configured 2p on each of 4 busses. If this is the standard configuration
    for your 8-way system, then this is a "standard" result. If (as seems
    more likely) the standard configuration for an 8-way system is 4p on
    each of two busses, then this is a "nonstandard" result.

    For historical reference: once upon a time, HP created a separate
    model number for a similar "partially populated" system. I was then
    able to publish the results on the "standard" table under the model
    name of the "partially populated" option. Look for V2600_Enterprise
    vs V2600_Technical in the STREAM results.



    This archive was generated by hypermail 2.1.4 : Thu May 15 2003 - 10:41:18 CDT