Great initiative!

Article: A Look at SPECpower
By: David Kanter (dkanter.delete@this.realworldtech.com), May 13, 2008 11:09 pm
Room: Moderated Discussions
Henrik S (hstahl@nospam.bea.nothanks.com) on 5/13/08 wrote:
---------------------------
>Some comments:
>
>> Sun had some of the most power optimized products,
>> courtesy of Niagara"
>
>I think it's worth pointing out that Sun has made no submissions on this benchmark
>yet, in particular not on Niagara hardware. Now, that may be because their JVM has
>been lagging in performance on SPECjbb (and by extension, on SPECpower) but now
>that they seem to have fixed that I hope they will remedy this quickly. Let's see
>them back up those "performance/watt" claims with hard >numbers.

When it was released, Niagara definitely had an edge. I don't know how well it would fare today.

>> SPECpower is probably the most complex benchmark that
>> SPEC has undertaken to date
>
>Not sure what you mean here... SPECjAppServer2004 and >SPECweb2005 are both much
>more complex to set up, run and optimize for.

I mean that the benchmark itself is very complex. Simply measuring performance is non-trivial, but measuring performance and power is much much more complicated.

Small things like the # of fans in a system suddenly make a much larger difference.

>> It is to AMD's credit that they were willing to
>> participate, given the degree to which their performance > lags behind Intel (roughly a factor of 3X, 698 vs. 203).
>
>Also kudos to Intel that they published a result on an old Xeon (Netburst, Irwindale):
>
>http://www.spec.org/power_ssj2008/results/res2007q4/power_ssj2008-20071129-00015.html
>
>
>which of course has a really poor score compared to AMD's submission as well as newer Xeons.
>
>> AMD's sole submission uses dual-core Opterons, rather
>> than the newer Barcelona - it would be nice to see tests > for AMD's newer server processors.
>
>Yes, and for other hardware. Right now all up to date publications are with Intel Xeon:
>
>http://www.spec.org/power_ssj2008/results/power_ssj2008.html
>
>...which makes for a boring read. We need submissions on other configurations on
>this benchmark to make it really interesting. SPECjbb2005 has been more successful
>in that to date (in large part because of competition >between the JVM vendors!).

I agree the results look a little empty, nothing from HP, IBM, Sun or Fujitsu with their proprietary systems.

DK
< Previous Post in ThreadNext Post in Thread >
TopicPosted ByDate
A Look at SPECpowerDavid Kanter2008/05/12 09:41 AM
  Great initiative!Henrik S2008/05/13 02:49 AM
    Great initiative!Wes Felter2008/05/13 09:42 AM
    Great initiative!David Kanter2008/05/13 11:09 PM
      Great initiative!Henrik S2008/05/14 02:32 AM
        Great initiative!Wes Felter2008/05/14 09:17 AM
          Great initiative!Henrik S2008/05/14 10:00 AM
Reply to this Topic
Name:
Email:
Topic:
Body: No Text
How do you spell purple?