Phil (phil@phil.com) on 2/6/09 wrote:
---------------------------
>someone (someone@somewhere.com) on 2/6/09 wrote:
>---------------------------
>>Phil (phdny@mac.com) on 2/6/09 wrote:
>>---------------------------
>>>David Kanter (dkanter@realworldtech.com) on 2/5/09 wrote:
>>>---------------------------
>>>>Paul (no@thanks.com) on 2/5/09 wrote:
>>>>---------------------------
>>>>>Temp (not@this.time) on 2/5/09 wrote:
>>>>>---------------------------
>>>>>>"Itanium sales are roughly comparable to Xeon MP volumes." could mean so many things.
>>>>>>One could be comparing CPU unit sales, CPU dollars sales (before or after rebates),
>>>>>>or even server sales based on those CPU product lines. Roughly comparable could
>>>>>>means anything from being equal, to being on the same order of magnitude - not
>>>>differing more than by a factor of 10.
>>>>>>
>>>>>>David could surely comment and clarify.
>>>>>
>>>>>Or it could mean that the volumes are the same, Xeon MP is much lower volume than
>>>>>DP (or whatever they call dual socket capable Xeons these >days).
>>>>
>>>>I was being purposefully vague with that statement, but I really meant CPU shipments.
>>>>
>>>>My sense is that IPF and Xeon MP shipments are within a factor of 2-3.
>>>>
>>>>I believe that Xeon MP is around 2M/year, IPF is probably close to 1M.
>>>
>>>David-Get an account with IDC because you need to check your facts. Based on IDC's
>>>latest server tracker data, for Q1-Q3 2008, 40,366 Itanium servers shipped (166,623
>>>Itanium CPUs). Clearly a bit off from 1M unless you are accounting for unsold chips
>>>and systems. 153,542 4-socket Xeon/EM64T servers (611,469 Xeon/EM64T Chips) shipped
>>during same period. Thats 15x delta!
>>>
>>
>>I say you are flat out wrong about processor count.
>>
>>See slide 12 here:
>>
>>http://download.microsoft.com/download/3/7/7/3772d75c-bce4-4a1c-977c-2f6bb6ef7fda/IDC_Keynote_Frankfurt_2008.pdf
>>
>>IDC says the average Itanium system sold for HPC in 2007
>>had 7 processors and was priced at $8966 per processor.
>>That means 150,000 processors shipped just for HPC and
>>HPC was less than a third of IPF sales in 2007. IPF sales
>>in each of Q1, Q2, and Q3 2008 were higher YoY according
>>to public IDC and HP statements and I doubt system price
>>per processor went up *at all* in 2008 let alone by a big
>>factor.
>>
>>Your implied claim of an average of 4.1 processors per
>>Itanium system can't pass a basic sniff test.
>>
>>See slide 9 here:
>>
>>http://hp-user-society.de/events/itanium/2006/vortraege/itanium_und_integrity_im_rechenzentrum.pdf
>>
>>>
>>
>>The trend over three years is clearly to higher processor
>>count in Itanium systems sold. In Q4 2006 the portion of
>>revenue represented by 2s and 4s systems had fallen
>>below 30% of Itanium sales two years ago.
>>
>>Also see slide 17 here:
>>
>>https://www-950.ibm.com/events/wwe/grp/grp010.nsf/vLookupPDFs/%EB%A1%9C%EB%93%9C%EC%87%BC%EB%8C%80%EA%B5%AC_%EA%B0%80%EC%83%81%ED%99%94-%EC%84%9C%EB%B2%84%ED%86%B5%ED%95%A9_081030(%ED%97%88%EC%9A%B1)/$file/%EB%A1%9C%EB%93%9C%EC%87%BC%EB%8C%80%EA%B5%AC_%EA%B0%80%EC%83%81%ED%99%94-%EC%84%9C%EB%B2%84%ED%86%B5%ED%95%A9_081030(%ED%97%88%EC%9A%B1).pdf
>>
>>IDC says HP shipped about 1200 Superdomes per quarter
>>for Unix in 2007. Some of those were PA-RISC - about as
>>many as IPF Superdomes shipped with Windows or Linux
>>and not counted in those 1200. IDC and HP numbers for
>>Q1-Q3 2008 all show higher Itanium sales each quarter
>>YoY. It isn't unreasonable that HP sold >4000 Itanium
>>Superdomes in Q1-Q3 2008:
>>
>>http://h71028.www7.hp.com/enterprise/cache/107846-0-0-0-121.html?ERL=true
>>
>>>
>>
>>"December 2008: IDC released a quarterly update to their Worldwide Quarterly Server
>>Tracker, citing market share figures for the 3rd calendar quarter of 2008 (3Q08).
>>Consider these telling facts:
>>
>>[..]
>>
>>Fact 8: HP is the leader in the high-end UNIX market, with a 41.1% revenue share, and a 42.9% units share.
>>
>>HP was the only major UNIX vendor to grow high-end UNIX revenue and gained 4.6 points revenue share year over year."
>>
>>
>>Your claimed processor numbers are totally incompatible
>>with these four linked presentations with IDC numbers
>>and public IDC press releases to date.
>>
>>Feel free to either post a link defending your claim or
>>retract it.
>>
>
>I'll need to figure out how to do this but I used the data directly from IDC's
>latest Quarterly Server Tracker data which is the December 2, 2008 release as listed here:
>http://www.idctracker.com/AllProducts.jsp?groupid=SRV
>
>For all "EPIC" servers shipped in 2008 this is what IDC shows:
>Sum of Units 40,366
>Sum of Factory Revenue (USM$) $3,616
>Sum of Customer Revenue (USM$) $3,802
>Sum of Core Count Shipped 292,161
>Sum of Die Count Shipped 166,623
>
>Maybe Die count does not equal CPUs sold??
>
>Also, spreadsheet shows HP shipped 942 x Itanium based SUPERDOMES from Q1-Q3 2008.
>
>I'll need to see where this latest data is located on net.

The revenue and server units agree with various public IDC
data.

Die/core ratio corresponds to 24% single core, 76% dual
core by shipments. Single core portion seems high given
Montecito has been shipping since Q3 06.

Die/processor count seems very low compared to published
IDC data for both average processors per IPF system and
average IPF system cost per processor.

Superdome number is reasonable if the IDC graph in the
IBM presentation was showing annualized run rate at each
quarter data point rather than rolling average shipments
per quarter.