David Kanter (dkanter@realworldtech.com) on 12/5/05 wrote:
---------------------------
>>I could only find the Sungard app benchmark mentionned in >Intel promotional material,
>>before it shipped with the Bensley review systems.
>
>It was mentioned before, SunGard was demoing said benchmark and app at IDF. It
>was actually used in a demo that Pat Gelsinger did (I think). It was an onstage demo at any rate during a keynote.

I was aware of this, since Sungard mentions it in the press release I linked. To what extent IDF is a promotional event is a matter of personal appreciation, but I considered it as such in my post. I did not link to it yet, but I also found out that Intel used the Sungard benchmark result to claim that Xean was "Ideal for intense multi threaded applications".

target="_blank">http://cache-www.intel.com/cd/00/00/23/57/235782_235782.pdf

>Intel didn't have source code access to the app in question. I asked about what
>sort of access Intel had to it, and how they tuned it. The answer was that it wasn't
>specifically tuned for Intel at all. AFAIK, Intel helped to make the front-end,
>i.e. the UI that we use, and that's it. They didn't have any access to SunGard's code.

Who did you ask about this? What is the source for this piece of information? From Sungard's press release:

Don Wood, chief technology officer of SunGard’s financial systems business said "We work closely with Intel software engineers to deliver solutions to SunGard customers running our applications on Intel-based platforms."

If the "we" in your above sentence is you, when you used the benchmark, and the UI is the benchmark UI, then it seems to me that Don Wood is referring to a different kind of collaboration (it is safe to say that most Sungard customers have probably never heard about this benchmark, and I suppose so since I have participated to the evaluation of an implementation of Sungard). If I interpret you sentence incorrectly, would you be a bit more specific?

By the way, what is the license of the benchmark? Is it freely distributable? Would you make it available?

>I'm not under the illusion that Intel is going
>to promote benchmarks where it loses (they probably won't), but I think that they
>are keenly interested in any benchmarks that exhibit scalability, irregardless of
>who wins.

I share your opinion on both aspects. No matter what benchmark they show internal interest for, they are not going to promote a benchmark that shows their solution loosing, even wen it shows nice scalability. Internal technical interest is one thing, public relations and marketing another. Intel shipped the Sungard benchmark with some systems knowing full well that it performed better on their systems than they do on competitive offerings.

>You have to keep in mind that many things which are bad on the P4 today,
>will not cause problems for Merom in the future. What's the point of going out
>of your way to find a benchmark that favors the P4, to then explain later why it
>sucks because Merom might not do well? That doesn't make any sense to me.

At this stage, I suppose Intel has a very good idea about which benchmark will do well on Merom, and which will not. Same for K8, at least to a point.

>Your query was certainly phrased in a way that would lead
>the reader (at least myself and some others) to conclude
>that you believed that Intel required that the benchmark
>be used.

That sounds reasonable to me. While I indeed considered this possible and made it explicitely clear, mainly from the considerable similarity between the benchmark set used by Tecchannel and in your review, your answers to my question led me to reject that hypothesis.

Prejudice would have been to reject this hypothesis before asking you, in my opinion, or formulating the hypothesis and consider it as a truth without explictely trying to be informed the truth of the matter.

- Armand