3DVelocity logo


Content

Latest

News
Articles

Community

Forum


Intel @ ECTS 2002 - part 1
Our Summer 2002 interview with Intel happened at the ECTS 2002. It is a two part series, with the next part out a few days later. Last year's interview can be read here. Here are some pictures of the contents within the Intel stand.

     


Our interviewees go by the faces of :



Left is Mathias Raeck and on the right is Dan Snyder. Their positions and job descriptions are below. We think Mathias looks a bit like Tom Cruise in Mission Impossible, but we might be wrong. Note the biscuits to the bottom of the picture, we can throughly recommend the ginger ones.

Please introduce yourselves to our readers.
I am Dan Snyder and my colleague here today is Mathias Raeck. I run the developer relations group based in Munich where we mainly deal with software developers. Mathias is the UK press specialist.
{In this article most of the replies are from Dan Snyder}

Who and what was the Pentium 4 initially designed for?
When we initially launched the Pentium 4 it was designed for multimedia. At the time we had an advice model to find out what people were doing and this was mainly entertainment (MP3/MP4), digital audio/video, communications and imaging. So we designed the Pentium 4 around these uses.

We also took note of the boom in MP3/MP4 interest so the Pentium 4 is really good for encoding and decoding on those formats.
Did Intel capture the area of the market they intended?
Definitely. You look at it now with almost all our products sold at retail are Pentium 4 based or Celeron. This was the fastest transition we've ever had.
When the Pentium 4 initially came onto the scene, it generally got bad reviews. How damaging was this at the time?
Whenever we launch a few product, a few things happen. One is that you are testing a new architecture, so there is a time gap until applications start taking advantage of the new architecture.

Two, the Pentium 4 architecture was designed for speed. We designed it so that it could be cranked the megahertz so in the initial parts didn't take full advantage of all those economies of scale.

Using Itanium as an example, it took a while to enable the market fully - to get the applications there and the economies of scale. Now that's there, you can see that the Pentium 4 will reach 3 Ghz by the end of this year which confirms that the scalability of clock speed on the Pentium 4.
{Itanium is discussed in the second part of our interview}

How damaging is this in the longer term?
Not at all.
Could the Pentium 4 launch have gone better?
At the time, Intel was going down the e-business path, and a lot of the optimization and resources went down this route which is bearing a lot of fruit today where our server market segment share is growing. We are selling tonnes of Xeons and Itaniums so that's great for that, however we turned away from the gamers, consumers and so on. When the Pentium 4 launched we realized that we still had to work with these guys because the applications weren't there.

The dilemma is whether to first deliver the architecture to the market or the applications. Being a technology leader we usually go to market with a product then try to enable our environment/community to optimize games and other applications. This doesn't only apply to Intel, if you are a leader in any market you face this dilemma.

Overall the Pentium 4 launch was successful. Today we have over 500 applications optimized.
Was the lacklustre performance of the Pentium 4 initially the main reason for AMD to gain market share on Intel?
Whenever you have a transition there is a transitory period. When the Pentium 4 came out the software was lagging behind and the CPU was priced a little high, so there is a period of vulnerability and AMD swept in and by using a few deceptive naming of processor (such as 2200+) but with a good product nonetheless. However we kept going with a good product and with 3 Ghz coming out at the end of this year we will have almost 1 Ghz lead over AMD.

Over the last two quarters we've started to regain market segment share.
This time last year you mentioned that 3 Ghz+ CPUs were already in testing, what speeds are you testing now?
At IDF {Intel Developer Forum} we showed a 4 GhZ part which was several months ago. With the fall IDF coming up in a couple of weeks you can expect to see something even faster - possibly a 5 Ghz part.

With the Terahertz announcement a few months ago, it showed that we have proven a path in the labs to a billion transistors and 20 Ghz in 5-6 years.
Why is there such a long lead time if you have had a part working at 3 Ghz for a year now and we only have a 2.8Ghz part for sale?
My {Dan Snyder} background is engineering. I worked in our testing department doing quality/failure analysis. The thing Intel is absolutely against and will not do is launch a part before it's ready. We don't want parts coming back on fire, we don't want things failing so we test the hell out of our products - I know because I have done it myself.

So yes we may have one part working, however we don't launch a product because we have one part working. We only launch it when we can say 2.8 Ghz is out and Michael Dell phones us and says we want 10 truck loads next week and we can provide it. That's the difference between a paper launch and a real launch.
How important was the recent increase of front side bus speed to the performance of the Pentium 4?
It was pretty important, along with moving to 0.13micron core. Roughly we see around 10% increase in speed due to those enhancements alone.
Has DDR overtaken Rambus as the memory of choice for Pentium 4 owners?
We support three memory standards now. We still are firmly behind RAMBUS. It is for the high end, the leadership platform. We appreciate that pricing issues, with RAMBUS memory prices not coming down as much as customers and we would have hoped lead to demand for other RAM variants. So we had to consider people who wanted a price point over performance, we do that through our 845 chipsets.

If you look at the sales figures of our motherboards it clearly shows that DDR based solutions are the most popular. SDRAM is a small percentage, catering for the value market with RAMBUS also being a small percentage catering for the high end, leadership platforms.
Does Intel officially support DDR RAM now?
We fully support DDR RAM. We say that RAMBUS is the highest performance RAM you can buy for the Pentium 4 platform. We acknowledge and say that DDR offers great performance per Dollar. If it's on our motherboards we support it.
In the second part we talk about Intel's Itanium 2 processor, 64 bit desktop processing, OEMs and money matters. Be sure to check that out.
comments powered by Disqus