Intel Developer Forum 2002 March 04, 2002 Sarju Shah

Introduction

Page:: ( 1 / 3 )

Rolling over to SF for IDF is about as easy as it gets for us when it comes to trade shows. From our offices, we have a direct line of sight to the Bay Bridge. Why bother with traffic reports when you can just look out the window. Compared to other conferences, like E3 and Comdex, IDF is a relative breeze when it comes to registration and getting about. Even though security was supposed to be very thorough, there weren’t any delays because of them. The only thing that got in the way this time was I. With school and midterms, getting around to IDF proved to be very difficult.

Just like at Comdex, tons of new stuff is announced at IDF. The big difference here is that it’s mainly Intel’s day in the limelight, or shall I say week. Intel itself announced quite a few things on the first day of the show.

On the general Intel info stuff, they now have five fabs running at .13 micron, with plans for a new plant in New Mexico. They also started using 300mm wafers at a plant in Oregon, along with a .13-micron production line; this yields roughly four times as many processors as an .18-micron process paired with 200mm wafers. It seems as though Intel doesn’t seem consumption going down the coming years, in spite of the economic downturn.

Keynote

Starting with Craig Barrett’s (CEO of Intel) keynote speech, Barret explained how Intel was going to get through the recession. It was going to spend itself silly. In doing so, they hope to create loads of new products that will stimulate the economy. This does make some sense, as sitting idle and waiting for things to get better sure isn’t going to help, especially if you want to be the one on top in the end.

The excitement began when Barret and his cohort proceeded to create an entirely digital movie using an air-cooled Pentium 4 3GHz machine. All the animation was done by a program that extrapolated just about everything it needed from a series of digital still shots. After that things died down for a bit, until they started to feed HDTV broadcasts through these computers. In one single demo, they showed that “fast” just is not enough. By having the computers decode HDTV broadcasts, the machines were almost completely taxed. The network bandwidth between these two computers was nearly saturated. The need for speed is out there, even before we get “there”.



SIDEBAR: The show floor at IDF starts to close the day before the last day. Remember that before you drive out for no reason.

Drives, and Processors

Page:: ( 2 / 3 )

Itanium

Going along the Itanium lineup, Intel continues to expand, announcing an entire lineup. Codenamed McKinley, Madison and Deerfield, and the far off Montecito. The Madison and Deerfield lines are scheduled to be built on the .13 process, while Montecito is to be built on a .09-micron process sometime in 2004. McKinley is supposed to pop into production pretty soon with a whopping 3MB of L3 cache. Moving a little further down the timeline, Madison appears with a 6MB L3 cache.

Plumas

With the upcoming Plumas E7500 server chipset, Intel has decided to use dual channel DDR SDRAM, forgoing the expected RDRAM path. The new Xeon processors that go in the chipset will run at 1.8,2, and 2.2GHz. It has been quite some time since Intel released a server level chipset, and it means quite a bit that they went with DDR SDRAM. If one looks at Intel’s current chipset pipeline, it’s rather interesting to note that Rambus doesn’t seem to play a large role.

Serial ATA II?

After getting everybody all aboard for Serial ATA, it was time to up the stakes on this fledgling technology. Whereas the original specification was designed with the home PC in mind, Serial ATA II intends to bring all these benefits to the server market. This new spec will bring with it second generation signaling rates. Normally the spiffy stuff trickles down, not up.

For those not in the know, Serial ATA is the planned replacement for the current connector system for IDE hard drives. Instead of utilizing a wide data cable, Serial ATA will turn that into one very thin wire. Not only does this result in cleaner wiring systems for systems, motherboards design benefits from routing fewer traces due to the reduced pin count. When the spec. gains more momentum in the desktop scene, those two wires should come down to one for both power and data. All you case modders out there will have a field day with this stuff when it finally comes out.

Other companies that have come out in support of Serial ATA include the following companies – Adaptec, Maxtor, NEC, Seagate, Eurologic, and Molex.


SIDEBAR: The food in the Press lounge is rather impressive. Their cookies are delicious!

Cases, Ethernet, and RIMMs

Page:: ( 3 / 3 )

Gigabit Ethernet

New on the Intel networking side is the introduction of three single chip Gigabit Ethernet chips. Covering the range from server to desktop PC these chips provide absurd transfer rates while at the same time reducing size and power consumption. The new controllers are dubbed the Intel 82546EB, 82546EM and the 82540EM. The first two controllers are meant to enter the server market, providing dual port access with a single chip. The 82540EM is meant for the desktop PC, the current size of the chip is smaller than a dime, resulting in minimal changes for motherboard manufacturers and thereby reducing the time to market.

While these chips have little direct affect on the little guy surfing the web at home. It is only because of advances in technologies such as these, that he is able to do so. According to the report by Sean Maloney (Executive Vice President, Intel Communications Group), the amount of data tossed around has grown at an alarming rate. Starting in the year 2000, under 40,000 Terabits were sent by Enterprise level companies. By the year 2002 we have approached 120,000 Terabits. Using projected growth rates, by the time we reach 2006, that number will have increased by over thirteen times, reaching 1,600,000 Terabits. The need for greater bandwidth is indeed there; just wait till the rest of the world gets online.

More on the Intel networking side is a little thing called XScale. In an attempt to make modular designs in the communications sector, where most products are normally of proprietary nature, Intel has designed the XScale processor. This would be the server side equivalent of their gigabit network chips. These new server chips will bring about some really nifty things. With XScale enabled hardware, just about anything under the sun will be able to interact with each other.

Form Factors

Not too much information was available on this rather interesting tidbit. There are quite a few different case form factors in the pipeline at the moment. All want to reduce the noise, and space requirements taken up by the modern PC. New designs in power supplies, heat flow and case design will help to reduce noise and improve on aesthetics that have been shoved under the mat for nearly two decades.

Intel has two programs running at the moment: Tidewater and Big Water. Tidewater is supposed to make the microATX standard even smaller, while Big Water is supposed to shoot for the stars and come up with something in the revolutionary.

Rambus

While Rambus and Intel seemed to have parted like a pair of scorned lovers out of some Jerry Springer rerun, Rambus is still trying to put out. On display were RIMMs capable of 1066 and 1200. Soundly trumping anything DDR has to offer for quite a while, and at the same time seemingly an afterthought to the company that has had 800 out for years now. According to Rambus, 1066 should be cakewalk to manufacture. It seems to only be waiting on the release of the 533MHz Northwood.

Why do we care? The lower clocked Northwoods are almost demon-like in their overclocking abilities. With reports of 1.6-1.7GHz versions revving up to 2.3-2.4 GHz, it’s no wonder that people are excited about Intel. Currently most of the success has been had on the 845DDR, but once the i850E is out, 1066 is sure to cause a stir. With costs nearly on par with DDR, if not lower, better performance and stability could be had out of these systems. People have already started calling it the second coming of the 300A, and with any luck we’ll have them by Easter.

Also on display at the Rambus booth were 32-bit RIMMs. Another major complaint with Rambus has pretty much been leveled. Many a person complained about having to purchase two RIMMs in order to make a system function with the dual channels. Now that 32-bit RIMMs are here you only need one RIMM to make a system function.

Alas

That should be wrap for our IDF coverage. IDF plans on coming around twice a year; with any luck we’ll have better coverage of it in the fall. Until then wait patiently for the new toys. As a teaser, stay tuned for a new 1066 article paired with the Northwood.





SIDEBAR: What are you looking forward to? Northwood? Serial ATA? Speak in the news comments

!