Archive for February, 2007

45nm Fabrication Plants

Wednesday, February 28th, 2007

Intel recently reiterated its commitment to ship 45nm products in 2007.

The “company remains on track to begin 45nm production in the second half of this year”.

Intel shall have two 45nm fabs in production by the end of the year.

An additional fab shall come online in the first half of 2008. Then another fab in the second half.

That will make four 45nm fabs.

Penryn is the codename for “Intel’s next generation 45nm family of products”.

Not too long ago Penryn booted its first operating system on first silicon, and the processor is now running both operating systems and applications.

Intel intends to retool an existing plant, Fab 11X in New Mexico, for its fourth 45nm manufacturing facility.

Fab 11X was Intel’s “first 300mm, or 12 inch, high-volume manufacturing facility”.

It came online in 2002 and currently manufactures 90nm products on 300mm wafers.

Test chips based on 45nm technology are currently being produced at fab D1D in Oregon. I wonder if the D in D1D stands for development?

Fab 32 in Arizona is scheduled to start producing 45nm wafers in the second half of this year.

Fab 28 in Israel in the first half of next year.

Then Fab 11X in the second half of 2008.

Source: Intel News Release

Interview with the Veep: The Intel Chronicles (2007)

Tuesday, February 27th, 2007

Pat Gelsinger, an Intel senior vice president and co-manager of the Digital Enterprise Group, gave an interview at a conference on power management in data centers.

Itanium and x86

Itanium is not only a different microarchitecture than Core. It’s a different architecture.

The instruction sets are different. The market segments at which the processors are targeted are different, too.

So anytime anyone starts talking about the convergence of Itanium with x86, one needs to keep these broad differences in mind.

On the other hand, certain technologies are common to both. Both use the same process technologies, for instance.

Intel hopes to continue leveraging technologies that it uses in its x86 products in Itanium.

These technologies include or will include things like design libraries, process and power management technologies, and circuit techniques.

Quad-core Itanium shall be “the next step in the product family”.

This next Itanium will move to common system architecture elements and align on design tools and process with Core.

On-die Architecture

The front side bus (FSB) is a bottleneck. Intel is therefore moving in the direction of an on-die interconnect.

Currently “near the CPU is a great sucking sound … the CPU starts hauling everything in”.

The system architecture of today will therefore give way to an on-die architecture.

More and more technologies shall “become part of the die”.

Transistor Architecture

There’s two things to keep in mind about the architecture of a transistor: the structure of the transistor and the materials of the transistor.

The materials of the transistor Intel does not expect to change much anytime soon.

The structure, on the other hand, seems to undergo constant research and development.

Intel is already looking at a tri-gate structure to replace the current transistor structure.

That being said, not even the structure of transistors is expected to change much throughout the 45nm and 32nm process technologies.

And 45nm is not even out the gate yet.

Source: The Big Interview: Pat Gelsinger

Data Center Design

Tuesday, February 27th, 2007

One of the senior vice presidents at Intel and general managers of the Digital Enterprise Group spoke at a conference on data centers.

Did you know that Intel has 140 data centers?

In data center design, the trick is to multiply the performance of servers while staying within the power budgets of those servers.

“The question is how much performance you can deliver within that thermal envelope”.

Source: Bryan Betts