Print 24 comment(s) - last by Calin.. on Feb 28 at 2:55 AM

Current Intel "Montecito" Itanium die
Intel's Itanium and Xeon architectures will have more similarities in the future

Intel’s enterprise Itanium and mainstream Xeon server platforms have remained completely independent from each other, in terms of technology. There has never been platform sharing, chipset sharing or anything else in that regards. However, Intel expects to converge the two server and workstation platforms. 

Pat Gelsinger, Intel’s senior vice president and co-general manager of the Digital Enterprise Group, outlined future convergence plans of Xeon and Itanium product lineups in an interview.

When asked how its Core and Itanium architectures will intertwine, Gelsinger responded “The first realization of that is Tukwila [quad-core Itanium] in late 2008, the next step in the product family, where we move to common system architecture elements, as well as full alignment on design tools and process.”

The upcoming Itanium Tukwila will take advantage of Intel’s common system interface, also known as CSI.

As Intel moves its Itanium and Xeon processors converge towards the common CSI-bus, the processors will share more similarities. Cache architectures of the two processors will become similar in the future, as there are no compelling reasons to have different cache architectures between Itanium and Xeon processors.

CSI-bus aside, Intel’s Tukwila-core is the first multi-core Itanium architecture. Tukwila is also the successor to Montecito, which Intel launched last year.

Intel expects to have common platform development as well. Common platforms will allow OEMs to have a single platform for Xeons and lower-end Itanium processors – easing OEM developments.

Those expecting Intel to produce a hybrid x86-64 and IA64 compatible processor will be disappointed, as Intel does not intend to take Itanium and Xeon convergence that far. “I don't see it getting that far, but I am driving these things to be as common as possible,” said Gelsinger.

Update 2/27/2007: The reference we had to Tigerton supporting CSI in 2007 was incorrect and has been removed.

Comments     Threshold

This article is over a month old, voting and posting comments is disabled

By Zandros on 2/26/2007 10:16:47 PM , Rating: 2
(last I heard, some variants of Tukwila will be well over 2 billion transistors)

Actually, the vast majority of those transistors are the L3 cache. In an old Anandtech article*, we see that the Itanium 2 core have significantly fewer transistors than a NetBurst Xeon. I think it may be 20 million compared to 80 million, but you can look it up for your self.

It is good to see that there still is life in EPIC and the Itanium.

Also of note is that POWER5+'s L3 cache is off die, which makes it seem smaller, when it might not be the case.


By killerroach on 2/26/2007 11:59:06 PM , Rating: 2
Agreed completely, although having the L3 off-die should do wonders to help yields (granted, at the expense of added latency). However, considering the price that a Montecito-core Itanium sells for, I think that's hardly a concern that Intel has at all.

By killerroach on 2/27/2007 12:05:27 AM , Rating: 2
In reference to what was mentioned about transistor count, the breakdown of the 1.72 billion transistors in a Montecito processor, from the Wikipedia article:

core logic — 57M, or 28.5M per core
core caches — 106.5M
24 MB L3 cache — 1550M
bus logic & I/O — 6.7M

So, excluding the L3, Montecito is roughly a 170 million transistor CPU, 63.7 million if you're just counting logic portions.

"Can anyone tell me what MobileMe is supposed to do?... So why the f*** doesn't it do that?" -- Steve Jobs
Related Articles
Intel Responds to AMD's "Torrenza"
September 27, 2006, 9:30 PM
The Count of "Montecito"
July 16, 2006, 3:39 PM

Copyright 2016 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki