Changes are a little more common on the server side of things. We see the Xeon equivalents of the new 6xx series desktop processors, as well as some changes in the Itanium roadmap. Intel has also cancelled the 3.80 GHz Nocona part due to its proximity with the 3.80 GHz Irwindale. The Irwindale has 2 MB of L2 cache - essentially the Xeon version of the Prescott 2M - and it should be noticeably faster than a Nocona with 2 MB of L3 cache.

It has been a while since the last server roadmap update, so we will include all of the recently launched Xeon "F" parts, which add support for EM64T along with XD support on the E-0 and later steppings. All of the Irwindale parts, incidentally, will include XD support from the start.

Intel Pentium 4 Xeon DP Lineup
Processor Speed Cache FSB Launch Date
Xeon DP 2.8F2.80 GHz1 MB800 MHzAlready Available
Xeon DP 3.0F3.00 GHz1 MB800 MHzAlready Available
Xeon DP 3.0F3.00 GHz2 MB800 MHzQ1'05
Xeon DP 3.2F3.20 GHz1 MB800 MHzAlready Available
Xeon DP 3.2F3.20 GHz2 MB800 MHzQ1'05
Xeon DP 3.4F3.40 GHz1 MB800 MHzAlready Available
Xeon DP 3.4F3.40 GHz2 MB800 MHzQ1'05
Xeon DP 3.6F3.60 GHz1 MB800 MHzAlready Available
Xeon DP 3.6F3.60 GHz2 MB800 MHzQ1'05
Xeon DP 3.8F3.80 GHz1 MB800 MHzQ4'04
Xeon DP 3.8F3.80 GHz2 MB800 MHzQ1'05 (late Q1)
Xeon DP 4.0F4.00 GHz1 MB800 MHzQ1'05

Intel Pentium 4 Xeon MP Lineup
Processor Core Speed L2 Cache L3 Cache FSB Launch Date
Xeon MP 2.0Gallatin2.00 GHz512 KB1 MB400 MHzAvailable
Xeon MP 2.2Gallatin2.20 GHz512 KB2 MB400 MHzAvailable
Xeon MP 2.7Gallatin2.70 GHz512 KB2 MB400 MHzAvailable
Xeon MP 3.0Gallatin3.00 GHz512 KB4 MB400 MHzSoon
Xeon MP 3.16Cranford3.16 GHz1 MB 667 MHzQ1'05
Xeon MP 3.50+Potomac3.50 GHz1 MB8 MB667 MHzQ2'05
Xeon MP 3.66Cranford3.66 GHz1 MB 667 MHzQ1'05

Most of the Xeon processors have been available for some time. The big news mirrors the desktop platform, where we see the addition of the Irwindale processors with 2 MB of L2 cache. One difference between the two is that the EM64T support has been disabled (for now) on the desktop processors, but it is fully functional on the workstation chips. The Irwindale also supports running in dual processor configurations. There are also uniprocessor Xeon chips available, and their launch dates mirror the DP Xeons.

The Xeon MP processors are still based on the old Gallatin core, and Intel extends the life of these platforms with the introduction of the 4 MB L3 cache chips, which will allow for a simple upgrade for companies that feel it is worthwhile. Moving forward, Intel has the Cranford processor, which is a Prescott core with multi-processor support running on a 667 FSB. Then in Q2'05, Intel will launch the Potomac core, which is a Prescott core that includes an addition 8 MB of L3 cache, again running on a 667 MHz FSB. The launch target lists a speed of greater than or equal to 3.50 GHz.

Itanium 2 DP Lineup
Processor Core Speed L3 Cache FSB Launch Date
Itanium 2 1.0 LVMadison1.00 GHz1.5 MB400 FSBAlready Available
Itanium 2 1.3 LVFanwood1.30 GHz3 MB400 FSBQ4'04
Itanium 2 1.4Madison1.40 GHz3 MB400 FSBAlready Available
Itanium 2 1.6Madison1.60 GHz3 MB400 FSBAlready Available
Itanium 2 1.6Fanwood1.60 GHz3 MB533 FSBQ4'04

Itanium 2 MP Lineup
Processor Core Speed L3 Cache FSB Launch Date
Itanium 2 1.3Madison1.30 GHz3 MB400 FSBAlready Available
Itanium 2 1.4Madison1.40 GHz4 MB400 FSBAlready Available
Itanium 2 1.5Madison1.50 GHz4 MB400 FSBQ4'04
Itanium 2 1.5Madison1.50 GHz6 MB400 FSBAlready Available
Itanium 2 1.6Madison1.60 GHz6 MB400 FSBQ4'04
Itanium 2 1.6Madison 9M1.60 GHz9 MB400 FSBQ4'04

The Itanium platforms have mostly remained static of late, with the only changes being delays and removal of parts from the roadmap. The first Fanwood parts have been delayed to Q4'04, and the 1.7/4M Fanwood has been removed altogether. Intel isn't entirely clear on how the Fanwood and Madison cores differ, but Fanwood appears to be a tweaked Madison core with better power characteristics. Madison 9M parts experienced a similar shift, with the 1.6/9M part moving to Q4'04 and the 1.7/9M part disappearing as well. Intel is not halting the Itanium platform of course, and there is a vague "TBD" in Q2'05 for a 667 MHz FSB Itanium platform.

As far as chipsets go, Itanium will continue with the E8870 for the forseeable future, while Xeon chipsets will have Tumwater at present with more options on the way. Lindenhurst support for 2P systems is starting to ship, and that will last for the next year at least. 1P systems are now available with the Copper River chipset, with Mukilteo coming in the second half of 05. On the MP front, Intel has their Twin Castle chipset planned for the launch of the Cranford and Potomac processors.

That about wraps it up for this edition of the Intel Roadmap Update.

Index
Comments Locked

6 Comments

View All Comments

  • stephenbrooks - Saturday, September 4, 2004 - link

    Jarred, thanks for explaining my query in #1!
  • JarredWalton - Friday, September 3, 2004 - link

    Yeah, what Kristopher said. Intel does not use the same naming mechanisms in the servers/workstation lineup as they use in the mobile/desktop segment. At least, not yet. They're sticking with clockspeeds, letters and cache sizes for now. I prefer this mechanism of "naming" chips anyway.

    Something I didn't mention in the article, notice how there are actually multiple chips with the same processor number? The Celeron D and Celeron M both have chips that use the same model number. 330, 340, and 350 for sure, although there may be others. I'm just waiting for the Model number 386 and 586 chips to show up on their roadmap! :-D
  • KristopherKubicki - Friday, September 3, 2004 - link

    ViRGE: I think you answered your own question, they are just E-0 Pentium 4s with EM64T. the Pentium 4 "J" series processors are following the Model number naming convention, the "F" series are following the GHz naming convention.

    Kristopher
  • ViRGE - Friday, September 3, 2004 - link

    Correct me if I'm wrong, but aren't all EM64T enabled Pentium 4F's using the E-0 core? And if so, why aren't they being called the P4J instead?
  • JarredWalton - Friday, September 3, 2004 - link

    The "F" suffix indicates activated support for EM64T. There are Xeons available without that support as well, and they have the "E" suffix, but they have all been shipping for a while now, so they aren't in the charts. Also, all steppings "E-0 and later" include XD support on the 1 MB cache chips.
  • stephenbrooks - Friday, September 3, 2004 - link

    Aren't some of the Xeons listed as things like '3.0F' actually '3.0E' because both the 1MB and 2MB parts are in that list. I'd've thought the 1MB ones would need the E suffix not F.

Log in

Don't have an account? Sign up now