The third Itanium:
Two years after the release of the first 64-bit Itanium (code named Merced) and 1 year after the release of Itanium II (code named McKinley) the third version of Itanium is about to be released (code named Madison).
Intel finds itself in the utterly unexpected position of being in a "do or die" situation with the third Itanium model. In some ways, this is really only a second version, since the first Itanium (Merced) was more for development and evaluation than production use. Because Itanium uses IA-64, a completely new and very different architecture, it was necessary to seed the market with an early version. "Wait for McKinley" (Itanium II) was the semi-official response to any questions about the light volume of Itanium I sales.
But Itanium II did little better than Itanium I, and recently was pretty much pulled from the market due to a very minor bug that was found in a few of the chips shipped in production systems.
None of this used to matter, since Intel pretty much had forever to get its 64-bit chip to market. SUN's 64-bit SPARC was suffering from low performance, the powerful Alpha was being dropped by HPAQ, and the PA-RISC line was being merged with Itanium itself - there was no competition for Itanium, so it really didn't matter how long Intel took to bring its 64-bit chip to market.
This all changed 2 months ago when AMD began shipping its 64-bit Opteron server platform. The current version of the Linux Journal is one of their periodic "focus on hardware" issues, and the cover story is on Opteron. The issue includes 11 ads for servers with Operon, 5 ads for servers with Itanium, and 2 for servers that can be had with either chip.
Given that the segment of the server market that's seeing the most growth is the high density Linux server market, this is a significant indicator of where the market is going. To put this in perspective, Opteron has been shipping for 2 months, and Itanium has been shipping for 2 years. Intel's historic revenue share of the server market has been about 100 times as large as AMD's.
Although servers only represent about 5% of the units sold, they account for a much larger share of revenue. And since the chips don't cost that much more to produce, they can account for an even larger share of profits (depending on whether or not volume is high enough to cover development costs).
The move from one generation to the next in CPUs always seems to happen suddenly, and somewhat unexpectedly. Zilog pretty much owned the microprocessor market (invented by Intel, then lost) with its famous 8-bit "Z80" until Intel came along with the 16-bit "8086". The 16-bit 8086 family was all anyone would ever need until the 32-bit "80386" came along - and suddenly 16-bits just wasn't enough. (there was an 80286 in there, in between, that had some 32-bit features, but it had some serious design issues that resulted in Bill Gates' famous "brain dead" characterization).
Judging by the count of server systems being advertised in the current issue of the Linux Journal, the server market is making the move from 32-bits to 64-bits - now. And that's why I say that Intel is in a quite unexpected "do or die" position with the third version of the Itanium - if AMD continues to dominate in the 64-bit server market, Intel could wind up losing a big chunk of its revenue and all of its profits.
Market Data 
Markets 
AI
