Judge Dismisses Oracle Fraud Claim; Filing Shows HP "Strategically Screwed" by Itanium Cancellation
The Situation According to Oracle
According to Oracle, HP has paid Intel $690 million since 2008 to continue building Itanium processors and representing the processor as alive, well, and a crucial part of Intel's roadmap. The problem HP has is that it built an ecosystem around the Integrity server line and its HP-UX platform without ever providing an effective way to migrate such customers to Xeon-based systems. HP-UX doesn't run on x86 hardware, and the company abandoned attempts to port the OS several years ago. "As HP's Senior Vice President and General Manager in charge of its Business Critical Systems unit put it, HP was "strategically screwed."
The filing continues: "HP also did not want to reveal that much of the Itanium roadmap is, in its own words, "more an illusion than of technical significance." The secret agreement HP has with Intel does not obligate or incentivize Intel to develop great Itanium chips--with the performance gains one would expect from one generation to the next. Its purpose, again in HP's own words, is simply to "extend the Itanium roadmap to create market perception of long term viability." The 2008 agreement thus set minimal performance benchmarks for the last Itanium chip, Kittson, precisely so that Intel would be relieved of any implied obligation to make normal performance improvements."
Oracle claims HP internal documents include phrases like: "HP-UX is on a death march due to inevitable Itanium trajectory," and "The Itanium situation is one of our most closely guarded secrets and we have not wanted to let the region/field know about it since all it would do is give them another reason not to sell." In the wake of Oracle's initial announcement that it would cease supporting Itanium, an HP executive supposedly rang up Intel noting that the company needed to be able to communicate that Itanium was alive and well. "This is a MAJOR MAJOR issue. We need to be able to tell the market that you never told Oracle about EOL plans for Itanium. This is a CRITICAL element of the HP/Intel relationship. I don't view this as optional."
Oracle claims that it would never have entered into any agreement to continue supporting Itanium if it had known about these issues. It also claims that HP and Intel have deliberately suppressed information indicating that future Xeon processors would share a socket with Itanium chips. Given that no such product currently exists, it's not clear exactly how much supression has taken place.
It's entirely possible that Intel intends to EOL Itanium; the company has made no announcements regarding the future of the CPU after Kittson, the upcoming 22nm die shrink. Oracle's allegations that HP somehow engaged in deception by paying Intel to manufacture and design Itaniums is absurd. The <b>entire foundry model</b> is built on the idea of paying someone else to manufacture (and possibly improve) a processor design for you.
Oracle's allegations that Intel was under no onus to improve the chip's performance is interesting, but not terribly relevant in light of Poulson's improvements.
Other changes not listed here include the addition of a 32MB L3 cache, support for HyperThreading, and support for Intel's Instruction Replay Technology. There's also a set of Itanium New Instructions and the core is 12-issue up from 6-issue on the Itanium 9300 series. In short, it's one heck of an advance for a project that supposedly had minimal performance goals.
HP has effectively been Intel's sole Itanium customer for years. It's possible that the company really does want to kill Itanium. That doesn't make it unfair for HP to pay them not to do so and then to claim that the chip's future is assured. Intel's Poulson targets are aggressive and should yield a dramatically improved processor -- a fact that leaves us wondering what, exactly, Oracle thought it was protecting HP customers from?