There’s nothing like a vendor war to keep things interesting, and Oracle has launched a great one with its recent decision to stop development for future Itanium processors.
According to Oracle, Itanium is the deadest of dead parrots - not pining for the fjords, not tired from a prolonged squawk, but well and truly dead. Intel begs to differ.
To get a handle on this, I’ve put together a short survey (about six minutes) asking about Oracle’s aims, potential next steps, and impact on the industry. If you take the survey, we’ll send you a link to the results when they’re published. You can take the survey here. We’re clearing the decks to get this one done quickly, so get your responses in now - we’ll crunch the results fast and show ‘em to you as soon as we can.
What's going on
The latest round in the fight came during Intel’s analyst/press briefing on their new Xeon E7 x86 processor, in which Kirk Skaugen, GM of Intel’s Data Center product group) spent a fair amount of time spent discussing Itanium – namely, how it fits into Intel’s present and future. Unlike our pal TPM (see his story here), I thought Skaugen came out about as strongly as could be expected for Itanium. He positioned the newest Xeon and Itanium as Intel’s mission-critical, scalable plays.
Skaugen also made a statement that I hadn’t heard before: the choice between Xeon and Itanium processors in the future will be dictated by customer O/S choice. If you’re running HP-UX or NonStop, you’ll run Itanium processors; for Windows or Linux, you’ll buy Xeons.
He said we can expect to see the processors leapfrog each other in terms of performance. Itanium will have less frequent ‘ticks’ (speed bumps), and its ‘tocks’ (process shrinks and other enhancements) will take it to the forefront of performance during its two-year rev cycle. Xeon will have more frequent ticks along the way, as is typical. This implies that Itanium will move along at a quicker pace than before – something that will have to be proven before it is believed, given Intel’s track record on this processor.
When positioning the new E7 chips vs. RISC competition, Skaugen said that Intel ships around 18 million chips vs. 250,000 or so on the RISC side. This gives Intel much better economies of scale, enabling it to spread development and production costs over a much bigger base. Compared to Xeon, Itanium processor shipments are low. The same can be said for IBM’s POWER processors, as well as SPARC chips from Oracle.
However, IBM has its own fab facilities and also uses these chips in products ranging from videogame consoles to supercomputers – making continued POWER development well worth the money. Oracle has a fabless model, which takes some of the financial pressure off, but t it has few if any other uses for SPARC chips beyond putting them into their own servers. So the business case for further SPARC development isn’t quite as easy to see.
So what about Itanium? I asked Skaugen and the assembled Intel'ers if these economies of scale worked for or against Itanium. They pointed to shared Xeon and Itanium components, namely the QPI links, I/O chipsets, Scalable Memory Interconnect, and Millbrook memory buffers that reduce unique Itanium content. According to them, this significantly drops the cost of Itanium development and makes the business case for further development a clear win in their minds.
Oracle doesn’t believe this – or, seemingly, much of anything that Intel or HP say about Itanium. Oracle’s announcement, titled “Oracle Stops All Software Development for Intel Itanium Microprocessor” didn’t mince words. In the weeks since, we’ve seen Intel and HP defending Itanium and their future plans, while Oracle continues to assert that it’s well and truly dead, with only nails holding the dead parrot to its perch. We’ve also seen a lot of speculation about Oracle’s motivation in all of this.
Is it simply looking to cut development costs by removing a platform from the porting mix? Or is this the first move in a plan to use its power as a dominant enterprise ISV to move customers to an all-Oracle hardware, o/s, and software stack? What are the implications (if any) for the industry overall? So what do you think? What have you heard?
Step this way for the Oracle survey. ®