HP: last Itanium man standing
Nehalem lives the dream
Itanium fared no worse than the market at large, although there is every reason to believe that with big Opteron and Xeon iron out there and Microsoft and Red Hat pulling the plug, Itanium may not bounce back or even hold its position in the coming years. Itanium is just in a niche, not the dominant platform as Intel and HP had dreamed.
While that makes Itanium the butt of many jokes and a sore disappointment to HP, Intel, many server OEMs, operating system makers, and customers who endorsed the platform and had high hopes for it, that doesn't mean Itanium servers are not a good business, at least in the near term, like mainframes or proprietary minis rehosted on anything other than an x64 chip. If you chant Itanic three times and click your heels together, it is still not going away. HP and Intel could shoot it in the head, of course, but not for a few years.
What Intel has its mind set on, and what Itanium was always part of, is attacking the remaining bits of the server racket that it does not own because this is where the big money is. The funny thing is, once Intel tears down the midrange and the high-end and commoditizes it, the chip maker may find that the money vanishes, like a mirage, because all the revenue Intel can't get its hands on with x64 chips is there precisely because porting applications off other platforms is such a herculean task. Once everything runs on x64 iron and there is no upward pull on server prices, maybe the whole damn thing goes commodity and no one can make a dime off this racket any more.
El Reg sat down recently with Shannon Poulin, director of Xeon platform marketing at Intel to discuss the "Nehalem-EX" Xeon 7500 launch and its impact on the Itanium business. And here is the chart that sums up what annoys Intel to this day:
In this chart from Poulin, MSS is short for market segment share, and as you can see, the Intel share of the server revenue pie, which includes both Xeon and Itanium server sales, has been trending upwards at a steady clip. But at this rate, it will still take Intel another 10 to 15 years to get Intel's revenue share anywhere near the dominant shipment share it enjoys in the server racket, which is north of 96 per cent most quarters.
Poulin says that great progress has been made in getting companies to adopt x86 and x64 platforms. In 2000, some 1.6 million non-x86 systems shipped worldwide, and by 2009 he estimates that some 200,000 non-x64 systems (with around 400,000 processors) shipped against some 7.5 million total servers consumed by the companies and governments of the world. But for all Intel's efforts, it still has less than 60 percent of the server revenues, and it wants more.
That means not getting religious or even philosophical about Xeon versus Itanium, a fight that was over in 2003 when AMD shipped the first Opteron and it became clear that Intel would have to counter with much-improved Xeons rather than try to undercut AMD with an Itanium roadmap that had serious issues.
"We're not going to hold Xeon back in any way," explained Poulin. "We're going to put as many capabilities as we can in both products." As for what ISV and server partners might or might do, Intel seems pretty resigned, but also pretty confident that the Xeon 7500 lineup can meet the requirements of 90 percent of the so-called "mission critical" server segment. "Whatever happens, happens. We have two horses in a four horse race." That's Xeon and Itanium versus Power and Sparc, in case you are wondering.
You didn't actually think Intel would admit that the Opteron is suitable for mission-critical systems, did you?
So, as HP launches its revved Itanium-based systems this week, the talk will turn, as it always has in recent years, to what HP will do with HP-UX, OpenVMS, and NonStop. The answer will be: as little as possible to make the money. And that means sticking to the Itanium roadmap and staving off the day when HP will have to port its own operating systems to x64 iron and try to get as many of the 14,000 Itanium applications as possible to jump to yet another platform.
Two thoughts. First, wouldn't it be funny if the "Sandy Bridge" Xeon chips had an optional Itanium emulator that would plug into the chip package and help run emulated Itanium code? (The Itanium chips had an x86 emulator, you will remember, and also emulated some PA-RISC instructions that HP-UX needed). And wouldn't it be far easier if HP just bought Red Hat and asked its HP-UX customers to make just one more port and compile, to RHEL running on x64 iron? Novell would be cheaper to acquire and still supports Itanium with its SLES Linux, and that would also do. ®