IBM x450: Stuck between Power and a hard place
IBM's Itanic struggle
Analysis IBM's release this week of its first Itanium 2 server gave the public a glimpse into the internal schisms affecting a company trying to be a one-stop server shop while still showing its mettle as a vendor with enough engineering expertise to go it alone.
On one level, the x450 Itanium-based system is just another server. IBM already has a strong line of Intel-based hardware running on 32bit Xeon chips, which makes a four processor 64bit box look like a natural addition to the family. All the servers run Windows and Linux, and they use a common chipset -- Summit.
Like its Xeon companions, the x450 also performs well in benchmarks. Armed with four 1.0GHz Itanium 2 chips, IBM delivered a record score running a SAP benchmark on DB2 and Windows Server 2003. The box outpaced HP's comparable rx5670 server. (PDF)
IBM, however, has not uttered a peep about this benchmark, which is one in a long list of signs that the x450 is more than just another server. This product sits at the heart of a (T)itanic struggle within IBM.
The x450 is the bastard child in IBM's server family. The company said in July of last year that it would deliver an Itanium 2 server sometime in 2002. By August, it already had product up and running in demos, but somehow it took eight more months before IBM put the system on sale.
IBM says it wanted to wait until the Windows 2003 Server operating system arrived before shipping the x450.
This reasoning does not jive with IBM's intense backing of Linux or the adoption trends of Itanium. The x450 demoed last August already ran mySAP and DB2 on top of 64bit SuSE. In addition, Itanium 2 has seen the majority of its success in technical computing accounts where Linux is the favored operating system. So why would a company backing Linux with a billion dollars wait for Redmond?
The more likely scenario behind the x450's tardy arrival stems from a conflict between the xSeries Intel server folks and pSeries RISC server team at IBM.
The first Itanium chip was a disaster, and IBM was burned by agreeing to join Intel on stage with HP, Dell and others to champion the new processor. IBM and Dell showed their displeasure with Merced by giving McKinley -the second Itanium- the cold shoulder at its release last year. IBM and Dell muttered at times about their support for the chip, but left HP to take the stage alone with actual servers.
The slow adoption of the chip must have made IBM's xSeries and pSeries teams nervous. IBM revitalized its Unix server line with the 64bit Power4 processor, and the company did not want to cannibalise this. The pSeries executives spoke openly about their dislike for Itanium and desire to make the chip an afterthought.
With Dell holding out on Itanium 2 as well, IBM knew it could put some public pressure on HP by keeping its systems in the labs. HP was the only major vendor with an Itanium 2 server and looked a little silly sitting by itself on the deck of the good ship Itanic.
Compare IBM's strategy here to its support for the Opteron chip from AMD. IBM was the only big name vendor to bless this x86-64bit processor on its launch day and promised an Opteron server next quarter. With Opteron, IBM would stand alone, while HP chuckled nervously.
The Opteron chip poses less of a threat to Power and more of a threat to Xeon, so this move makes sense.
Itanic, on the other hand, runs square into the higher margin Power business. IBM prides itself on designing top class processors and Power4 stands as a shining example of research and development paying off.
The chip is the first from any RISC vendor with dual processor cores and has rivaled Itanium 2 in performance. What motivation could IBM have for supporting a rival's chip over its own homegrown product?
Out and proud. Sort of
Truth be told, IBM came out with an Itanium 2 system to save face and keep up its image as a complete IT store. IBM always stresses that it will provide whatever customers need, and, for the most part, sticks close to this strategy. Selling hardware based on a wide variety of technology helps open sales for the software and services teams that generate the most revenue at IBM.
When something like Itanium pops up, however, being all things to all people has certain costs.
It's for this reason that IBM's marketing team debated whether or not to even put out a public statement about the x450's availability, according to sources.
IBM was willing to ship Itanium; it just didn't want a lot of people to know about it. To the best of our knowledge, IBM only pre-briefed CNET about the system instead of doing the usual massive media outreach. We certainly were not on the briefing list, but that's alright since we announced the x450 to the world first.
In the end, IBM did the deed and buddied up to Intel for a day instead of championing its own product.
But now even with the product out the door, IBM's commitment to the x450 remains in question. The benchmark mentioned above used to be displayed with pride on Big Blue's server page. Thanks to Google for cacheing this page because after one call from The Register about the benchmark, it disappeared from IBM's performance hall of fame. (Update: The link was restored several hours after our story posted.) ®
Sponsored: DevOps and continuous delivery