Martin Hingley's Blog

Itanium Update Q210

• Intel launched the Itanium 9300 family in February 2010
• Both Microsoft and Red Hat have dropped Itanium as a supported chip for future server operating systems
• The news is negative for the very few non-HP Itanium users
• Itanium continues as the engine for all HP Mission Critical servers
• IA64 had been widely adopted and heralded as a world-beater before its delayed launch in 2001
• HP uses Itanium-based Integrity servers as merger platforms for HP-UX, OpenVMS and NSK customers
• It is unlikely that Intel or HP will give up on Itanium any time soon
• Itanium will help HP become more vertically integrated

Microsoft And Red Hat Drop Itanium Support – Few Consequences For HP

Both Microsoft and Red Hat have decided to drop support for Intel’s Itanium processors in the last few months – more specifically that Windows 2008 R2 and Red Hat Enterprise Linux 5 – will be the last of their operating systems supporting Itanium. This is despite the introduction of the new Itanium 9300 family by Intel in February 2010. My view is that this is insignificant in terms of the amount of business they are making from installing their operating systems on these machines and also for HP’s server business. Itanium remains an important platform from which HP builds all of its Mission Critical servers and, although Intel and AMD’s chips encroached on its power and features, it is extremely unlikely that HP will jettison Itanium any time soon – not least because it creates a vertically integrated platform other server suppliers are unable to ‘change the engine’ for. If anything the movement of suppliers to becoming what I call ‘matrix integrated’ vendors means that we’re going to see increasingly numbers of proprietary servers in the next few years.

HP And Intel Looked Onto A World-Beater When Announcing IA64

The beginning of Itanium came with HP’s decision in the early 1990s to cease its chip production business. Given the increasing costs of fabrication it threw its hand in with Intel in 1994 to develop a replacement for its PA RISC processor, which was to have an Explicitly Parallel Instruction set Computer (EPIC) architecture. By moving the logic of how parallel instructions work from the chip to the compiler, they would be able to handle Very Long Instruction Words (WLIW) running up to 6 instructions through the processor at the same time. Between them the two companies persuaded just about everyone that IA64 was going to be a world-beater – it would be the natural successor to proprietary RISC chips and there were a number of successful attempts – later dropped – to convert operating systems such as Sun Solaris and IBM’s AIX to run on it. It would also offer x86 server vendors like Dell an easier step up into the 64-bit minicomputer market.

Why Itanium Failed To Be Adopted Widely

There were a number of reasons why Itanium (as it became known in 1999) failed to live up to its promise. In particular:

• It was late, eventually shipping for the first time in the middle of 2001
• It was initially underpowered – offering far less performance than expected
• Writing compilers was difficult, although HP did a great job with HP-UX
• It allowed its suppliers (HP in turn for PA, Alpha and NSK) to offer customers a slow withdrawal from older machines, limiting its take up
• There was a surge of sales of Sun’s Sparc based machines in the late 1990s and into the early 2000s, making it look like ‘fab-less’ chip design was an alternative
• IBM rediscovered its advantages of designing and fabricating both the z and Power processors
• Linux was successfully ported, but neither Solaris nor AIX were ever sold on Itanium
• AMD shifted x86 to 64-bit in 2003 with the introduction of its Opteron chips, catching Intel napping and speeding up its introduction of 64-bit Xeon processors

Itanium Helped HP In Acquiring Compaq As A Merger Platform For OpenVMS and NSK

However it continues to be a success for HP. It went beyond just being ‘fab-less’ to persuading the world’s biggest microprocessor vendor to share the design and take on manufacturing of its next high-level server chip. For many years I thought that a failure of Itanium might kill HP, but hardly scratch Intel. But it helped in the acquisition of Compaq begun in 2001: HP’s Integrity servers became destination platforms for both of Compaq’s minicomputers – OpenVMS and (eventually) Non Stop Kernel. The fact that three months before the acquisition announcement Compaq had decided to stop developing its Alpha chip and move to Itanium removed a number of objections, since the road map from both companies was firmly with Itanium. I’ve estimated HP’s Mission Critical server business by CPU in Figure 1, which gives rolling q4 totals.

Some Conclusions – HP Is Intel’s Biggest Customer And Is Highly Unlikely To Drop Itanium

Unlike in 1994 HP is Intel’s biggest customer today, so it is unlikely that Intel will drop Itanium any time soon. I certainly don’t expect HP to voluntarily move its three operating systems, HP-UX, OpenVMS and NSK, to x86 platforms, no matter how powerful they become – they help the company a greater level of control and allow it to address more sophisticated vertical market solutions than it could with x86 machines. Neither Microsoft’s nor Red Hat’s decision to stop supporting Itanium will have much effect on their business with HP – who is almost certainly the biggest customer for both. It will have a negative effect on any users who use their operating systems, rather than HP’s. However a) there aren’t many of you and b) there are powerful alternative x86 servers to move to instead. I believe that HP’s moves towards a more vertically-integrated strategy is actually helped, as Itanium moves away from its potential as an ‘industry standard’ system.

Do you run Itanium servers? Which operating system do you use? Let me know by commenting on this post.

Filed under: HP, Server Tagged: | AMD, HP, HP-UX, Intel, Microsoft, NSK, OpenVMS, PA, Red Hat