Itanic: Davy Jones' Locker
By dcb on Mar 29, 2005
In the year 2000, just as the first Itanium processor from Intel hit the market, IDC predicted that 2004 Itanium server sales would hit the $28 billion mark! But IDC missed their projection slightly. They were off by about $26.6 billion, or ~95%. Ouch!!
Of the few Itanium-based servers that were actually sold in all of 2004, HP lead the "crawl" and accounted for 76% of them. But HP, as of mid-2004, joined Sun and IBM in the Opteron-based server market, so expect Itanium sales at HP in 2005 to slow at a faster rate than HP's general server sales numbers. IBM came in 2nd with 10% of the Itanium market, but has strongly hinted that they are killing off their Itanium-based server offerings in favor of Opteron, Power, and traditional Intel processors. Dell captured 3rd place with just 5% of the tiny Itanium pie, and so far Dell has resisted selling Opteron-based servers... but how long will Michael watch from the sidelines?
For those who like to look under the hood, it seems to me there are three server-oriented processor families that deserve attention and will still be important in 2010:
- Sun's (and Fujitsu's) SPARC-based CMT families (US-IV, Olympus, Niagara, Rock, etc)
- IBM's Power family (Power4, Power5, Power6, etc)
- AMD/Intel's x86/x64 families:
- Opteron/AMD64 [Egypt, Italy, etc]
- IA-32/EM64T [Nocona, Potomac, Smithfield, Tulsa, etc]
It will be fun to watch. They all have well funded R&D, aggressive rates of innovation, compelling roadmaps, and market/ISV traction. I believe all three horses will be in the race five years from now, but only two will be perceived as the market leaders. Unpredictable market dynamics and execution challenges will likely cause one of the three to stumble and fall behind. But anyone's guess as to which will stumble would be just that - a guess. Intel can survive a $25 billion dollar mistake, and learn from it; and AMD is actually delivering new processors faster than their roadmaps suggest (an amazing feat for a processor design shop)! IBM's roadmap and processor technology look great, but massive CMT could explode and their Cell Processor could turn into the next Itanic for server applications. Sun has Olympus to compete with Power6, and very exciting new yearlings (Niagara and Rock) that could, well, Rock the world soon. Single-threaded deep pipeline performance processors, throughput-oriented massive-CMT chips, and price-efficient desktop/presentation CPUs are all up for grabs. I doubt one horse will win the Triple Crown. Stay tuned.
Of course, OS traction will dictate this to some degree (Solaris, Linux, and Windows64 are all interesting candidates), as will J2EE -vs- .NET adoption and COTS app support. I think that security and efficient/reliable virturalization technology will be key drivers of platform selection in future years.
The one thing we can predict with near certainly is that Itanium (aka: Itanic) is headed to Davy Jones' locker.