Intel Delays Itanium Update

Arthur Cole

If anyone is still waiting for Intel to make good on its promises for the Itanium processor, it looks like you'll have to wait a little longer.


Seems that the launch of the newest quad-core version of the chip, the Tukwila, originally expected back in 2007, will be pushed back a couple of months so Intel can boost its memory capabilities with new DDR3 technology. The idea is to provide a "scalable buffered memory: by linking several DDR3 modules to each CPU memory interface, offering greater memory capacity than native memory configurations.


This is essentially the same approach that AMD tried several years ago with the Socket G3 Memory Extender (G3MX) system, says The Inquirer. That design was supposed to replace the FB-DIMM approach with dreams of quadrupling available DDR3 memory. Trouble is, the idea never took off and both AMD and Intel scrapped plans to place any kind of discrete buffer between the CPU memory controller and the memory bank, until now.


Why is Intel still beating the dead Itanium horse, wonders PC Magazine's John Dvorak. The entire technology was oversold prior to its introduction and actually led to the demise of some truly remarkable technologies like the MIPS and DEC Alpha processors, simply because the entire world was about to fall to Itanium. Even Sun was seriously considering moth-balling the SPARC program. Now here it is, 10 years later, and Intel says it's ready to move Itanium out of the high-end boutique niche it's found itself in? Too little, too late.


Still, there are those who say the Itanium should be given its due-Sun for one. The company recently finished up a project with the Itanium Solutions Alliance to port Java Platform, Standard Edition (Java SE) to the Itanium, offering an enterprise-class development environment for Windows and Linux. And companies like Sybase are on board as well, looking at Java SE 6 as the base for running Java code under the Adaptive Server Enterprise 15.0.3 platform.


So is this a case of a battered technology getting off the mat for a knock-out punch? Or is Intel just trying to recoup whatever it can of the billions in development dollars that went into the Itanium before putting it out to pasture? We won't really know until the device actually debuts.


In the meantime, it doesn't look like x86s are going anywhere soon.



Add Comment      Leave a comment on this blog post
Feb 13, 2009 1:02 AM Kirk Shipley Kirk Shipley  says:

The delays in initial Itanium release added cost.  Then when confronted with x64 cpus, HP (who else sells Itanium?) was unwilling or unable to gain market share for the technology by pricing it competetively.

It's still probably the best performer for large SMP systems, but the sticker shock is substantial on small servers.   Combined with slow adoption by software ISVs, the promise of this platform has been sadly eroded.

Perhaps if IBM had stuck by Project Monterey, things might have been different.  But they didn't, and now Power and Itanium are struggling independently to withstand the x64 onslaught.

Sun has already conceded the war, leaving their high-end SPARC work to Fujitsu and focusing on multi-core.

Maybe it's time for IBM and Sun to join HP in using Itanium for their single-threaded performance systems.  Of course, they will face the same issues with binary compatibilty that have caused HP so much grief.

IMHO, all three are faced with grim choices.

Unless something changes, it does look like the Itanium is going to join the ranks of great tech that didn't achieve its potential, like the Alpha, which once ran Windows in x86 emulation mode faster than native cpus could.

Reply

Post a comment

 

 

 

 


(Maximum characters: 1200). You have 1200 characters left.

 

null
null

 

Subscribe to our Newsletters

Sign up now and get the best business technology insights direct to your inbox.