Earlier this month, the computer community learned how late in life Itanium was living. The chip architecture was going to rule the world when HP and Intel first announced it in 1994 as a project called Tahoe. Intel ruled the world with x86 architecture then in the lion's share of PCs. Literally, as in the true definition of lion's share: all of it.
It's taken 25 years, but Intel has called Time's Up for its design it co-created with HP. Intel told its customers that the final order date for Itanium 9700 series processors is January 20 of next year. The final Itanium processor shipments end on July 21, 2021.
Itanium was essential to the HP decision to stop manufacturing HP 3000s. Itanium was going to be the future of all enterprise computing, the company figured right after Y2K. There was not enough money in the R&D budget at HP to fund the redesign of MPE/iX for a new processor. VMS, sure, HP would do that for a market that was four times the size of MPE/iX.
By now HP has split in two and Itanium is nowhere but in the HP Enterprise servers, the ones running VMS and NonStop. HPE says it will support its Itanium-based Integrity servers until 2025. A superior article in the the EE Journal includes this summary.
Itanium’s developers sought a path to much faster processing. Unfortunately, the theory behind Itanium’s development was just plain wrong. While VLIW architectures do appear to work well for specialty processors running well-behaved code, particularly DSPs, they’re just not appropriate for general-purpose applications thrown willy-nilly at server processors.
And so, we remember the fallen.
It's taken awhile to understand the inertia that occupies the energy of the computer industry. HP seems to have a side of itself that learned such lessons more slowly than most enterprises. HP was late to Windows (who needed GUIs?) and got well behind the pack on the Internet (Sun crowed over HP's sluggish pace by the late 1990s.) After creating its own RISC chip in PA-RISC, HP figured that another chip developed along with the creator of the x86 was a slam dunk.
There was a time for dictating the way forward in computing with a new architecture for chips. That time passed well before Y2K. HP hung on for more than a decade in full denial, even as it revved up the ProLiant enterprise servers using x86.
It's not easy to see a clean future in the years beyond 2025 for the companies which are invested in Integrity systems. But no one could see how the PA-RISC servers of the 3000 were going to be anything but a write-off for their users, either. The strength of the operating environment, as well as a long history of efficient computing, gave the 3000 a longer lease on life. Now's the time to see if the HP-UX and NonStop environments are going to make the jump to x86. We're also watching how well they leap and what the HP of the 2020's will say.
Not so long ago, in the months before HP had to do its split-up, the company announced a project to create a computer with a clean slate. It was simply called The Machine, an echo of the hubris from the era that built Itanium after a too-long construction period.
No customer who cleaned house of their MPE/iX investment will feel much vindication at the news of the Itanium demise. Industry wags started to call the architecture the Itanic when the chip fell well short of expectations right off the bat. The lifeboats are out in the water now for the latest survivors of the great HP business server purge. The torpedos were first launched in 2001. VMS futures are now in third party hands. World domination turned out to be a very long shot for the Very Long Instruction Words of Itanium. Investing such faith in a single vendor's vision feels foolish now. During the 1990s, sailing on the Itanic looked like a smart berth indeed.