« 3000s continue to fly free: but how many? | Main | HP 3000 work still surfaces, on contract »

October 08, 2009

Itanium: Failing HP-UX futures, or more?

ItaniumRising We take it on faith today that Intel produces most of the world's popular processors. Even Apple, once a Motorola and IBM POWER stronghold, now uses Intel chips in Macs. But the HP 3000 never got a real chance at having Intel Inside. Now that 3000 emulators are in the works and testing soon, it looks like skipping over Intel's Itanium might be a good thing for MPE/iX users.

This might come as heresy to the 3000 advocates who lobbied HP long and hard for a shot at 64-bit processing, the Valhalla of the journey via Itanium. But look at what HP-UX customers got for their waiting -- including the 3000 sites that migrated sooner than later -- and you can wonder if the delays were worth it. The 3000, and MPE/iX apps, are now more likely to find a future on an mainstream Intel chip.

This matters now, in the gray time of HP's Unix system migration. PA-RISC is old tech, but it's running a large share of the migrated 3000 sites. The Itanium failure to dominate relegated HP-UX to a niche market, a place HP couldn't imagine setting up shop. The 3000 was supposed to be the small market, even if HP didn't say so while Itanium was so new it was called Merced.

Since Hewlett-Packard plowed its engineering into Itanium, HP's Unix customers cannot host their applications on a standard computer, something HP sells very well (think ProLiants, and Linux or Windows). These Industry Standard Servers, as HP calls them, are so strong that HP is thinking of folding its printer business into a combined PC-printer organization. This would offer little help to HP-UX customers. The merger is supposed to jump-start HP's printer sales.

Back in the 90s, HP trumpeted vast plans for the chip that now represents the Only Home for HP's Unix. Then the market had its say. One PC columnist, whose last name is the same as a failed keyboard layout, asserts that Itanium hobbled more than HP-UX options, since it failed to live up to its promise. John Dvorak says the chip killed the computer industry.

Now Dvorak has been a splashy computer writer for a long time, which can boost a fella's readership nicely. (It helps to be published by PC Magazine, which recently dropped printing altogether to retreat to the Web.) Dvorak told his version of Itanium history this year as a cautionary tale. He reminded us that promises of world domination by any technology should be viewed as fables until the future arrives.

His column does a good job of summarizing the hubris of Itanium, nee-Merced-nee-Tahoe, a flight plan HP cooked up in its top-notch Labs but had to take on Intel as a co-pilot in order to fly. The flight of Itanium was as anticipated as any Spruce Goose test run. HP told all of its customers to expect all other chip architectures to evaporate. Who could take on the industry clout of Intel and the brainpower of HP's Very Long Instruction Word designs? And so by degrees we lost the Alpha, the SPARC, and more. Computers made by Dell, IBM and Sun would be powered with chips created by HP and Intel.

I've covered Itanium since these two companies were calling their joint project Tahoe in 1994, then naming the chip architecture Merced in '95. By '96, the 3000 community was eager to learn what Hewlett-Packard would decide about including the HP 3000 in the world domination party. Early in '97, the 3000 customers were told, in a special TV teleconference, that they weren't invited to the 64-bit party.

PA-RISC, said HP in 1997, provided plenty of processor for the 3000s future. As it turned out, HP sold PA-RISC to all of its MPE and Unix customers for another 6-11 years. We wrote in 1997:

[HP] indicates a long lifespan for the 64-bit processor that now powers the 3000. Remember, Merced still isn't a tested solution anywhere, and few expect it to be available before 1999 in HP's processors. What's more, HP still hasn't shipped PA-8200 chips in either HP 9000 or HP 3000 systems. There's a lot of PA-RISC lifetime still left to live.

Only in 2007 did the number of HP-UX servers sold for Itanium/Integrity pass the sales of PA-RISC computers. HP stopped selling PA-RISC last year, 14 years after it crowed about Itanium ruling the marketplace.

Dvorak says that the high-water mark of the computer industry was 2000, and he adds that Itanium pulled the business into the basement in the years since then. It doesn't look like he's accounting for the Y2K swell that put your community at its crest. But he's right about one thing: The chip that hosts the future of HP-UX, the one that will give those users processor headroom for years to come, never came close to the $38 billion it was supposed to earn way back in 2001.

HP and Intel were late, over and over, in delivering something to beat PA-RISC. Hewlett-Packard was hoping for a repeat of the miracle of MPE. HP rolled out PA-RISC in 1987 and the 3000 apps written for 16-bit CISC processors ran in Emulation Mode on the new chips. That's why an emulator for the 3000 hardware will have traction and generate sales for a company that makes it available. Emulators have a good track record with 3000 enterprise customers.

What better not happen: A series of big promises and Itanium-like delays for these hardware emulators. That's why nobody, not Stromasys or Strobe Data or anybody, is promising when the emulator solution will be ready. It's worse to miss a milestone than to release no schedule. People budget for products months and years in advance. Changing your mind is often expensive, and IT expenses remain on many chopping blocks.

Itanium has carved a niche for some apps, so it's not an utter failure. It provides the fastest engine for HP-UX, although there's no chip even racing in second place. No amount of cheery industry measurements can pull the only current HP-UX processor into the mainstream market. Such a market is important to a future without costly changes. HP 3000 owners have learned that business practice from Hewlett-Packard. Sales and market share make at difference at HP. Perhaps any project to emulate PA-RISC on industry standard Intel chips will have an even bigger set of customers: HP-UX sites looking for a longer future for their PA-RISC investments.

08:11 PM in History, Homesteading, Migration, News Outta HP | Permalink

Bookmark and Share

Use our search engine to find 20 years
of HP 3000 news and articles



The comments to this entry are closed.