When the chips are down

Two months ago Intel launched its long awaited 64-bit Itanium chip without fanfare and two years late. It has yet to become a mass-market product and won't become anything more than a tool for developers until Intel releases the McKinley version of Itanium sometime next year. However, it is already having an impact on the enterprise server market. At that time, two of the market's big guns (and now likely to be one very big gun) - Hewlett-Packard and Compaq - threw their weight behind it.

Both are phasing out their Risc-based processors in favour of Intel's IA-64 architecture leaving only Sun and IBM to carry on with Risc strategies. However, even IBM will embrace IA-64 with its 32/64-bit Summit architecture to be used in its xSeries eServers, effectively giving it one leg on either side of the fence.

While they still see 10 or 15 years left in Risc, IBM executives agree with others in the industry that by 2004/5, Intel's Itanium architecture will play a dominant role in enterprise IT servers.

Intel plans that Itanium will become the first low cost, mass produced, high-end microprocessor and despite the delays, the response to Itanium today has been positive.

Ideas International senior analyst Peter Steggles says the writing is on the wall that there will be fewer and fewer architectures in the future.

He predicts that it will be at least 2004 before Intel achieves market dominance but says it is inevitable that it will happen.

"Two of the four major vendors are committed to that happening so that really only leaves Sun to continue with a plan to develop its own chips, with Texas Instruments as the manufacturer," Steggles says.

He adds that the high cost of developing processors is forcing others out of the market and both HP and Compaq see the commoditisation of the server processor as the way to go, because they can bring the cost down with mass production.

"Risc will probably continue for the next three to four years, possibly longer depending on how long it takes to get applications, compilers and so on completely tuned up for the Itanium processor.

"Sun sees Risc continuing, but HP and Compaq have both firmly stated they won't be continuing and that will only leave IBM that is taking an each-way bet by going on to Itanium but also continuing development of its Power chip."

He says there is a lot of speculation that Sun has painted itself into a corner with its Risc strategy.

"But Sun is not stupid. If it ultimately saw there was value in porting its Solaris operating system on Itanium, it would. It has not made anything public yet, and it is quite right in what it is saying that it's early days yet for Itanium. It will be at least next year before Itanium goes mainstream with the release of McKinley and will take time to work its way into the market," he says.

Both Compaq and HP are allowing themselves two to three years to make the transition, but by 2003 the switch should be complete.

Itanium was developed jointly by Intel and HP, so it is not surprising that HP should announce that it is phasing out its PA-Risc architecture in favour of it.

Itanium is based on a concept called EPIC (explicitly parallel instruction computing). All modern CPUs have some capability of running multiple instructions (low-level commands such as add, multiply, or read from memory) simul taneously. Most CPUs analyse software on the fly, looking for opportunities to process instructions in parallel but EPIC shifts responsibility for this analysis from the CPU hardware to the programming language compiler used to create the application.

This is where Compaq claims it will play a major role in the future development of Itanium. It has sold its Alpha technology - in particular the Alpha compiler - to Intel for integration into the IA-64 architecture. In doing so it has allowed Compaq to ease out of the expensive 64-bit processor arena, where it has been since taking over Digital and inheriting the Alpha chip. Compaq saw the deal as win/win - it no longer had to pour money into a seemingly bottomless pit developing better and faster microprocessors; it got a healthy payout from Intel; and it had time to move its existing customers over to IA-64 as well as develop its own Itanium products.

Manager of Compaq Australia's business-critical server group Rob MacLean says the company has no intention of letting Alpha die until 2004. It has been around since 1992 and there are still two new chips scheduled for release before then - and that was what surprised the industry when Compaq announced in June that it was joining forces with Intel.

McLean says the decision had only been made three weeks earlier and the technical details have still not been worked out; however, Compaq decided it was better to get it out in the open than to let the industry speculate for the next 12 months.

Compaq will proceed with plans to release its EV-7 and EV-79 chips over the next two years; however, a third new chip planned for release in 2004 has been abandoned.

"While Alpha was the fastest chip available and Compaq could maintain that position over the next few years, it looked at what the position was likely to be in 2004 and found that it was reasonable to expect that Intel would have its Itanium chip relatively competitive by then," MacLean said. "While we would still have had an advantage it would be whittled away over time and the question was how much would it cost to maintain the competitive advantage? Did Compaq want to be a solution company, a chipmaker or a solutions and chipmaker?

"So we thought if we sell our important stuff to Intel to be integrated into the Itanium then the Itanium would effectively be where we expected the Alpha to be in 2004. Not only that, we would not have the expense of developing chips," he said.

The ‘important stuff' in this case, according to MacLean, was the Alpha compiler.

But HP sales and marketing manager for Asia Pacific Peter Hall disputes that. He says Intel already had perfectly good compiler technology thanks to its joint venture with HP and that the decision to the deal with Compaq was just a way of taking the Alpha out of the market.

Hall says HP's road map for the future is very clear - Itanium is the next generation of 64-bit technology. He believes EPIC has significant advantages over Risc, in particular its ability to anticipate calculations in advance.

"It's like you waiting in the queue at the bank and the tellers see you coming and have both a deposit and a withdrawal slip waiting for you, knowing that the likelihood is that you will need either one or the other and all they have to do is fill in the amount."

But IBM's eServer xSeries brand manager Alex Yost believes there will still be a place for Risc for some time to come, particularly in the Unix space. He says despite Risc having been around for a decade it could have 10 to 15 more years and IBM would continue to develop its Risc-based Power architecture, with Power4 due out later this year.

However, at the same time he agreed that Itanium would become a major force over the next few years and IBM was adopting IA-64 and integrating into its 32/64 Summit architecture, which had been in use for 32-bit applications for more than a year. Summit will become IBM's Windows and Linux enterprise server platform.

With three of the four major players committing to Itanium, there has been speculation that Sun's Risc-based Sparc architecture could be in jeopardy, particularly as it appears to be having trouble advancing Sparc while Intel has been able to come out with a faster processor at its first attempt in the 64-bit space.

Sun national sales product manager John Fennell says nobody is arguing that Intel has a faster processor but it has to be put into context.

"It is not a case of speed alone. You need a properly balanced system," Fennel said. Our argument is that because we have been doing it for such a long time, we really know how to create large multiprocessor servers, but Intel is starting from scratch. While it has been focusing on chip speed it hasn't been focusing on how it put the chip into large machines with multiple CPUs."

Fennell says Risc has at least another generation because the advantages that made it popular in the past are still valid today. He says Sun has a five-year road map and it will progress its roadmap as the market moves, but he doubts that Sun will port its Solaris operating system to Itanium.

Whatever Sun does it seems clear that the market dynamics will change over the next three to five years, with Intel becoming a dominant player in the high-end enterprise server area. Despite its late arrival and underplayed launch those who have used Itanium say it is solid and the EPIC technology is impressive. As a result while Risc will be around for some years, it will no longer be the force it has been in the market.

Join the newsletter!


Sign up to gain exclusive access to email subscriptions, event invitations, competitions, giveaways, and much more.

Membership is free, and your security and privacy remain protected. View our privacy policy before signing up.

Error: Please check your email address.

More about CompaqHewlett-Packard AustraliaIBM AustraliaIdeas InternationalIntelTexas Instruments Australia

Show Comments