There was a report last week at the “The Inquirer” that gave us a little insight into the prospects for the future arc of Intel’s Itanium server chip line.
For the past two years with the advent of the Opteron line of server chips that offer 32 and 64-bit native capability for the X86 code base, there have been loud rumbles about where Itanium now fits in the CPU landscape. And even Intel has acknowledged that Opteron has created problems for Itanium in the market place as reported in September 2004 saying "I would be remiss to say the impact was zero, but the impact was mostly noise and confusion," Talwalkar (
Abhi Talwalkar, general manager of Intel's Enterprise Platform Group) said of the decision, referred to as EM64T. "It set us back a few months, I think, with the general audience."
Yesterday’s report from The Inquirer adds a little resolution to the picture. It reports that at the ceremony where the Intel CEO reins were handed from Craig Barret to Paul Otellini, “A former employee asked what Intel's plans were for the Itanium. Paul Otellini said it was a RISC replacement processor, typically running Linux. There are a number of companies who run SAP on Windows where the Itanium fits into the infrastructure nicely. Intel wasn't walking away from the Itanium, he said.”
Itanium is running Windows in some cases but it’s “typically” running Linux according to Intel. Though there are some Itanium servers running Windows, Windows on Itanium is a dead-end path as Microsoft pointed out in 2004 saying then that Itanium just doesn’t run the Microsoft stack very well and it never will. I guess growing up on X86, as Microsoft has, will do that to a code base.
If Itanium’ s future is pinned mainly to Linux, that says an awful lot. It’s probably not likely that Itanium has much future unless Linux makes massive gains in market share over the next 5 years. While that’s entirely possible, Itanium will still be just one of many architectures chasing a share of the market and it will be starting from a price/performance point that is not nearly as attractive as AMD’s Opteron for nearly all but a very few number crunching applications and even those are looking very challenged with the recent dual core Opterons from AMD.
Taking the more reasonable route that Linux will continue to along its growth curve in the server, cluster, workstation space and take somewhat longer to reach the desktop and laptop client space, then you have to wonder even more loudly about the future of Itanium. Can an entire architecture survive on a very small fraction of the Linux server market? Probably not but that’s a bridge we will cross when the time comes.
The most interesting aspect of this story to me is that Linux is nearly single handedly providing life support to the Itanium. If Itanium ‘typically’ ships with Linux, what would Itanium be shipping with if Linux were not around? Do you think this was Intel’s plan when it was rallying Sun, HP, IBM, Fujitsu, and the rest to follow it down the Itanium road in 1996 and 1997? I think it’s safe to say ‘no’ to this question.
All along, Itanium’ s biggest problem has been software. It’s enough of a different architecture that traditional X86 code runs poorly on it even after some massaging. And it presents enough of a programming challenge for software folks that not a lot serious porting activity has taken place or what activity has taken place has run into problems and had to cut bait. Microsoft went after a Windows product for Itanium but it just did not work out as X86 performance continued to scream upward in the late 1990s breathing new life into Windows on the client and now, in the server room.
As usual, in rides Linux. Our little Linux brings with it top-shelf credentials for being easily ported and for being a good performer. Linux, you would think, is not the enterprise match for Itanium that Intel had in mind in 1996. The proprietary Unix operating systems from Sun and others are probably better aligned with who Intel thinks Itanium should be used for but one by one, all the other players have dropped away. Linux is the only player that can’t leave the table and so it remains, as a crutch for Intel’s 64-bit server chip.
An interesting aside, if Intel truly believes in Itanium, then it has to do away with Windows. Windows is not coming to Itanium. While killing off Windows is probably a pipe dream, even for a company with the resources of Intel, that’s what would need to happen to bring the industry to the point where Itanium is running the most widely used code base. But that’s not really in Intel’s plan either. The whole reason for Itanium, aside from moving the industry to a 64-bit platform, was to eliminate competition in the x86 CPU space by eliminating x86. To eliminate x86, Intel needed Microsoft to embrace Itanium fully at which point Intel could move the industry to Itanium as volumes increased and prices decreased. Intel would have been in a nice position of having an entire CPU space all to itself and as long as it maintained a value proposition that would steer the x86 code base toward legacy status. To do this however, Intel needed Windows. With Windows, Intel has a proprietary code base that, eventually, only runs and gets maintained on Itanium. Without Windows, Itanium is running on Linux and Linux runs on anything which means Itanium is one of many instead of THE one.
AMD’s Opteron and 64-bit client chips effectively took X86 Windows into the 64-bit space in a way Windows never would have with Itanium. Windows performance in 64-bit mode is generally 5-10% better and in some cases is over 100% better where a large memory footprint is needed, like in terminal services and large databases. Just as importantly, application developers can now take advantage of huge memory spaces and this should lead to nice gains in the next few years as application software catches up with hardware and now OS.
So, I put it to Intel this way: if Linux is the crutch propping up Itanium, let’s get after it and put some major development dollars into making Linux an unstoppable force on the client and on the server. Oh, but wait, if you do that, then it will be an open playing field where the best CPU will have a chance to win because … Linux runs on everything.
In the end, I don’t think Intel has any hope here of winning with its Itanium. Top shelf X86 is where the lion’s share of engineering focus is happening on design and process technologies. X86 is probably where the future is for at least the next 5 years. Linux will open up the playing field in the CPU space when it becomes the dominant code base and at that time, Microsoft will begin to start looking more like Apple does today than anything else – a quirky old OS for users that don’t mind a lot of hassles, a difficult to maintain product, and like something different – what a change that will be!
But in the meantime, Intel should decide if being a bit player in a niche server market with no hope of proprietary control of the OS is really the road it wants to travel. I would say no, Intel says yes. I am betting it will change its mind before I change mine.
Subscribe to Linux Email News Alerts
Subscribe via RSS
