I guess this means that it is ending 15 years early?

From this article one gets the impression that Windows will not be supporting Itanium anymore.
Way back during the initial marketing onslaught of Itanium, it was said to be the architecture for the next 25 years for Intel. That was a decade ago. It seems to be losing software support fairly rapidly though. Its hard to see this lasting another 15 years … let alone 5 years.
Linux still has Itanium support for now, but fewer users of it are out there. Important subsystems (like accelerated video drivers) aren’t being built for it anymore … there is no real market for them, on Itanium.
We still have an Itanium2 box in the lab. Haven’t turned it on in more than a year. We no longer have any customers with these systems. Not that we have lost customers, just that the customers have thrown away these systems.

This is, IMO, a good decision on the part of Microsoft. Itanium wasn’t and isn’t the future of processors. They realized that they wouldn’t make back their investments going forward. I suspect they will be tightening other areas up soon as well … if your business is only loss making, you shouldn’t expect it to continue as a going concern. Which opens the question of Windows HPC’s future. Too small of a market for them with insufficient ROI. How long will they continue to push this? Took them 10 years to finally give up on Itanium. Maybe we are at the halfway point now. Who knows.

3 thoughts on “I guess this means that it is ending 15 years early?”

  1. I agree, and our customers do, on Itanium. We continue to support it only to support existing customers who made serious investments in Altix several years ago. But we’re seeing the opposite with Windows HPC. We expected to see interest only among small customers with no Linux expertise. To our surprise, we’re seeing interest in large enterprises with considerable installed Linux infrastructure. Even they have more Widows than Linux sysadmins, allowing more flexibility in maintenance assignments, and even their Linux admins are impressed with ease of setup & administration. Our benchmarks are just as good as on Linux. You’re correct that the market is small for Microsoft, but I suppose they are banking on the commoditization of HPC.

  2. It’s a bit of a shame, really. The Itanium memory architecture helped some sparse matrix codes run at >90% CPU utilization compared with maybe 40% on IA32 (and AMD64) and Power. Not all such codes, but ones based on physical models tended to run like gangbusters with a little autotuning. Performance on day-to-day tasks is abysmal relative to the IA32 line, and eventually Intel Intel’ed itself.
    As far as Windows HPC, well, a better POSIX layer would render the rest of the user-level issues moot. Codes that already have to move between Linux and AIX (and ex-Solaris, ex-Irix, etc.) often are sufficiently portable to handle any sensible POSIX layer. I haven’t seen such a layer for Windows, however, particularly if you like using mmap to share data between MPI processes on the same physical node…

  3. Red Hat announced back in December that they were dropping support for Itanium in RHEL 6, so RH will only officially support IA-64 until 2014 now (though apparently some OEMs will offer support to 2017).

Comments are closed.