The Linux kernel community is moving to finally kill off support for Intel’s Itanium architecture. And this time, it looks like it’s actually going to stick.
A recent discussion on the Linux kernel mailing list, reported by LWN.net, has laid bare the state of ia64 — Intel’s once-ambitious 64-bit processor line that shipped from 2001 to 2021. The architecture has been on life support in the kernel for years. Now, developers are making the case that maintaining it is an active drag on forward progress.
Why Itanium’s Time Is Up
The core argument is straightforward: nobody is running ia64 hardware in production anymore, and the maintenance burden is real. Ard Biesheuvel, a prominent ARM and EFI developer, has been pushing to remove ia64 support, arguing that the architecture’s quirks force other subsystems to carry workarounds and special-case code that slows development for everyone else.
This isn’t a new conversation. Intel discontinued Itanium chip shipments in 2021. HPE, the last major vendor shipping Itanium-based servers (the Integrity line), ended support orders even earlier. The writing has been on the wall for the better part of a decade.
But the kernel community tends to be conservative about removing architecture support. There’s always a lingering question: is someone, somewhere, still depending on this? In ia64’s case, the answer appears to be effectively no — or at least, no one who has spoken up to defend it.
The discussion, as LWN details, centers on the practical costs. Itanium’s instruction set architecture was radically different from x86-64. It used a VLIW (Very Long Instruction Word) design philosophy called EPIC (Explicitly Parallel Instruction Computing), which offloaded scheduling decisions to the compiler rather than the hardware. Clever in theory. A commercial failure in practice. Compilers never got good enough, and x86-64 — AMD’s answer to 64-bit computing — ate Itanium’s lunch.
That architectural oddness means ia64-specific code paths litter the kernel. Memory management, interrupt handling, EFI boot support — all carry ia64 accommodations. Removing the architecture would let maintainers simplify these subsystems significantly.
The Broader Pattern of Architecture Cleanup
This fits a broader trend. The kernel has been shedding dead or dying architectures with increasing regularity. Support for several obscure platforms has been removed in recent years, and the community has grown more willing to make these calls. The logic is simple: every line of architecture-specific code that nobody tests is a potential source of bugs and a guaranteed source of maintenance friction.
Linus Torvalds himself has historically been open to removing architectures when they no longer serve a real user base. The bar isn’t zero users — it’s whether anyone is actively maintaining and testing the code. For ia64, the maintainer situation has been thin for years.
Some context on just how dead the platform is. Intel’s last Itanium chip, the 9700 series (code-named Kittson), launched in 2017 on a 32nm process — already ancient by the standards of that year. No major Linux distribution still ships ia64 builds. Debian dropped it. Gentoo dropped it. The toolchain support from GCC remains, but it’s not where anyone is investing effort.
So what happens next? The removal likely won’t land all at once. Kernel developers typically deprecate an architecture first, giving a release cycle or two for any objections to surface. If none materialize — and the consensus from the mailing list discussion suggests none will — the ia64 code gets ripped out. We’re talking about tens of thousands of lines of code that can simply disappear.
For enterprise IT teams, this changes nothing operationally. If you’re somehow still running Itanium iron, you’ve been frozen on older kernels for a long time already. But for kernel developers and downstream distribution maintainers, it’s a meaningful reduction in complexity. Fewer architectures means faster development, cleaner abstractions, and less cognitive overhead when modifying core subsystems.
One less ghost in the machine.
The full technical discussion is available to subscribers at LWN.net, which continues to be the most authoritative source for kernel development news. If you’re tracking Linux internals, it’s worth reading the thread in full — the back-and-forth between developers reveals just how much hidden cost a defunct architecture imposes on a living codebase.
WebProNews is an iEntry Publication