Skip to content

Software won’t be far behind those 64-bit ARM chips

For three years until August 2013 I spent writing about the semiconductor industry, primarily on The INQUIRER but a couple of other places too.  Despite outward appearances, the semiconductor industry is an extremely interesting one and one that has many bona-fide characters. Since August 2013 I have been lucky enough to be a part of the industry in a small way, working at AMD.

arm_arch64All of that should serve as some context for what follows. One of the things I was covering when writing professionally was 64-bit ARM (technically ARMv8 architecture) processors, chips that will be used in servers. 64-bit ARM processors are already here thanks to Apple’s A7, however building a “server” chip is somewhat harder than one for consumer electronics – the memory controller, the branch prediction units, the validation and a whole heap of other stuff is considerably harder to develop and test for, and thats not even touching on the more rigorous on validation that is required.

One of the tactics that Intel is using to combat these upcoming chips from AMD and a number of other vendors, is to claim that x86 processors such as the Xeon or even AMD’s traditional Opteron units have a huge library of software. It’s a tactic that works well, software is absolutely vital to hardware adoption and it can be a vicious cycle, just look at Intel’s own Itanium processor.

However, 64-bit ARM processors are not having to rely on traditional software vendors to provide the ecosystem needed to power their sales. It is expected that Linux distributions such as Ubuntu, Red Hat Enterprise Linux (RHEL) and SUSE Linux Enterprise Server (SLES) will be operating systems of choice, not Microsoft’s Windows. Being open source community driven distributions – RHEL and SLES significantly use Fedora and OpenSUSE respectively – it allows companies to contribute directly to the operating system including the Linux kernel. Intel knows this and that is why it is one of the biggest contributors to the Linux kernel.

Not only is the operating system an area where companies and individuals contribute to a greater good, or even their own self-serving needs, but projects such as Apache Hadoop and OpenStack are community driven and open source, meaning interested parties can contribute – in a selfish manner – if needed.

The point here is that 64-bit ARM won’t be like 64-bit x86 or any other emergent instruction set. It won’t need Microsoft or any other company to bring the software to the hardware. Work on supporting the the ARMv8 (64-bit) architecture has been going on for the best part of two years and that’s not even taking into account that ARMv8 processors have backwards compatibility with ARMv7 and ARMv6 code – something that Linux and other software vendors have supported for years. Linaro, the non-profit body that oversees a number of projects to bring support for new hardware and new features to Linux software, has been beating the drum for ARMv8 and

It would be foolish to say that 64-bit ARM is a big change, but if Intel is banking on software being the stumbling block to ARM adoption in the datacenter then it will have a real shock next year. Far too many semiconductor companies have put their weight, faith and most importantly balance sheets, behind ARMv8 to succeed that it isn’t leaving the software ecosystem to chance.

The question is just how much market share will Intel cede in the datacenter. It certainly won’t remain at the 90-or-so percent.