Analysis Servers powered by CPUs based on Arm Holdings IP sound like a good idea.
They’re a good chance of being denser, more powerful and less costly to run than the Intel CPUs that handle the overwhelming majority of server workloads (and scoop most of the server CPU profits too).
Because Arm dominates the mobile world, server CPUs using the company’s tech are also a good chance of attracting more developers than can be bothered working on server alternatives like SPARC or POWER. Linux and lots of other important enterprise software already runs on Arm.
Little wonder that Microsoft and Google both expressed an interest in Arm-powered boxen in their clouds. Microsoft even ported Windows Server to Qualcomm’s Centriq silicon. Throw in the fact that hyperscalers of all sorts are cost-hypersensitive operations that know their users won’t care what CPU lies behind SaaS or services consume over APIs, and the world looks ready for Arm-powered servers.
What can we use to hit Intel between the eyes, thinks Qualcomm – a 10nm ARM server chipREAD MORE
Cavium brings the Thunder as a fine alternative for servers, but also well-suited to devices like storage arrays or white-box networking kit. Which is realistic: no top-tier cloud has openly invested in Arm servers, with OVH’s Cavium-powered cloud the most visible such effort. Nor have server-makers rushed to offer hardware. Gigabyte has a server design and NVIDIA has come to play to make sure that HPC users that like Arm can use the GPUs they crave. Yet a glance at supercomputer vendors products shows Arm-powered kit lags machines with Intel inside. Big brand server vendors have nothing on their public price lists.
Why is a seemingly good idea failing to gain traction?
A hyperscale engineer of The Register’s acquaintance suggested Arm needs to do more to foster an ecosytem. The engineer told us that Arm doesn’t offer much assistance to either chip-makers or server-makers. Without that help, it’s hard to make products that big server-buyers will want. Such buyers sweat their assets hard, so without confidence about the minutiae of a CPU they’ll stick with safer options.
Qualcomm’s previously signaled it wants to control costs after piling resources onto its server side from its mobile flagships. The effort required to build an ecosystem around a CPU family is substantial. Intel goes to considerable lengths to make Xeons relevant, cutting custom products for big buyers and ensuring each new generation of its products is tuned to emerging workloads.
Perhaps someone out there will do the same for Arm server silicon. But if Qualcomm really is about to quit the business, who else has the muscle, the bank balance or the will to step in and take on Intel? ®