Even as x86 chipmakers like Intel dream of getting a piece of lucrative smartphone and tablet chip market dominated by ARM Holdings plc licensees, ARM is ready to take the fight to Intel. Already preparing to invade the laptop space, courtesy of Microsoft incoming support with Windows 8, ARM has just taken a major step towards establishing a beachhead on Intel's most fertile and fast growing empire -- the server market.
I. HP Throws Weight Behind ARM
After weeks of rumors, Hewlett-Packard, statistically tied [source] with International Business Machines, Inc. (IBM) as the world's largest server maker in terms of revenue, has announced that it will be deploying ARM-based servers.
HP is a major server supplier to Facebook and Google, the world's two top firms in terms of internet traffic.
Dubbed "Project Moonshot", the new ARM servers will extend HP's veteran ProLiant brand, which has traditionally relied on x86 chips from Intel and AMD. HP says the new ARM servers "for select workloads and applications" will reduce energy costs by 89 percent, reduce server room space by 94 percent, and reduce overall costs by 63 percent (versus equivalent x86-based servers). Those are some pretty incredible numbers, and it remains to be seen how well they hold up in the real world.
As speculated, HP's partner in this endeavor is a fresh-faced startup from Austin, Texas, Calxeda. ARM Holdings maintains a substantial stake in Calxeda and has been working with the startup to design the first server-optimized ARM core design, dubbed the "EnergyCore™ ARM® Cortex™". Little has been revealed about how much processing power these chips pack, but Calxeda's press release reveals that they will allow a razor-thin power budget of 1.5 watts, with a total platform budget of 5 watts.
The first product from HP and Calxeda will be the "HP Redstone" processor-dense server. HP writes:
HP Redstone is designed for testing and proof of concept. It incorporates more than 2,800 servers in a single rack, reducing cabling, switching and the need for peripheral devices, and delivering a 97 percent reduction in complexity. The initial HP Redstone platform is expected to be available in limited volumes to select customers in the first half of next year.
Calxeda's website says that the ARM servers clusters will "fit in your hand." It also boasts that the full server package "packs ten times the performance at the same power, in the same space." However, the target application for this comparison or the x86 processor compared to are not given.
II. Top Linux OS Makers Pledge Support to ARM
Alongside the HP Redstone trial deployment, HP announced two other programs. The first is "The HP Discovery Lab". Essentially an open laboratory where corporate and academic partners can come in and test the new Redstone ARM servers, the first Discovery Lab will be located in Houston, Texas and will open in January. Additional sites are planned for Asia and Europe.
OS-wise HP's biggest partners thus far are RedHat, Inc. (RHT), makers of CentOS, RedHat, and Fedora; and Canonical Ltd., makers of Ubuntu. These OS makers are joining HP, ARM Holdings, and Calxeda in the third announced HP program -- "The HP Pathfinder Program".
Curiously AMD is also joining in this program, though it remains to be seen where its interests lie. AMD recently quietly snuck into the DRAM market and also is a top seller of GPUs (and backer of OpenCL). Alternatively, exploration into ARM CPU designs also seems feasible for AMD, given that the company needs something to keep pace with Intel's process-driven energy efficiency improvements.
The Pathfinder Program will help develop open standards to define the Project Moonshot servers, making it easier for other companies like IBM to jump onboard with future ARM support.
III. The Big Picture: Who Will Win x86 or ARM?
Will ARM displace x86 in the server market? It certainly stands a shot, but several obstacles and unknowns stand in the way of its mass deployment.
One big mystery is whether the server OS counterpart to Windows 8, likely to launch next year, will support ARM processors. Microsoft has already made it clear that Windows 8 will support ARM, but given that until right about now no major firms had ARM server plans, there was no word on Windows Server support. An endorsement by Microsoft could help ARM gain even more ground in the server market.
Looking ahead, likely the biggest short-term obstacle for ARM in the immediate future is Intel's 3D FINFET technology. By adopting a 3D "fin" for the transistor gates, Intel has dramatically cut leakage. Thus while ARM may enjoy certain power efficiency advantages from an architectural perspective, Intel has the ARM chipmakers wholly beat on the process end.
It’s a toss-up as to which company will capture power efficiency crown when the dust settles.
The only other elephant in the closet for ARM servers is application compatibility. X86 binaries/executables can't be run on ARM processors, and vice versa, as they have different instruction sets. While open source applications can simply be recompiled on the target system (meaning most Linux applications should be available), proprietary closed source apps must be recompiled by their author.
It's easy to either understate or overstate what effect compatibility concerns may have on ARM server deployment.
On the one hand, compatibility may certainly keep some specialized businesses from jumping on the ARM train. On the other hand, with Windows 8 supporting ARM the "if you build it, they will come" process is in full effect. Mass market software like Microsoft Office and Adobe Creative Suite, will almost certainly pop up in ARM-compatible form in the next year.
And while ARM may be a bit bigger compatibility shift than most, it's important to remember that scores of similar transitions have already successfully occurred many times before. To give one example, every time there's a new web browser, page makers and web application designers have to readjust their content to play nicely with the new engine. Or to give another example, many versions of Windows have broken compatibility with extremely old legacy applications -- particularly those that require specialized hardware support.
Plenty of Windows releases have been accompanied with dire predictions, "Windows XXXX will never be adopted at this firm as it will break legacy app XXXX."
Sometimes these predictions indeed prove true, and firms reject the upgrade. But the bottom line is that if there's a cost savings to the upgrade, most firms will go for it, perhaps maintaining a handful of legacy systems to run the bits of old incompatible code. The same will likely hold true with ARM, should HP make good on its power efficiency and cost claims.
Calxeda and HP have their work cut out for them, as they have yet to deliver a solid finished ARM server product. Nonetheless, with the announcement of Project Moonshot, ARM has taken a big step forward in its goal of global CPU dominantion. Intel should be concerned, as the project does threaten to at least slightly cut into to its fastest growing business.