This week, AMD announced it had formed an alliance with ARM to create the next generation of cloud servers. This announcement has been expected since Rory Read became CEO of AMD, and it now becomes the showcase initiative for his strategic vision. His vision is to stop chasing Intel and to help drive a new generation of blended solutions that would put Intel and the other ARM vendors on different paths. AMD is carving out its own space, which is risky but far less hopeless than trying to catch a much better-funded Intel from behind or trying to catch NVIDIA on tablets or Qualcomm on smartphones.
Let’s talk about 2014--the year AMD plans to shake up the server market with ARM technology.
However, first let me share an interesting Microsoft back-story. A few years back, Microsoft came up with this idea of the data center as a server with processors that function the way RAID arrays do in storage. Essentially, you’d build a data center with networked parts, and rather than having individual servers, you’d have a mass of inexpensive processor cores optically connected and dynamically managed, so that the data center itself became one massive server which could handle constantly changing virtual loads.
Microsoft's idea was that in an age of virtualized servers, you don't need individual servers. Instead, you basically could have a lot of parts which could be connected dynamically to provide the point performance you needed. You could then, at a very granular level, apply just the right resource to the job. And because the whole thing was built with inexpensive processors with lots of cores, you’d have the ultimate flexibility at the lowest possible price (at least with current processor technology).
The proof of concept was done with Intel Atom processors, but to work in servers this way, these processors needed Xeon-like technology. Intel presumably didn’t want to cannibalize its Xeon margins, so it tried to push Microsoft into their more expensive lines. Instead, Microsoft switched to ARM. Some say this move may have set the groundwork for Microsoft’s lack of interest in Intel’s ultra-mobile parts for smartphones and Microsoft's use of ARM in Windows RT.
Microsoft apparently is continuing to look at this ARM-based solution. All it needs is a processor partner who understands servers and whose brand can be considered enterprise-ready. Now, we have the potential for them to do to their server platforms what they just did with PCs and shift much of their back office efforts to ARM, with AMD (now that AMD is doing ARM) being the likely vendor to benefit.
As we move to more and more mobile dependency, the apps that these devices use are thinner and thinner. Given that an increasing number of these devices are always connected, the chance to provide a new client-to-server end-to-end solution has never been greater. However, a mass switch from x86 to ARM would take a long time given how slowly this market moves and how quickly ARM technology moves. As a result, the solution would need to be blended and able to handle both traditional loads as well as new loads from mobile devices.
When you add to this the GPU computing trend, you get only one vendor with the capability to address the legacy x86 market, the emerging ARM server market, and the growing need for massively multi-threaded and increasingly intelligent GPU computing solutions.
While x86 to ARM is certainly interesting, the ability to mix and match cores with a GPU component could create the most flexible servers in a market screaming for flexibility. That could immediately shift AMD from a follower to a leader. Intel divested itself from ARM (it still has an ARM license but is tightly committed to x86) when it sold Strong ARM to Marvell, and the firm has been light on GPU computing technology. The other ARM vendors don’t have x86 licenses and currently aren’t positioned well for workhorse servers, which are AMD’s bread and butter.
Of course, Intel won’t be standing by, but in this next battle, AMD will have the combined resources of ARM and Global Foundries to supplement its own, plus stronger Microsoft backing, which should offset Intel’s historic advantages. In short, for once Intel won’t be the assured victor. That is critical to AMD’s ability to emerge from under Intel’s cloud.
In 2014, AMD plans to start shipping its ARM server solution, which will heavily leverage its Sea Micro acquisition, and AMD apparently has firms like Dell and HP waiting eagerly for it. The key isn’t the fact that AMD is moving in this direction, but the fact that it is the only vendor with ARM capability, x86 capability, GPU capability and a ton of server experience. That gives it a chance of creating a market it can defend better than anyone else can currently attack.
If it succeeds, it won't mean just a new AMD but also a new class of highly flexible servers which likely will have Microsoft as a core backer. Apparently, Rory Read and his team are kicking a bit of butt over at AMD, and that effort is starting to bear fruit.