ARM is all the rage these days. Microsoft has long participated in the ARM ecosystem via its Windows CE-based offerings. Recall that its first real success in the PDA space was with Compaq’s iPaq, which was based on Digital Equipment’s StrongARM processor. And now Windows Phone is focused on ARM processors made by Qualcomm (and, in the future, others). But why bring mainstream Windows to ARM? Most observers suggest this is because Intel can’t seem to make itself competitive in the low-power processor space. Intel, of course, thinks it can. Why else would it have jettisoned StrongArm (which it renamed XScale after acquiring the family from Digital) just as the mobile space was exploding? That Intel has failed to deliver to date is one explanation for Microsoft’s new-found love of ARM. But that isn’t a sufficient reason. A friend who recently joined MIPS pointed out that the majority of Android Tablets sold to date have had MIPS processors in them, not ARM processors! So why didn’t Microsoft just resurrect its support of the MIPS architecture (which was an original target of Windows NT) rather than bless ARM?
It’s all about the ecosystem. The ecosystem around ARM has matured into one that rivals the x86. Even as Intel (or AMD) produces x86 processors that are competitive in the low-power space for Tablets, the ARM ecosystem will march on. And Microsoft wants a piece of the action. With the number of classic PC manufacturers shrinking it wants to make sure that OEMs growing up from the mobile device market (e.g., Nokia and HTC) can extend into the Windows PC space. ARM is the architecture of choice for the larger mobile device manufacturers, and so it is the right target for Microsoft to grow its own ecosystem.
Of course it isn’t just the mobile device manufacturers that Microsoft cares about. Ever since the x86 vanquished all other CISC processors (e.g., the Motorola 68000) and the P6 (Pentium Pro) proved the x86 could keep up with RISC architectures, semiconductor manufacturers have been looking for a way to compete with Intel in general purpose processors. AMD took advantage of a licensing quirk to directly compete in the x86 market. Everyone one else pretty much was out of the game. Gradually both previous industry leaders, such as former #1 (and current #3) semiconductor vendor TI, and innovative newcomers such as Qualcomm and NVIDA adopted the ARM architecture so they could target the growing mobile device market. What about other leaders? Motorola was a major player in the PowerPC RISC processor market. It spun off its semiconductor unit into Freescale who, lo and behold, is now producing the i.MX family of ARM processors. Samsung (#2) produces ARM processors as does Toshiba (#3) . STMicroelectronics (Europe’s #1 semiconductor maker), ARM. All the big boys, with the exception of Intel (and IBM, but they are a special case), are focused on ARM. With so many semiconductor firms focused on ARM a great deal of innovation is bound to happen in that world. Microsoft can’t afford to miss out on the ability to take advantage of that innovation, and so this becomes the other reason for their support of ARM.
What about the current debate over supporting so-called Desktop Applications on ARM with Windows 8? Microsoft could just assume that x86-64 based Tablets will take up the slack for customers, primarily businesses, that want to support Desktop Apps. But this penalizes OEMs who want to focus on ARM. For example, if Nokia wants to stick purely with ARM and also wants to target enterprises then it will be at a significant disadvantage to Dell, HP, Lenovo, and the other big x86 devotees out there who can either stick with x86 or bifurcate their offerings into ARM-based consumer devices and x86-based enterprise devices. Indeed some, like Acer, have already done this with Android/ARM tablets for consumers and x86/Windows 7 Tablets for enterprises. Now Acer has the choice of ARM or x86 for consumer Windows 8 Tablets too.
And what if Intel can’t really make the x86 competitive with ARM? It could always buy MIPS so that it had a competitive architecture it also owned. MIPS was an early leader in RISC processors, and an original target platform for Windows NT, and its architecture remains extremely popular for microcontrollers. Even many of the ARM licensees I mentioned also license and manufacture MIPS processors as part of communications products and various microcontroller offerings. Recent developments such as this $100 Android tablet sporting a MIPS processor show that MIPS and its licensees are capable of playing in the Tablet space. Not only would Intel immediately have a major presence in Android tablets, it no doubt could convince Microsoft to (re-)add MIPS as an architecture upon which Windows runs. (Note that absent a player of Intel’s stature on the semiconductor front, or a Dell or Nokia on the mobile device front, committing to MIPS for general purpose mobile processors I doubt Microsoft will support that architecture. The costs simply outweigh the benefits.)
And so I now add another wrinkle to the discussion about supporting Desktop Applications on ARM (Tablets or otherwise). In an earlier post I said a lot of this had to do with what ISVs are telling Microsoft about their intent to port Desktop Applications to ARM. Now I add the other side of the coin. Do Microsoft’s OEMs, new or existing, care about having Desktop Applications on their ARM devices? Or are they telling Microsoft they are perfectly happy ignoring those applications entirely (e.g., players new to the Windows ecosystem) or addressing them purely with x86 (e.g., existing Wintel players)? That’s what Microsoft is weighing.