Microsoft’s Big Bets on ARM are Really About Intel (Premium)

In recent months, Microsoft has embraced the energy-efficient ARM microprocessor platform on the desktop and in the datacenter. But ARM offers no notable advantages over Intel on these platforms when it comes to real world use.

Which leads me to an obvious conclusion: These moves are really about Microsoft pressuring Intel to be better in ways that matter to Microsoft, and, collectively, the industry.

And on that note, Microsoft is almost certainly doing the right thing for all of us here. After all, what company knows better than Microsoft about the negative impact that a monopoly can have on that business? Losing your competitive edge is only the tip of that particular metaphorical iceberg. Less choice, higher prices, and slowing development are natural byproducts as well. And that is all absolutely happening to Intel right now.

The timing of these moves is interesting, however.

For decades, AMD served as a competitive counter to Intel, and while they’ve had some ups and some (big) downs, that smaller company is now delivering well-regarded competition on the desktop and in servers/datacenter. But AMD is arguably bouncing off the bottom as I write this, and Microsoft’s ARM initiatives were likely in the works for quite some time. Meaning that they began when AMD was in a particularly bad spot.

But here’s the most important thing to remember about Microsoft/ARM, the one constant that is true for both PCs and servers: ARM only offers advantages over Intel in narrow use cases that do not benefit the vast majority. And Intel has major benefits over ARM that, frankly, benefit virtually all use cases. This is, in fact, a one-sided fight, and Intel is on top.

So let’s examine why Microsoft chose ARM in both scenarios, using their own public explanations. And then see whether this answers any real needs.

Microsoft in December announced that it was bringing full Windows 10 to “truly mobile, power efficient, always-connected cellular [ARM-based] PCs” in partnership with Qualcomm, starting with that firm’s Snapdragon 835 SoC.

But how does this processor compare to Intel’s mobile offerings on the PC?

Poorly.

From a performance perspective, the 835 will almost certainly be Atom-like, falling below even the performance of Intel Core m/Y-series processors. Which I think we all realize is not in any way enticing.

But that “truly mobile [and] power efficient” bit is interesting, too. Few would argue that Qualcomm-based phones deliver great mobility, in the sense that these devices can provide great battery life and are small and easy to carry. But PCs are more complex and have different computational needs, and they are of course much bigger and heavier than phones. They can also deliver great battery life: A comparatively big and heavy Surface Book with Performance Base provides over 11 hours of life on a charge in real world use in my own tests, for example.

Of course, those PCs have humongous batteries too. But even a super thin and light device like the LG Gram 15 can provide several hours of battery life. Even if an ARM-based laptop with a similar weight could provide twice as much battery life, the resulting performance issues would more than overcome that advantage, and that’s before we even consider the performance hit that Win32 emulation brings. Such a device would almost certainly be a non-starter, especially in the first generation.

And then there’s that “always connected” bit. Because modern ARM processors have evolved alongside the mobile industry, chip makers like Qualcomm have also created and integrated ever-efficient radios that often provide universal, worldwide coverage. And the systems that run on these devices—Android, iOS, whatever—are likewise smart about seamlessly switching between Wi-Fi and cellular networks on the fly.

But the onus of making Windows 10 work efficiently on ARM in a similar manner falls on Microsoft, not Qualcomm. And to date, there is no evidence that they will succeed: Windows has long supported cellular networks, yes, and has offered basic Wi-Fi/cellular switching since Windows 8. But few PCs ship with cellular connectivity, still. That is a market reality, not a problem: Customers who need this feature just tether their phones. It’s like we’re chasing a problem few people really have.

Microsoft’s push to ARM in the datacenter seems even more tenuous. Here, the software giant speaks of very specific workloads that will be more efficient on ARM. Which sounds great until you realize that micromanaging hardware and software in a datacenter to address such concerns is both complex and costly. And in this market, Intel has an even greater advantage over ARM than it does on PCs: Its Xeon D processors, in particular, offer double the performance per watt of the Cavium ThunderX. (Microsoft this week was touting the ThunderX 2.). Not double the performance. Double the performance per watt. Put simply, Intel isn’t just better in some narrow category. It’s just better.

Speaking of narrow, what this relegates ARM to is certain server workloads where overall energy consumption is more important than performance. Do these workloads exist? Sure. But the conversation here is the same as it is on the PC: While there are some niche instances in which “energy efficient” is the only concern, for most people, for most use cases, Intel offers dramatically better performance and is in fact energy efficient enough.

On that note, AMD is in a better position to compete with Intel in both of these markets than is ARM.

And as I’ve noted, I think that’s the heart of the matter: With AMD still limping along, it’s often not part of the conversation. And that’s why Microsoft is propping up ARM now, I think. To ensure that there is, in fact, a conversation. And that conversation is about scaring Intel into believing there are other options.

Microsoft’s ulterior motives here aren’t all that ulterior. And again, if they are successful, Microsoft will initiate some changes that will benefit all of us.

The first benefit is lower prices, and this is true even if you think ARM is a real threat or, if you’re particularly starry-eyed, that ARM could somehow “beat” Intel in either of these markets: Competition can and will drive prices down, period. The more real the threat is, the lower those prices will go, and all the more quickly.

Second, Intel—like Microsoft and so many other tech giants—does its best work when competitors are breathing down its neck. With its unchallenged dominance of both desktop and server computing, Intel is under a lot less pressure to be better across the board, and it will simply move more slowly. And get sloppy.

Which has, of course, happened. Microsoft’s early embrace of Intel’s Skylake platform in Surface Book and Surface Pro 4 had disastrous repercussions. And you have to think that this has played a role in the software giant’s embrace of ARM. Intel screwed over Microsoft with Skylake, leading to such craziness as Surfacegate and a (since scaled-back) pledge to not support older Windows versions on the platform. Think about it.

On the PC, Microsoft’s goal is no less than getting those damned, fingernail-damaging Intel stickers off of our laptops, to restore a sense of competition. And in the datacenter, the goal is no less lofty, to lower prices across the board, which is the real efficiency need in such installations. This is about competition. And we’ll get there whether the competitive threat is real, and driven by market conditions, or fake, and driven by Microsoft.

I welcome Microsoft’s ARM moves on many levels, but if you look back at my article ARM-Based Windows 10 Portable PCs!? Hell Yes!, the real excitement there is about pushing PCs forward more quickly. “I am suddenly very excited for a future in which portable PCs can be true best-of-breed devices that meld the best of the PC with the best of today’s mobile devices, and do so in a way that isn’t burdened by compromise,” I wrote at the time.

Exactly so. And we shouldn’t care which platform sits under those PCs as long as it’s the best available. Intel, Microsoft has served notice. It’s your move.