Macs on Both ARM and Intel?
February 2nd, 2017As Intel gets later and later in releasing processor upgrades that offer only slight performance boosts from previous processors, the impact on Macs is obvious. New models are late not so much because Apple doesn’t care about Macs, but because of Intel’s roadmap. Sometimes Apple sticks with the previous year’s parts, which incites the critics. Regardless, benchmarks may be just a few percent better than the pervious model.
Indeed, it may take a few years before a Mac offers a significant performance upgrade compared to older gear. So why upgrade? It does appear that folks are sticking with their vintage hardware longer and longer. Indeed, even though Mac sales were up slightly in the last quarter year-over-year, Apple says more than half of the units sold went to Windows switchers. That would appear to mean that most Mac users are content with what they have.
The theory goes that Apple could do another processor switch. They’ve done it twice already; Intel replaced the PowerPC in 2006. Since Apple has managed to improve its custom ARM silicon to roughly match the speed of many Intel notebooks, how well would they do if they attempted to go all out on producing parts that worked on traditional notebooks and desktops, where power requirements are less restrictive?
Of course, a processor switch of that sort, while certainly doable, would have tradeoffs. By using Intel processors, Macs can run Windows natively in Boot Camp, and with most of the performance of a PC via a virtual machine. That is a significant sales pitch for Macs, since they can run both macOS and Windows with great performance. If Apple went to ARM, they’d have to run Intel operating systems in emulation, which would mean a potentially significant performance hit. Do you remember the days of the PowerPC?
I’m not dwelling on forcing developers to switch. Apple already delivers Xcode developer tools that build apps for both Intel and ARM hardware, so changing over may not be near as difficult as it was in previous processor switches.
By and large, however, putting developers and Mac users through such a migration probably wouldn’t make sense at this point. Apple works hard enough just keeping Mac sales relatively flat, and the cost of a migration probably isn’t justified. It’s not the same as the PowerPC, where development pretty much stalled. Intel is still actively improving its processors, even if the pace of those improvements isn’t what some would like.
At the same time, Apple’s skills with exclusive ARM silicon means they can do things that PC companies can’t do with a single processor. The Late 2016 MacBook Pro with Touch Bar features two operating systems and both intel and ARM parts.
So it uses a standard Intel Core chip, the same ones many PC makers use. But the Touch Bar and Touch ID fingerprint sensor is operated by an ARM chip apparently derived from the Apple Watch, the T1, and an operating system apparently derived from watchOS.
Now there’s a published report in Bloomberg that Apple is working on adding more Mac features to the ARM hardware, thus taking some of the load off Intel. If this story is true — and Apple is certainly not going to confirm the news unless there’s a product to announce — the new ARM processor will be known as the T310.
The T310 would reportedly be used to control a Mac’s Power Nap feature, a sleep mode on steroids. While the Mac is in low-power mode, software updates are downloaded, iCloud data is synced, email is downloaded, and other functions are handled without waking the Mac or putting a load on the Intel hardware.
So if there’s less load on the Intel hardware, battery life could be extended some, at least when the computer is idle. Very likely other hardware functions could also be passed on to the more efficient ARM processors and OS, which may maximize performance potential in some respects. I suppose there are other advantages that hardware designers could let us in on.
Now I don’t pretend to have the full picture of how this dual-processor/OS setup would work in practice. The Touch Bar may be just the first step of a long-range plan that would add extra functions in a way that couldn’t be realized near as well with Intel hardware alone. Since Apple develops its own chips and mobile OS, PC makers would have no way to compete except in a clumsy fashion. It’s not that Dell, HP or Lenovo are likely to consider building their own custom processors. It’s not that they couldn’t hire the expertise to do so. Perhaps they could poach a few Apple processor designers to get started, although it would have to be done in a way that doesn’t infringe on Apple’s many A-series chip patents.
Even if it could be accomplished, it would take years to deliver shipping products. Maybe Microsoft would try, but it’s efforts at mobile hardware with ARM silicon have been total failures. The rest of the PC industry consists of companies that build me-too hardware based on generic components. But if the Bloomberg story is true, Apple may be taking the Mac into uncharted waters in a way that vindicates its ongoing commitment to the platform..
Print This Post
When the 2016 MacBook Pros came out with the new Touch Bar, I thought it was an innovative way of using the Function keys area and adding Touch ID. I do use those keys for other functions. I read articles about how it works, do an OK and moved on. But the latest Bloomberg piece made me step back and look at the bigger picture. To coordinate the work load of two major architecturally different processors is amazing. If more Intel work get offloaded to ARM, Apple is “thinking differently.”
I do think about the relationship between Apple and Intel. The 2016 15-inch MacBook Pro uses an Intel HD graphics processor. The prior MacBook Pro used the Intel Pro Iris one. When Intel finally released the Skylake chips with the Pro Iris graphics, Apple did not use it. Was it due to performance, cost, or schedule issue? Or a rocky relationship, I believe that Intel dropped Pro Iris for Kaby Lake.
For at least five years, Apple has been working on delivering applications in a compiler intermediate form that can be turned into native code as needed. That means that compliant apps, those compiled with the appropriate compiler settings, could run on ARM or Intel with the decision made at the very last moment. If the power cord is plugged in, they can run on the Intel. If operating on battery power, they can run on ARM. If you just use your computer to browse the web, watch video, edit text and read mail, MacOS might decide to run everything on the ARM processor and not even turn on the Intel processor. If you start editing video, processing audio, compiling packages or something more heavy duty, it would turn on the Intel chip and you’d get full power.
The big implication is less about the Mac than about Intel. One of the problem with Intel is that they have to trade off performance and power, and they do this by providing all sorts of internal modes, clock speed and so on. If the OS could decide only to use the Intel chip for things that really need it, they could build a high performance chip with less concern for its power usage, particularly in mid-level operation. Apple could produce a laptop with good battery life and the full power of a workspace chip.
I’m not sure we’ll see exactly that, but I’m getting a sense that this is where we are going.