Dead hardware provides an opportunity to intimately examine a once-unique, now-trendsetting Arm-based PC: the Microsoft Surface RT.
Way back in June 2012, I covered Microsoft’s launch of its first branded mobile computers, in both x86- (Surface Pro) and Arm-based (Surface RT) flavors; one week later, I compared them against each other. And nearly two years after that, I revealed that I’d bought refurbished models of each for ongoing use (and editorial fodder, of course).
The first-generation Surface Pro is still humming along, closing in on a decade later, at least the last time I checked; I even updated it from Windows 8 to 8.1, and more recently to Windows 10, all in a relatively unmemorable (albeit imperfect) manner. That said, it’s currently in long-term storage, since it’s been replaced for day-to-day use by Surface Pro 4 and LTE-enhanced Pro 5 successors.
The Surface RT (more accurately: “Surface for Windows RT”) system, on the other hand was ready for disassembly.
As I mentioned in my May 2017 coverage, I’d noticed that the Surface RT would no longer respond to power-button-press attempts back in August 2015. I put it on the teardown pile, although I’d periodically pull it back out to fruitlessly see whether or not it had spontaneously resurrected itself. At this point, although Windows RT 8.1 is still covered by Microsoft’s Extended Support for another 2+ years, ongoing software development has effectively ceased (a situation further crippled by the fact that the only apps available came from Microsoft’s own Store, unless you were to jailbreak the system). And although a battery swap brought my Kindle back to life, I don’t think that’s the root of the issue with this particular system (assuming I could even track down a replacement battery for the Surface RT at this point), since it won’t even power up when AC-tethered, and judging from the glow at the end, the AC adapter itself still works just fine:
Instead, I’ve decided to follow through on the teardown I mentioned back in 2017. I’ll begin with some as-usual pre-surgery shots, alongside the obligatory 0.75″ (19.1 mm) diameter U.S. penny for size comparison. The unit itself is 10.81×6.77×0.37 inches (274.60×172.00×9.30 mm) in size, and weighs 1.5 lbs (680.40 grams). Not included (or shown) are the optional touch and type cover keyboards, which still work fine on my first-generation Surface Pro and therefore didn’t go “under the knife.”
Reflected in the 10.6” diagonal widescreen glossy LCD is yours truly, along with his trusty Google Pixel 3a:
Atop the HD (1366×768 pixel) display you’ll see the 720p front camera; its matched-resolution rear-mount twin is shown in the back-cover shot, in the middle of a plastic strip that, every time I see one, is suggestive that there are RF antennas underneath (hold that thought):
Extend the single-position kickstand (with enhanced positional flexibility in the successor generation) to prop the device up, and you can more easily capture a photo of the right-side ports (top-to-bottom): speaker, proprietary digital A/V, full-size USB 2.0, under-kickstand microSDXC, and magnetic power connector:
Their left-side counterparts are the speaker, analog headphone/mic jack, and up/down volume toggle:
Time to dive inside. Step one is to remove that kickstand; extended, a set of product markings on the main chassis are visible underneath (the image also shows the magnetized keyboard connector):
At each hinge connection, there’s a T5 Torx screw that requires removal:
At that point, the kickstand lifts right off, revealing more product markings underneath it:
Here’s a closeup; particularly interesting (IMHO) info includes:
And here’s that aforementioned microSD card, which I’ll be pressing into alternative service going forward:
Now for that plastic strip. This teardown indicates that it should pop right off, but my experience didn’t match. Maybe it’s because mine was a refurb unit, therefore with a plastic strip reinforced with sturdier (adhesive) stuff. Or maybe I’m just a neanderthal.
Here’s an overview of its topside, post-removal, and its underside, along with a closeup of the rear camera (also the ambient light sensor hole alongside it) and one of the microphone openings:
Here are three shots of what the backside looks like after plastic strip removal, including an early glimpse at a bit of the insides. Note, for example, the rear camera and ambient light sensor, along with the MEMS microphones’ rubberized “vents.” No visible antenna yet, but trust me, they’re there:
Next, I had to remove 17 more T5 Torx screws located around the circumference, including one underneath a warranty-busting sticker:
With that mission accomplished, the two halves were now separable. See anything curious, right off the bat?
See that glow in this closer look?
I’d tried one final time to charge up the Surface RT and get it to boot before beginning the disassembly, and this is evidence that my working theory of the battery not being the root cause of the system failure has “legs.” Admittedly, it takes far fewer electrons to illuminate a single miniscule LED versus an entire computer.
In what follows, keep in mind that the battery subsystem is at the back of the unit, with the other half (containing the motherboard, LCD, etc.) at the front. This also means that, for example, when you’re looking at a circuit board on the left side of a particular photo, it correlates with connectors found on the right when you’re viewing the system from the front.
And speaking of which, let’s look more closely at that battery subsystem, which consists of two cells and connects to the remainder of the system via a single flex PCB cable:
Perhaps obviously, it consists of two distinct lithium-ion polymer cells, together delivering 7.4V at a 31.5 Wh rating:
Now for the other half, which at first glance looks somewhat similar (containing two white roughly square regions, this time for the Samsung LTL106AL01-002 display’s backlight) but will turn out to be far more interesting:
Everything from this point forward is held in place with a combination of plastic clips and T3 and T4 Torx screws, the latter of which my iFixit 64-bit driver kit handled with its usual aplomb. First, let’s get that right-side (left in the photo, remember) assembly for the charging connector and microSDXC reader out of there:
Above it is the right-side speaker plus power switch assembly; after removal, the transducer’s access to the outside world is even more evident than before:
And here’s its left-speaker sibling:
Below it is the analog headphone/microphone jack plus volume control assembly:
Below that is what this teardown curiously calls a “third speaker” but I’m pretty sure is a coin cell battery for retaining system settings in the absence of main-battery charge (agree or disagree, readers?):
And at the very bottom is the touchscreen controller PCB. On the top are (among other things) the three Atmel (now Microchip Technology) mXT154 touch controller slave ICs, which mate to an Atmel mXT1386 master touchscreen controller SoC on the other side:
Here’s what’s left over after this particular PCB is removed:
Did you notice that I saved (arguably) the best for last? What a tease, eh? Let’s finally take a look at the motherboard at the top of the assembly along with the post-removal leftovers:
The first thing you might be surprised by is the complete absence of a system fan; then again, though, that’s not so unique for thin-and-light laptop and tablet designs. Perhaps more notable is the complete absence of any sort of (normally hulking big) passive aluminum heat sink in the design. But then you might recall, as I did, that this isn’t an x86 CPU-based system, but a more traditionally power-thrifty (therefore thermally friendly) Arm-based one.
Before we dive inside those Faraday cages, let’s first turn the PCB over:
There’s not much to “write home about” here, but the two array microphone ports are now clearly visible, as are the two antennae alongside them.
My guess is that one of the antennae handles 2.4 GHz Wi-Fi communications, along with Bluetooth, while the other is devoted to the 5 GHz spectrum. Then again, though, perhaps the Bluetooth antenna is PCB-embedded, instead. Thoughts, readers?
Back to the front side; let’s first remove the camera modules (again, with identical 720p resolution):
And now let’s get those Faraday cages off, shall we?
Now revealed (among other things) is the only nod to heat management, a dab of thermal paste. Underneath it, unsurprisingly, is the NVIDIA Tegra 3 SoC, in a quad-core (plus one) Arm Cortex-A9 configuration. Alongside it is a Winbond 25Q32BV 32 Mbit SPI serial flash memory, presumably holding the boot BIOS code.
To one side you’ll see the four Samsung K4B4G0846B 4 Gbit DDR3 SDRAMs together comprising 2 GBytes of system memory. Also visible, for example, is the Marvell Avastar 88W8797 2×2 WLAN/Bluetooth/FM SoC, and its accompanying RF Micro Devices RFFM8200 and RFFM8500 Wi-Fi front end modules.
On the other, you’ll find (for example) the Sandisk 32 GByte flash memory storage module, Texas Instruments’ TPS65911 power management unit (PMU), a Wolfson 8962E stereo audio codec, and Cypress Semiconductor’s CY8C20466A CapSense 8-bit PSoC. Also note the curiously unpopulated “J14” (naming indicative of a connector) site in the midst of this PCB section.
The Surface RT may be no more, but its lineage lives on in Microsoft’s Surface Pro X line, based on more powerful Arm-based silicon co-developed with Qualcomm and capable of running not only native Arm-compiled code but also supporting 32-bit (and coming soon, 64-bit) emulation for x86-compiled software. And of course there’s also Apple’s now-underway migration of its various computer product lines from Intel’s x86 to its own Arm-based SoCs, as a potential proof-of-concept. Thoughts on the Surface RT design, the Surface Pro X successor, or anything else discussed here? As always, sound off in the comments!
This article was originally published on EDN.
—Brian Dipert is Editor-in-Chief of the Edge AI and Vision Alliance, and a Senior Analyst at BDTI and Editor-in-Chief of InsideDSP, the company’s online newsletter.