Surface Pro upgrade offers more memory plus LTE connectivity

Article By : Brian Dipert

Insufficient memory, coupled with a longstanding desire to be a mobile computing warrior, fuel yet another system upgrade.

In a recent post, I discussed my then-newly purchased (albeit factory-refurbished) Microsoft Surface Pro 4 in comparison to its first-generation Surface Pro precursor, specifically in the context of ongoing microprocessor architecture (and foundation manufacturing process) evolution and their positive impacts.

The Intel Core m3-6Y30 CPU in the Surface Pro 4 was (and remains) perfectly adequate for my relatively modest needs, with the added benefit of low power consumption that (along with clever “hybrid liquid cooling system” design by Microsoft) results in a completely fanless, therefore completely silent setup. However, I closed out my prior coverage with the comment, “my first stab at a newer Surface Pro system ended up having at least one notable shortcoming, which I discovered shortly after I began using it.” The details on that shortcoming, and how I resolved it, are the subject of this follow-on piece.

Among my daily routines is an every-morning simultaneous launch of 28 Firefox browser tabs to peruse the day’s online comic strips (yes, I know, I’m a child at heart). The first time I did this on the Surface Pro 4, it brought the system to its extremely sluggish knees. See for yourself:

Surface Core M CPU usage

Granted, part of the problem was that I was used to running Firefox for MacOS, which is dramatically more performance- and power consumption-optimized (beginning with v70) by virtue of its heavy leverage of the GPU as a co-processor, for example. And part of the problem was that I’d initially neglected to install and enable the WebControl add-on for Firefox, which when activated enables me to kill background scripts (several dozen per tab, to be precise, as “free” comics compensate by serving up lots of ads and other cruft it turns out).

But even with WebControl turned on and doing its thing, the system performance was still quite sub-par. My gut feeling was that the CPU wasn’t to blame, and further analysis bolstered my theories. Look closely again at the above graphic. Yes, the CPU is maxed out, but so too is system memory usage. That undesirable scenario meant that Windows 10 was frantically frequently transferring data into and out of the “swap space” sitting on the SSD; much faster than if I’d had a HDD, granted (at least for reads), but still quite a hamper on performance.

The entry-level Core m3-6Y30 version of the Surface Pro 4 only comes with 4 GBytes of DRAM, and because the memory is soldered down on the motherboard, aftermarket capacity updates aren’t possible. That meant I’d need to upgrade to a different model with more DRAM built in in order to alleviate the memory bottleneck. The higher-end Intel Core i5-6300U- and Core i7-6650U-based variants of the Surface Pro 4 come in 8 and 16 GByte DRAM options, but unfortunately neither is fanless, which was a deal-breaker for me.

Instead, I turned to a refurbished 5th generation Surface Pro (which Microsoft oddly only officially referred to as the “Surface Pro;” it’s also commonly colloquially referred to as the “Surface Pro 2017“). Microsoft based the “Surface Pro 5” on Intel “Kaby Lake” processors, one generation newer than the “Skylake” CPUs in the Surface Pro 4. Both processor generations leverage Intel’s 14 nm lithography, but Kaby Lake’s process tweaks coupled with architecture enhancements (along with Microsoft’s aforementioned “hybrid liquid cooling system” innovations) enable the Core i5-7300U variant of the Surface Pro 5 to also be fanless.

Here’s that same 28-browser-tab experiment run on my Core i5- and 8 GByte DRAM-equipped Surface Pro 5:

Surface Core i5 CPU usage

Once again the CPU is maxed out (and ironically, the Core i5-7300U offers 25% less aggregate L3 cache than its Core m3-6Y30 predecessor). But note that now system memory is “only” 85% utilized. Trust me when I tell you that the resultant absence of “swap space” delays makes a BIG difference in overall load time, tab-to-tab switch latency, and other factors.

I also took advantage of the system upgrade to make one further feature set uptick. The Surface Pro 5 came in a variant with an integrated LTE Advanced modem (which is still available for purchase to this day, in fact). I’d been paying AT&T for an unlimited-data 3G (later 4G LTE) plan ever since June 2010, when I got my cellular data-supportive first-generation iPad. Frankly, I’d rarely used even a modicum of the amount of monthly data needed to rationalize the “unlimited” tier, and then only when I was surreptitiously leveraging jailbroken tablets as Wi-Fi hotspots. But AT&T no longer offers that particular plan; although I was “grandfathered” in, once I canceled it I could never get it back. So I soldiered on.

On a laptop, on the other hand, a hefty monthly data allocation makes much more sense. When I took the AT&T LTE nano SIM out of my 9.7″ iPad Pro and popped it into the Surface Pro 5, the cellular connectivity option popped right up:

AT&T cellular

Here’s the new system in action at one of my favorite breakfast stops, 20th Street Cafe in downtown Denver (which also has free customer Wi-Fi, but where’s the fun in that?):

20th St cafe

Surface Pro 5 LTE

The LTE-equipped Surface Pro 5, like my jailbroken iPads of days past, can also act as a Wi-Fi hotspot for other devices, thanks to software such as that sold by Connectivity. And standard and LTE-equipped systems are visually virtually indistinguishable. The thin, very slightly off-color polycarbonate strip across most of the top of my Wi-Fi-only system (seen on top) extends all the way around the top on the LTE-enhanced model to comprehend the additional antennas (internal images of which you can see here):

Surface Pro polycarbonate strips

Surface Pro LTE polycarbonate strip

Next to the microSD card slot underneath the kickstand on the Wi-Fi-only model is the nano SIM slot for the LTE-enhanced model:

Surface Pro 4 microSD

Surface Pro 5 nano SIM

Consider me quite pleased with the end result, no matter that it took me two purchases to get there! Questions or other thoughts, readers? As always, please sound off in the comments!

Brian Dipert is Editor-in-Chief of the Embedded Vision Alliance, and a Senior Analyst at BDTI and Editor-in-Chief of InsideDSP, the company’s online newsletter.

Related articles:

Leave a comment