Review: Comparing Google Wifi to other mesh networking heavyweights

By JIM SALTER for ARS Technica.

Here at Ars, we’ve been rocking some high-tech Wi-Fi setups for a while—in particular, Senior Technology Editor Lee Hutchinson and I are fond of Ubiquiti’s UAP line of wireless access points. A lot of people either can’t or don’t want to run cables through their house at all, though, and that’s the niche Wi-Fi mesh kits seek to service.

I’ll save you some time up front: I’ve played with just about everything out there, and nothing comes close to the performance of multiple access points with full wired backhaul like the UAPs. To be fair, nothing Wi-Fi at all comes close to the performance of wired Ethernet itself, so don’t get too excited about the “3.2 gigabits per second!” that AC-3200 Wi-Fi router promises you. You’ll never actually see such speed. But if you don’t want wires and you don’t want the possibly intimidating controller systems like Ubiquiti’s UniFi interface, mesh might be for you. Luckily, today happens to be a bit of a boom for mesh offerings.

What we tested (this time)

Our trio of contenders are Google Wifi, Plume, and AmpliFi HD. We’d originally intended to test the new Eero v2.0 firmware, too, but unfortunately two of our three test units refused to make it through the update process. At press time, we don’t have a fix from the Eero team, so we regretfully had to put Eero aside for now.

We tested each device using iperf3 to get raw throughput numbers in several sites throughout the house, but first and more importantly, we’re going to look at heatmaps of the Wi-Fi signal produced by each kit. Hutchinson introduced me to NetSpot, a free-as-in-beer visualization tool for Mac or Windows that lets you walk around your house and map out the signals throughout. I was impressed enough that I actually installed Windows on a laptop just so I could run it—and it was worth it. The visuals it generates are an absolutely invaluable tool if you want to geek out hard and get the most out of your Wi-Fi coverage.

AmpliFi HD

AmpliFi is Ubiquiti’s answer to wireless mesh networking. To be honest, I expected it to sweep the field clean based on my experience with their UAP line of traditional wired-backhaul access points. Our own Eric Bangeman has AmpliFi HD in his home and loves it, but my testing hasn’t left me as thrilled. AmpliFi HD has a seriously “hot” RF signal that carries well, but it’s coupled with a very not-kid-friendly, avant-garde physical design that leaves it prone to really bad locations and even physical breakage.

AmpliFi’s router is a little white cube that makes a reasonably good, if Star Trek-ish, clock, but its satellite units are magnetically coupled “towers” that plug directly into wall sockets. They’re big, they’re outlandish, and they’re fun to break apart.

AmpliFi’s HD model packs enough of a wallop in RF signal to largely, although not entirely, overcome the very poor usually floor-level placement its satellite design demands. I still wouldn’t feel comfortable letting my three-year-old roam loose around these devices, though. Why, why would you make your Wi-Fi gear fascinating to break?

This is a heat map of the Wi-Fi signal put out by AmpliFi HD’s base unit only. The image scale is from -30dBM to -75dBM. Basically, anything from red to green is great; pale blue is decent but slow, and once you dip all the way down to the deepest blue it gets iffy.

The big takeaway here is that the base unit alone is fine for the entire top floor of my house (around 2,800 square feet) along with a good chunk of the yard. This should make you ask the question: why mesh, then? Well… there’s a bottom floor, too. And there’s a foundation slab and a few feet of packed earth in between it and the wiring closet upstairs.

What you’re looking at here is a heatmap of my normal day-to-day Wi-Fi on the bottom floor, but only looking at the signal coming from the top floor. Until I set up my Ubiquiti UAPs—one on the top floor, one on the bottom floor—there basically just wasn’t any Wi-Fi down here at all.

Things look much better with multiple APs. Let’s take a look at AmpliFi HD’s signal on both floors, this time with all three nodes activated.

With both satellites added in, the top floor looks like an absolute war zone. You’re definitely going to get connected, no problem, wherever you wander around. The satellite on the right is heating things up nicely for the carport and driveway. The question then: is the one underneath the TV clearing that foundation slab and overcoming the problem area downstairs?

The satellite upstairs clears the edge of the foundation slab, and without a doubt, it makes things a lot better down here. Performance testing shows this is actually usable, but it’s still not something I’d want to bring my buddies over to brag about.

Google WiFi

I wasn’t quite sure what to expect from Google, but I was definitely excited to find out. Like AmpliFi, Eero, Luma, and most of the mesh pack, Google is pushing Google Wifi in packs of three as the default size. On a conference call with their team, the presenters told us to expect “a lot of signal” from even one Google Wifi node. They said this in the kind of careful tone that told me they really meant it. I’m used to Wi-Fi gear with very high-powered radios, though, so I wasn’t sure how much signal I should really expect from something that didn’t look out of place in my seven-year-old daughter’s room on a piece of furniture her great-grandmother used to call a “vanity.”

Still, this was Google, and the presenter didn’t gush about how amazingly powerful it was. He sounded more… cautionary. So I was pretty eager to see what the heatmap survey had in store.

Google’s not kidding. Google Wifi nodes have really powerful radios in them, and it’s incredibly obvious when looking at a heatmap of the router node only.

I was pleased to see that the app individually tested the satellite nodes’ throughput to give me a yea/nay on placement as I set them up. This might not sound like much, but it’s a big and much-needed improvement over what I’ve grown used to from other devices. Never again will you stumble through the inevitable “this LED flashed a funny color; is that a good funny color, or a badfunny color?”

With both satellites up and running, the heatmap looks like an RF war zone again—even hotter than AmpliFi HD’s.

At this point, you’re probably wondering why the satellite locations aren’t identical to AmpliFi HD’s. Rest assured, there’s a reason: the physical form factor of each device lends itself to different placement. AmpliFi’s right-in-the-socket design made it seem best served in the wall socket underneath my television where it wasn’t actually down at baseboard level. Similarly, one of Google Wifi’s little jar-shaped nodes would have looked funny just sitting there under the TV, and it wouldn’t be getting as much height to do its magic as it otherwise could.

Basically, it would’ve been cheating to put a node somewhere I wouldn’t put it if I was really living with the device day to day, so I didn’t. With that said… what’s the downstairs look like this time?

Google Wifi’s coverage downstairs is a marked improvement over AmpliFi’s, except, unfortunately, for the bedroom on the right. It’s actually quite usable there, too, but some adjustment of the satellite locations would’ve helped a lot here.

In addition to a smoking hot signal, Google Wifi offers a ton of accessible functionality. You can tweak the brightness of its LEDs to your heart’s content, you can pause Internet delivery to devices or groups of devices (get to the dinner table, kids!), you can even manage home automation gear. All of this functionality is exposed via a “cards” system that should be familiar to anyone using modern Android phones.

Google Wifi performed well overall—noticeably better than AmpliFi HD, even where the signal level looks a little “bluer”—but it didn’t take top honors either.

 

Plume

I went into this process really, really annoyed with Plume. At review time, there was no Android app, though there’s a beta version that’s expected to be released in the coming weeks. There was no Web interface either. You had to have an iOS device or pound sand. I gritted my teeth, bought a refurbished 4th-gen iPad, and secretly kind of hoped for an excuse to be mean.

For the kind of folks who are really into the aesthetics of Wi-Fi devices, Plume is probably about as good as it gets. Instead of going sleek and white like an iDevice, Plume went for “completely unobtrusive.” It works—each pod is about the size of a plug-in air freshener and looks something like an antique brushed-nickel closet door knob. I was actually a little worried that I’d forget one during tear-down, leaving me to manually hunt through the house to find it.

The attention to aesthetics continues in the app itself. Typically, I’m not really the kind of person who cares much about this. If it performs, and it lets me get to all the nitty-gritty, I don’t much care if it looks like a chainsaw made love to a pickup truck. But even I have to admit, I found Plume’s UI oddly charming. It’s kind of reminiscent of a twilight scene in a Final Fantasy RPG.

Plume doesn’t offer much in the way of technical feature accessibility; if you want to play with static routes and port forwarding and so forth, you’re mostly out of luck. You get a LAN, a WLAN, and a guest WLAN, and that’s pretty much all you get.

This isn’t as bad as it sounds, though. Unlike a lot of the mesh kits, Plume makes it really, really easy to just put it in “bridge” mode, potentially giving you the best of both worlds. You can let your router be a router and let Plume handle the Wi-Fi. This is going to be especially important for those of you with gigabit Internet connections who don’t want to sacrifice any of the high-end performance for their wired machines in order to get a little Wi-Fi convenience. It’s also important for those of you who might like to set up Wi-Fi for your more technically challenged friends and family. When they inevitably call their ISP after an Internet outage, and the ISP talks them into resetting their router, it won’t screw up their Wi-Fi!

Moving on from the UI, an individual Plume pod (unlike Google Wifi, AmpliFi HD, or one of my own Ubiquiti UAPs) is decidedly not a high-powered device.

This wasn’t a big surprise, though; what makes Plume interesting is that you can buy a six-pack of its pods for $329—only $20 more than a 3-pack of Google Wifi and $20 less than an AmpliFi HD kit. As we’ve seen to some degree already, placement is absolutely critical on these three-device kits in a large or challenging house. Netgear’s Orbi, which I reviewed recently for the Wirecutter, got around this problem by doing less with more. You get one router, one satellite, and Netgear’s simple deployment instructions: “put the router where the Internet connection is, and the satellite as close to the center of the house as you can get it.” (It worked really well.) Plume’s approach is the exact opposite: here’s a bunch of tiny cheap things, scatter ‘em like chickenfeed and win.

Does it work? Well, the sheer number of pods means this map needs a little explaining. The router is in the network closet as always; satellite pods one through four are upstairs on the back wall of the office, the back wall of the large bedroom on the left, under the TV, and one behind the microwave in the kitchen, on the right. The final pod (which you’ll note is grayed out compared to the others) is actually downstairs on the back wall of the den, deep in the Wi-Fi shadow my foundation slab creates. The intent was to force it to talk only to other satellite pods rather than letting it get a clear signal to the router pod in the network closet.

In sharp contrast to the heatmaps from AmpliFi HD and Google Wifi, Plume’s heatmap shows almost no red at all but a much more even distribution of yellow and green throughout the house. In testing, this translated to clients quickly connecting to the closest pod instead of getting confused and tending to connect to one halfway across the house. It’s also nice because some client devices actually perform worse when overwhelmed with too much signal. I’ve seen more than one thread where somebody was frustrated with speed tests… which didn’t get better until somebody told them to back a few feet away from the router with their laptop. Put it all together, and Plume’s gentler, more consistent signal strength is a good thing. (Its throughput numbers bear this out.)

Turning the tables, we can now see the upstairs pods half-grayed, with the one on the back wall of the den displayed as normal. A lot of the signal you’re seeing in the office (on the left) and the bedroom (on the right) doesn’t belong to that pod, though—it comes from the pods upstairs in the kitchen and in the big upstairs bedroom. While the multi-hop topology I forced the downstairs pod into does work, it’s worth noting this is significantly slower—I only got 30mbps when connected to that pod, which turned into an easy 80 mbps when I unplugged that pod entirely and let my laptop connect directly to one of the pods upstairs. Much like I discovered when I tested Eero for the Wirecutter, multi-hop here works, but it’s best avoided if you don’t absolutely need it.

Plume boasts a proprietary “adaptive Wi-Fi” protocol that routes the backhaul from satellite pods to the router pod over multiple channels and multiple paths. This is the kind of thing I tend to raise a dubious eyebrow at in press releases, and I wasn’t alone in that immediate impression. However, it did bear out in testing. I saw successive iperf3 runs in the same location trend upward from run to run until the fifth or sixth run, when they stabilized. I mentioned this in a call with Plume CEO Fahri Diner (who seemed pretty defensive after the scathingly poor review Dong Ngo gave Plume on cnet), and he said that was an indication of the algorithm adapting to the topology of my network. He said it’s the sort of thing a hasty “one run and done” review might tend to miss.

There’s a company blog post that goes into a little bit more detail about Plume’s adaptive Wi-Fi here. Between my own results testing Plume and seeing the numbers get better and a really in-depth (but unfortunately confidential) technical paper Fahri shared with me, I’m convinced—the adaptive Wi-Fi is really a thing. It really does help.

The big take-away from all this is that Plume just works without a whole lot of effort or planning on your part. You buy enough of them to scatter evenly throughout your house, you plug them in, they do the rest. If you don’t want a separate router, that’s fine; Plume can be your router. If you have a router and you like it, that’s fine, too. Plume goes into bridge mode to handle Wi-Fi only with two taps on your phone or tablet. Easy peasy.

Throughput tests

Hey, this is Ars Technica, you didn’t think you were going to get away without some numbers did you? I went a little easy on the number crunching this time, because I wanted to pay more attention to coverage, and generating those heatmaps takes a significant amount of time if you want to do it right. But the heatmaps, as useful as they are, still don’t tell the whole story. While your device will generally connect best at the same signal strength regardless of what it’s connecting to, how fast that connection is can change significantly from one access point to the next.

Rather than set up a giant table with a ton of specific places in the house, this time I wanted to demonstrate speeds at particular signal levels. So what I did was look at the heatmap for each device and test four sites apiece with iperf3. The physical location varied a little from device to device, but the signal strength was about the same for each. The client device used was a Fujitsu Lifebook running Windows 10 Pro with a Linksys WUSB6300 USB3 dongle (an AC1200 adapter that drastically outperforms any internal Wi-Fi I’ve tested so far).

If the heatmaps didn’t make it clear enough, the throughput testing does: Plume absolutely dominates the pack in my multi-floor “Wi-Fi hard mode” house. Google Wifi’s stupendously hot signal outperformed it in the car and barely eked out a win in the slam-dunk “we’re sitting right outside the network closet” test. Everywhere else, though, Plume’s more consistent, close-range connection wins every time.

AmpliFi HD, hampered by its awkward satellite design, did not take first place in any location tested. It’s worth noting that it’s still night-and-day ahead of the typical router and/or Wi-Fi extender, though (again, Managing Editor Eric Bangeman loves his). If you dig its funky design, it might still find a place in your home or apartment.

The one cautionary note about Plume, which I referred to earlier, is that in the downstairs test shown here I actually unplugged the downstairs Plume pod to get the result you see. With the downstairs pod plugged in, I only got 30mbps. That 30mbps was quick to connect and utterly reliable, but it was a lot slower than the 73.2mbps you see above blowing AmpliFi and Google Wifi out of the water. Even with Plume, it does pay to test.

Plume CEO Fahri Diner opined that Plume would probably have improved the performance of the downstairs pod if given another day or two for its algorithm to “learn” about my topology, but I’m still a little dubious about that. Plume’s results here were very similar to Eero’s when I forced it into multi-hop for the Wirecutter, which leads me to think that it’s just something best avoided if you possibly can.

Plume

The good

• Lowest cost per unit
• Unobtrusive design fades into your home
• No cable clutter. good performance
• Super expandable
• Great coverage
• No-brainer “one in every room or so” deployment

The bad

• Still a lot more expensive than a standalone router, and most people’s homes can be served fine from a single, good standalone router

The ugly

• It’s iOS only right now, though Plume has a beta Android app now and expects Android to be fully functional in a few weeks.

AmpliFi HD

The good

• Strong signal
• Avant-garde design that will appeal to Star Trek fans
• Ubiquiti Networks heritage

The bad

• Satellites can only connect directly to the base router.
• Not expandable.
• Placement is crucial, and placement options are limited due to the satellite design, especially for US people whose outlets have the ground prong on top, not bottom, of the outlet.

The ugly

• The pointless breakaway magnetically coupled ball socket on the satellites is useless for actually orienting the antenna, and fascinating-to-break for small kids, mischievous roomates, and possibly even dogs.

Google Wifi

The good

• Strongest signal of the bunch
• Covered the house well
• Good aesthetics
• Second-best performer
• Loaded with tons of extra features and things to tweak and customize, up to and including the brightness of its LEDs
• Familiar Material Design interface
• Expandable

The bad

• Placement is critical, and at the price, you probably don’t want to put one in every room like Plume

The ugly

• Remember Revolv? This is something of an experiment for Google, and the company has a history of abandoning its less-loved projects “whenever”.
• Since Google Wifi is cloud-managed, if Google decides to turn off the infrastructure, your Google Wifi nodes become instantly useless.