A Tale of Two Access Points

I’ve used Apple products for the last several years - let’s face it, I’ve become a fanboy. Recently, I began replacing switches around the house with Ubiquiti’s UniFi gear, so when I heard Apple was ceasing development on their in-house WiFi gear I opted to revamp the house with a new Ubiquiti access point. Like any good systems geek, I wanted to capture a before and after view - especially since I opted for a UAP-AC-HD (Ubiquiti’s 802.11ac “wave 2” 4x4 MU-MIMO access point).

At $350 (cheaper on Amazon) each, I wanted to make sure I got my money’s worth.

Surprisingly, I ended up with a bit of a mixed bag - and learned something about my Macbook Air in the process. I like graphs, so I figured I’d summarize and post the experience here.

WiFi at Home

My house is a large vinyl-clad box in Indiana. It’s around 3200 square feet, but it’s rather densely-packed into a 40-foot cube. The first floor had an older Airport Express (the wall-wart 802.11n model) in the kitchen at the back of the house, and an Airport Extreme (the modern pillbox with 802.11ac) in a more central location. The second floor just had a newer Airport Express (still 802.11n, though) in a central room at the front of the house - a fair ways away from the master bedroom in the back corner. Originally, I just had the two Airport Express units, but had poor reception around the living room (particularly after adding a number of WiFi devices, Lutron switches, and Hue bulbs). The Airport Extreme fixed that right up.

I initially measured three configurations - all three Apple APs as-is, the Airport Extreme alone (with the radios turned off of the two Express units), and a new UAP-AC-HD in place of the Extreme (which was unplugged). I later added a fourth round of tests for the UAP-AC-HD, with the transmit power raised to maximum and only one ESSID available (as my initial deployment had two networks, but I’d read that having more than one can decrease the performance in certain Ubiquiti APs).

Each measurement was done with iperf3 -c server.local -i 10 -t 10, running against iperf3 running on a well-connected server in my basement. I’m using a MacBook Air (6,2) for my wifi clients, with most other clients either idle or turned off. Bandwidth was measured from the living room couch, a chair next to the access point in the den, a green couch upstairs, and in the master bedroom.

Initial Charts and Graphs

To quote Lisa Simpson, I make lots of graphs. First, the base case:

Average Throughput, Initial Network

Merely looking at the averages here doesn’t really illustrate the network’s true behavior. While measuring, I went from the downstairs couch to the den, then to the green couch upstairs, then the bedroom - then repeated the process in reverse, and then again in order for a third run. Only three runs isn’t necessarily statistically significant, but even I need to draw the line at calculating a Chi Square test on my off time.

In this case, an interesting effect emerges looking at the three runs in isolation:

Individual Throughputs, Initial Network

The first run went at reasonable speeds - over 400 Mbps from my couch, nearly 600 Mbps in the den (right next to the Airport Extreme), and still 277 Mbps from the green couch upstairs. The bedroom, though, was just far enough away to cause my laptop to lose the 5 GHz signal. Dropping down to the 2.4 GHz radio was a big problem, since the laptop never recovered (at least, not until the WiFi adapter was cycled).

Removing the older Airport Express units (turning off the radios) helped somewhat:

Average Throughput, Airport Extreme Only Individual Throughputs, Airport Extreme Only

Most likely, once my laptop had dropped down to the 2.4 GHz radio it was connecting to the older Airport Express units (which were farther away from the various couches). Still, the fundamental issue remained; the Airport Extreme was great as long as I was on the 5 GHz radio, but once off I would never get back on again.

Ubiquiti’s UAP-AC-HD

Switching to the UAP-AC-HD, it was immediately obvious that the overall throughput was lower than the Airport Extreme:

Average Throughput, UAP-AC-HD Individual Throughput, UAP-AC-HD

Oddly enough, the Ubiquiti was able to better reach the upstairs bedroom - but much less capable when a few feet away in my den (316 Mbps vs 598 Mbps with the Airport Express). Worse, my laptop still suffered from the same behavior on the 5 GHz radio; a trip upstairs was enough to lose the connection, and it never fully rejoined (though the UAP-AC-HD seemed to have much better 2.4 GHz performance, as evidenced by runs #2 and #3 in the graph above).

Ubiquiti can set the AP strength in software; by default, it’s set to “Auto”. If I set this to “High” I get much more consistent results, for only a nominal increase in power consumption:

Average Throughput, tuned UAP-AC-HD Individual Throughput, tuned UAP-AC-HD

Yes, performance upstairs isn’t the greatest, and throughput is still less than that achieved with the Airport Extreme. But at least with the power set to High, I never lose the 5 GHz signal - meaning that however I roam around the house, I don’t drop off the faster radio.

Final Thoughts

I’ll be keeping this AP, despite the obvious peak performance issues. I don’t rely on a high streaming speed from my laptop, and the integration with the rest of Ubiquiti’s UniFi system (plus the insight it brings into what the kids are doing online) makes it more worthwhile. The recent rumblings about Apple effectively abandoning their access point lineup may have something to do with that decision as well…

All in all, it’s not a bad AP - though without a lot of MIMO-capable wireless gear, and a comparatively limited number of devices (far fewer than the 500 the AP claims to support) it’s honestly a lot of overkill for my house. Still, with the ease of setup and management, I could see myself buying another couple of these (or at least the UAP-AC-Pro predecessor) to round out the coverage in the old homestead.