Holidays are a good time to test mobile networks because they invariable require driving on crowded highways to visit the in-laws’ in the far-flung suburbs. So the long Thanksgiving weekend was the perfect opportunity to run AT&T’s (s T) Chicago LTE network and its new smartphones through the paces.
I did my tests in outbound and inbound stages. On my way from central Chicago to Arlington Heights, I basically tried to break the HTC Vivid and Samsung Galaxy SII Skyrocket with video. While on the way back into town, I ran a barrage of speed tests. In general, I was impressed with overall experience as well as the speeds I clocked, though I did have some concerns about how LTE’s inner-city network will perform when its truly loaded. But onto the results!
Starting at home, I turned off Wi-Fi on both phones and then downloaded the Netflix(s nflx) streaming app on the Samsung. It handled the 6-MB download almost instantaneously, but the real test would be how it actually handled watching a movie. I started some Britcom episode my wife and I had already seen (so as not to tempt her to watch while she drove), and while the video popped up instantly with no buffering, initially it played at a much lower resolution than I was expecting. Within 20 seconds, though, the show went into a much higher resolution and stayed that way for the remainder of the test.
Our trip took about 45 minutes, starting on Chicago’s streets (where it’s never really possible to go more than 25 mph) for 15 minutes and then onto the Kennedy Expressway for most of the remainder of the journey. The Netflix video ran flawlessly the whole way, only interrupted when the episode ended and I started a new one. I saw no discernable loss of resolution, no hiccups or pauses and no artifacts, except right when pulled into the driveway and there was a slight buffering delay. I don’t test as many gadgets as Kevin Tofel, but I’ve certainly played with quite a few, and I’ve never seen a phone on a wide-area cellular network do this before.
Networks are fickle creatures: bandwidth typically rubberbands between extreme highs and extreme lows as you move from the center to the edge of cells, as users join and drop out of the cell you currently occupy and as beams cross paths. At some point those wild fluctuations (and AT&T has plenty of them as we’ll see later) are bound to mess with your bitstream. To see a phone play continuous high-quality video without a single glitch while passing through dozens of different cells at freeway speeds and through crowded cityscape –- well, that’s truly impressive.
Of course, Netflix has a lot of buffering built in, which helps even out those peaks and valleys. So with the HTC, I surfed around a bit giving the device far less opportunity to cache. Websites came up almost instantaneously –- which was nice considering I’ve almost come to expect the “smartphone lag” –- while YouTube videos came in at full resolution without a hitch or buffering delay. Only at one point did I encounter a problem. A YouTube video halted mid-stream and I had to restart it.
Fast but not as fast you’d think in the city
On the trip back, I ran nothing but speed tests, which is where the fluctuations in bandwidth I expected became apparent. On the Samsung, I used RootMertics’ Coverage Map app to run a continuous speed test for the entire trip. (I would not recommend doing this with your own phone, especially on an LTE network. 45 minutes and nearly 90 test cycles ate through an entire 2 GB plan.) I split the test into two parts: freeway driving and city driving. Meanwhile, I downloaded Ookla’s Speedtest app on the HTC and used it to take individual speed measurements every two minutes or so.
The first thing I noticed is that AT&T’s LTE network is a lot faster outside of the city than it is within. In Arlington Heights I was getting measurements over 20 Mbps from Speedtest. Those speeds slowed down considerably as we started driving and got on the highway, but that’s to be expected. It’s much harder for a network to maintain a fast connection when moving. For the freeway trip, Root’s testing app ran through 68 cycles, recording an average download speed 10.8 Mbps and an upload speed of 6.1 Mbps, which is right in line with what Kevin T saw when he tested Verizon’s(s Vz) LTE network soon after launch. On the freeway trip, AT&T also had a 100% data connection rate, it never failed to find a signal over which to ship data.
Once we got to our exit, I started the tests over again. We were moving much slower but so was the network. Running for 21 cycles on surface streets, Root recorded an average download speed of 5.7 Mbps and upload speeds of 3.9 Mbps – almost half of what the devices clocked on the freeway. It also failed to connect on one of Root’s cycles and on one manual speed test. There’s much more population density in the city, yes, but there should also be a much higher cellular density to account for the greater number of devices packed into the same square footage.
I’m not complaining — 6 Mbps is plenty of bandwidth for anything you’d ever want to do on a smartphone (and most of what you’d do on a laptop). In only one of the tests did downlink speeds drop below 2 Mbps — only just barely — while upload speeds remained consistently high. In the 3G world, network speeds can easily drop down to the sub-100 kbps range, so as long AT&T can maintain a minimum of 2 MB range, I doubt its customers would ever notice anything like the capacity problems they experienced with the iPhone.
The issue is AT&T’s LTE network is practically empty today. It only launched here in Chicago in late September and its smartphones went on sale earlier this month. Once its network becomes loaded, with millions of devices vying for the same shared capacity, those lower inner-city speeds I recorded might become a problem — at least on the north side of Chicago. That said, AT&T has just started to build out its LTE network. In addition to adding more markets, it will likely increase the cellular density within existing markets as it adds more customers.