AT&T's LTE is the fastest network in Houston, but it's nowhere near as fast in Chicago. In our tests in Houston, we found that AT&T's brand-new 4G, LTE network achieved stellar speeds averaging 24.64Mbps down. But in tests in Chicago, LAPTOP Magazine found anemic speeds: a mere 5.59Mbps down. What's up?
The Big Blue carrier launched its new 4G network on Sunday with five cities: Dallas, Houston, San Antonio, Atlanta, and Chicago. As I pointed out in our test results story, Chicago stands out; it's not considered one of AT&T's strongest markets, and AT&T's LTE coverage there isn't as comprehensive as it is in the other launch cities.
Look at the results in the chart below. (The vertical axis is speed in Mbps, the Chicago data is from LAPTOP's study, and I'm really bad at making Excel charts on deadline.) What's striking is how consistent Verizon's speeds are compared to AT&T's wildly varying results.
I've asked AT&T about these differences, and it hasn't gotten back to me yet. But the reason appears to be spectrum, and it means there's going to be much sharper differences in AT&T speeds city by city than there will be with Verizon's speeds.
Why AT&T is Fast - And Slow
According to Phone Scoop, AT&T has 24MHz of 700MHz spectrum in its Texas launch cities, but only 12MHz in Chicago. That means it can run its LTE network on paired 10-MHz channels in Texas, but it has to use 5MHz channels in Chicago. Verizon, on the other hand, has at least 22MHz everywhere in the nation, so it can use at least 10MHz channels everywhere.
AT&T plans to supplement its network with its AWS 1700MHz holdings as well, which could help the situation in the future. It's just not using it yet. As this Phone Scoop map shows, AT&T (then Cingular) also owns AWS spectrum scattered throughout the country, including in Chicago.
So is this what 5MHz LTE channels look like? Fortunately, we have another example of a 5MHz LTE rollout in the U.S.: MetroPCS. MetroPCS is actually running 4G LTE on even narrower channels in some cities, down to 1.4MHz. But in Dallas, for instance, MetroPCS is running its LTE network in a 5x5 arrangement.
In our Fastest Mobile Networks test, we found that MetroPCS's Dallas LTE network averaged 2.64Mbps and peaked at 6.3Mbps. That was in a loaded, real-life scenario, too. So AT&T's Chicago result of 5.93Mbps without any load is within that range, showing that AT&T's problem in Chicago may be spectrum.
Early this year, AT&T's chief technology officer John Donovan told MuniWireless that the company needs 10Mhz channels for optimal speeds.
"You can deploy [LTE] using 5×5, but that doesn’t last you very long given the way data is being used today," Donovan told the site.
AT&T argues that the solution is to merge with T-Mobile and buy Qualcomm's 700MHz spectrum as well. Qualcomm has some 700MHz spectrum in the highly populated Northeast Corridor and California that AT&T has proposed to buy. T-Mobile doesn't have any 700MHz spectrum at all, though; the advantage there would be to increase AT&T's 1700MHz AWS holdings.
AT&T's plan also seems to assume the carrier won't use any of its existing 850MHz or 1900MHz spectrum for LTE, including the spectrum that's currently used for 2G EDGE. According to a chart at the bottom of this MuniWireless story, that's where AT&T's spectrum strength is: while Verizon owns more 700Mhz spectrum than AT&T, AT&T generally owns considerably more 1900Mhz than Verizon.
What This Means For You
Cities where AT&T has more 700MHz spectrum will probably roll out earlier and have better speeds. According to Phone Scoop, that includes Baltimore, Boston, Cleveland, Detroit, New York, Philadelphia, St. Louis, San Francisco, and Washington.
Cities where AT&T has less 700MHz spectrum will have slower speeds; that includes Miami and LA. And areas where AT&T has neither 700MHz or AWS spectrum will get LTE last of all; that doesn't include many major cities, but much of the states of Iowa, West Virginia, Kentucky and the Florida Panhandle all fall under this category.
T-Mobile has turned to HSPA+ 42, which it argues is a more efficient way to use small amounts of spectrum, but AT&T has said it won't use that technology, although it's compatible with the carrier's technology path. In our Fastest Mobile Networks tests in New York City, a strong T-Mobile market, we saw speeds on the carrier's then-HSPA+ 21 network averaging 4.57Mbps down. HSPA+ 42 should be faster than that.
If AT&T is denied the easy answer of swallowing T-Mobile and all its spectrum, the carrier is going to have to get creative to match Verizon's nationwide coverage and speeds.