T-Mobile’s 42 Mbps HSPA+: Fast, but it’s still no LTE


T-Mobile USA likes to talk up the capabilities of its souped-up 42 Mbps HSPA+ network, claiming it can rival even the new LTE networks deployed by Verizon Wireless(s VZ)(sVOD) and AT&T(s T). Well, over the last few weeks, mobile network tester RootMetrics has been holding T-Mobile to its claims, pitting the carrier’s new 42 Mbps smartphones against their LTE equivalents in real-world environments. Root found T-Mobile’s network can support some truly astonishing speeds, but they still don’t match up to the raw bandwidth of LTE.

Root tested the HTC Amaze 4G in five markets: Columbus, Ohio; Orlando, Fla.; Providence, R.I.; Richmond, Va.; and Tuscon, Ariz. Root found T-Mobile’s 42 Mbps HSPA+ network averaged download speeds between 6.2 Mbps and 7.7 Mbps and upload speeds between 1 Mbps and 1.4 Mbps. (You can see the full reports on Root’s website.) T-Mobile handily beat out AT&T’s (s t) HSPA and Sprint’s(s S) WiMAX services — in most cases more than doubling their speeds — though Root hasn’t yet tested Ma Bell’s recently released LTE smartphones. But in four of those markets (Verizon hadn’t yet offered LTE in Providence), Root pitted the Amaze against Verizon’s LTE HTC Thunderbolt, which averaged download speeds between 9.6 Mbps and 11.4 Mbps and upload speeds between 3.8 Mbps and 6.7 Mbps. It’s not hard to pick a winner.

These tests are important, because they’re the first apples-to-apples comparisons of rivaling network technologies. To say LTE is inherently faster than HSPA+ based on older tests simply isn’t fair because LTE networks like those deployed by Verizon and AT&T are using double the spectral bandwidth – 10 MHz on both the downlink and uplink – compared to a typical HSPA+ network. T-Mobile’s new 42 Mbps network, however, bonds two 5 MHz-by-5 MHz HSPA+ carriers together. For the first time in the U.S., we’re able to compare two networks, megahertz for megahertz.

That LTE still comes out on top may not seem like a surprise, but in truth, as an air interface, LTE is not that much more spectrally efficient than the latest generation HSPA technologies. The likely reason Verizon’s – and presumably AT&T’s – network is so much faster is the multiple input-multiple output (MIMO) smart antenna technologies used in LTE. MIMO sends two parallel data transmissions, allowing LTE in many cases to double up on capacity. HSPA+ can support MIMO as well, along with a host of other upgrades that can send its theoretical ceiling of 42 Mbps into the stratosphere. T-Mobile has simply elected not to deploy those upgrades – at least not yet.


Andrew J Shepherd


Some of your analysis insinuates that the RootMetrics comparison is an apples to apples, MHz to MHz test of non MIMO, Dual Cell HSPA+ airlink tech to MIMO LTE airlink tech. But that is not necessarily true. Rather, the comparison also includes other variables, such as cell site density, backhaul capacity, and, most importantly, network loading.

On the last count, we know that T-Mobile (according to its third quarter report) has 10.1 million W-CDMA devices on its network nationwide. Has VZW released its number of LTE devices in the field? And can we assume that the usage patterns of both sets of subs are comparable?

The greatest flaw in the RootMetrics test and/or your conclusions from it is the incorrect presumption of equal bandwidth. We do know that VZW has deployed LTE currently in only its nationwide Upper 700 MHz C block 22 MHz spectrum, and that allows for one 10 MHz x 10 MHz LTE carrier channel. But, in all of the markets tested, T-Mobile has multiple AWS 2100+1700 MHz licenses. And only Richmond is limited to 20 MHz bandwidth. See below:

Columbus (AWS A block 20 MHz + AWS E block 10 MHz) = 30 MHz
Orlando (AWS A block 20 MHz + AWS D block 10 MHz + AWS E block 10 MHz) = 40 MHz
Providence (AWS A block 20 MHz + AWS E block 10 MHz) = 30 MHz
Richmond (AWS D block 10 MHz + AWS E block 10 MHz) = 20 MHz
Tucson (AWS A block 20 MHz + AWS F block 20 MHz) = 40 MHz

As a result, in Richmond, T-Mobile can deploy only one twin set of 5 MHz x 5 MHz Dual Cell HSPA+ 42 carrier channels (i.e. the same 20 MHz total bandwidth as VZW’s LTE bandwidth). But, in Columbus and Providence, T-Mobile can deploy an additional single 5 MHz x 5 MHz HSPA+ 21 carrier channel. And, in Orlando and Tucson, T-Mobile can deploy a second twin set of 5 MHz x 5 MHz Dual Cell HSPA+ 42 carrier channels. That means that T-Mobile can spread the network loading over multiple carrier channels, thereby potentially affecting the MHz to MHz comparison.

So, the RootMetrics comparison is really just a reasonably representative test of the state of the T-Mobile HSPA+ network over a given span of time to the state of the VZW LTE network over the same span of time. A comparison over an earlier or later time period could produce different results, or a comparison between carrier Y HSPA+ and carrier Z LTE could produce different results. Hence, we should be careful not to use the RootMetrics results necessarily to extrapolate non MIMO, Dual Cell HSPA+ performance vs MIMO LTE at large.


Kevin Fitchard

Hi AJ, that’s quite a comment! Before diving in I should say that Root didn’t really make the comparison, per se. They provided the data, while I was the one drawing conclusions about LTE and HSPA+. Now let’s take your points:

I see your point: There really is no such thing as a true apples-to-apples comparison when talking about mobile networks. Each has its own configuration, topology, density and band scheme. I definitely admit to generalizing. My aim was to point out that this comparison carries more weight because the two networks have the same size carriers, while past comparisons have always pitted an HSPA network against an LTE network with double or quadruple the bandwidth. You’re right, there are a lot more nuances. I didn’t really want to go into all of them in a short blog post, but I did point out the big one (MIMO vs. SISO) in the final graph.

I also see your point on the congestion issue. T-Mobile has more subscribers on its HSPA+ than VZW has on LTE now, and its 42-Mbps phones have to share that capacity with 21 Mbps phones and regular old HSPA phones, while Verizon’s is essentially a brand-spanking-new network. But I don’t think its accurate to say T-Mobile’s network is loaded. If that were the case, then average speeds on its phones would be declining. But Root’s data doesn’t seem to show that happening. In some cases average speeds are actually increasing, which may just be better handset chips but certainly not a sign that load is starting to take its load. Verizon’s numbers have been dropping since the days its network was wide open and seem to have settled into that 10 Mbps range. You’re right a more accurate comparison would require an equal number of subscriber in the same distribution, but I’m not convinced it would make that much of a difference.

As for the T-Mobile’s extra carriers, are you sure that applies here? My understanding is that T-Mobile can aggregate two carriers right now, if not on the network side then most certainly on the device side. So if it has more than two carriers deployed in a market it definitely gets the benefit of capacity but can’t bond that third link into a single device connection.

The purist in me agrees entirely with you. No two networks are alike and comparing them is pointless. But the more practical side of me thinks that comparisons will be made anyway so it’s best to point out the ones that make more sense.

Keep up the comments! Make me honest :)


Andrew J Shepherd


First, I want to say that I came across your work for the first time just last month, but I have read with great appreciation your published pieces both here at GigaOm and at Connected Planet Online. So, my lengthy comments are not intended to criticize, only to add to the discussion.

As for HSPA+ carrier aggregation, you are correct that T-Mobile can combine only a twin pair of 5 MHz x 5 MHz carriers in Dual Cell HSPA+ 42, as UMTS Releases 8 and 9 allow for aggregation of just two contiguous carriers. Release 9, however, does allow simultaneous Dual Cell + MIMO operation. And speculation is that a future Release 10 will include Multi Cell HSPA+, which could allow for aggregation of three or greater carriers, possibly even among carriers that are not contiguous (or even in different spectrum bands). This is akin to many of the carrier aggregation capabilities in LTE Advanced, which is how AT&T, for example, plans to use the former Qualcomm Lower 700 MHz D/E block 6 MHz unpaired spectrum for LTE supplemental downlink.

That background aside, my point about licensed bandwidth and multiple carriers was directed toward network loading, not carrier aggregation. If we compare one VZW 10 MHz x 10 MHz LTE carrier to one T-Mobile twin pair of 5 MHz x 5 MHz Dual Cell HSPA+42 carriers, then that does seem to be an apples to apples, MHz to MHz comparison. And we do know that VZW has Upper 700 MHz C block bandwidth to deploy just one LTE carrier channel per market, currently. But what of those markets in which T-Mobile has sufficient bandwidth to deploy greater than two HSPA+ carriers?

In particular, take Columbus and Providence. In both markets, T-Mobile has a total AWS 2100+1700 MHz spectrum bandwidth of 30 MHz. That allows for one twin pair of Dual Cell HSPA+ 42 carriers and a single HSPA+ 21 carrier channel. Presumably, some of the RootMetrics trials in those markets could have been dynamically assigned by the network to the Dual Cell HSPA+ 42 carriers, some trials to a slower HSPA+ 21 carrier, unbeknownst to RootMetrics, thereby affecting the average results.

To use a road analogy, the RootMetrics tests strive to compare a VZW LTE “four lane road” to a T-Mobile Dual Cell HSPA+ 42 “four lane road.” If both VZW and T-Mobile have only one “four lane road” apiece, then that may be a fair comparison. But the comparison may get skewed when VZW definitively has only one “four lane road” per market, while T-Mobile may have one “four lane road” side by side another “two lane road” in some markets, even two “four lane roads” side by side in other markets. The VZW “traffic” is always confined to one “road,” while the T-Mobile “traffic” may be spread across multiple “roads.” Does that analogy make sense?


Kevin Fitchard

Thanks for responding, Wabtech. Definitely all good points, as are AJs, and I believe you’re both right. A true comparison can only be performed between two networks or equal size, density and loads. That said, I think these tests still have a lot of relevance because both networks use the same spectral bandwidth. I agree, load is an important factor, but I guess I just don’t feel that Verizon’s network is all uncrowded (it has nearly 4M subs many of them modem users) or that T-Mobile’s is all that crowded (if it were already experiencing congestion issues, it would be in trouble).

Kevin Fitchard

No worries, AJ, I didn’t take it as criticism. Just good open debate on issues I think are really cool (but few others don’t, which is why I limited the network discussion in the original post).

I see where you’re going with the multiple carriers being multiple roads analogy, but a) doesn’t the network and the phone duct tape together whichever two of the three carriers happen to available, and b) even if T-Mobile phones are bouncing between single and dual-carrier connections, aren’t Verizon’s phones doing the same, going between LTE and CDMA?

Just some thoughts. Your point makes sense. I’m just try to explore all possibilities here.

Andrew J Shepherd


Per my understanding of Dual Cell HSPA+, the twin paired carrier channels must be contiguous. In other words, the two carrier channels to be bonded together must be contained within the same license (AWS A/B/F block 20 MHz), or the two must be contiguous within two adjacent AWS licenses (e.g. AWS D block 10 MHz + AWS E block 10 MHz).

Continuing with the Columbus and Providence examples, since T-Mobile controls the AWS A block 20 MHz + AWS E block 10 MHz licenses in both markets, and as the AWS A block 20 MHz license is separated by 20 MHz from the AWS E block 10 MHz license, the AWS E block 10 MHz license cannot be used for Dual Cell HSPA+ 42, only HSPA+ 21.

See the AWS band plan here:


The capability to “duct tape” disparate blocks of spectrum together will not arrive until LTE Advanced, maybe UMTS Release 10, too.


Kevin Fitchard

You’re right, AJ. Got ahead of myself in the standards didn’t I? Still, what about Verizon. Wouldn’t it customers be shifting from CDMA and LTE just as T-Mobile’s are switching between single and dual carriers?

Andrew J Shepherd

I need to make a correction, Kevin. UMTS Release 9, which is the highest current standard, does already support Dual Cell HSPA+ across two non adjacent carriers. That said, I do not believe that T-Mobile has deployed the Release 9 update because its primary function is to enable simultaneous Dual Cell + MIMO. And T-Mobile, according to my sources, has shied away from MIMO so far. Thus, T-Mobile has deployed Release 7, Category 14 for HSPA+ 21 and Release 8, Category 24 for Dual Cell HSPA+ 42.


As for VZW devices switching between LTE and EV-DO carriers, I am sure that does occur, though it should be due almost entirely to LTE coverage issues, not LTE load balancing concerns. And the transition from LTE to EV-DO should be clearly indicated on the device under test. On the other hand, I do not believe that T-Mobile equipment displays a distinction between HSPA+ 21 and Dual Cell HSPA+ 42. Both are labeled simply “4G.” Only a field test debug screen (or similar app) would indicate the actual category of the current W-CDMA carrier.

Of course, another notable difference in the RootMetrics comparison is that the VZW LTE network does not presently carry voice nor SMS traffic, while the T-Mobile HSPA+ network does on both counts.



The biggest difference between Verizon LTE and Dual Cell HSPA+ is the number of subscribers on both networks. Yes, T-Mobile’s HSPA speeds are increasing but that’s only because they are increasing backhaul and spectrum carriers. The number of users and amount of usage on the network is also increasing bringing down the average speed. If the dual cell HSPA network was as empty as Verizon’s LTE network we would see some awesome peaks and sustained speeds much more of the time, but for now we get is what 7-12Mbps? It’s not a fair comparison. These speeds are also what Verizon says to expect on LTE, but obviously not right now when LTE is very uncrowded.

Raymond Padilla

The difference in ping should be noted too. The ping rates between HSPA+ 42 and HSPA+ 21 are non-existent. LTE absolutely trounces HSPA+ in terms of ping and upload.

Kevin Fitchard

Good point, Raymond Root’s studies don’t get into latency, but it’s easy to see the difference between and LTE and HSPA network. I have become so accustomed to the “mobile lag” when on a cellular data network I was almost shocked when Webpages loaded almost instantly on LTE networks.

Comments are closed.