3G Speeds in the US Measured in kbit/s???

A number of sources have reported this week about a recent performance measurement study that puts 3G downlink speeds of Verizon and AT&T in a region between 245 – 645 kbit/s and uplink speeds between 106 – 305 kbit/s. Unfortunately I don't have a copy of the study to look at the details but official publications usually do not downplay performance. Having said this, these numbers are really puzzling to me because they are so far away from what I experience in 3G networks all over Europe.

Here are a number of specific examples: In Germany, I consistently get 3 MBit/s and beyond in several networks if coverage is good, no matter the location and time of day. Peak speeds are beyond 6 Mbit/s with my HSDPA Category 8 stick. And with HSUPA my uplink speeds are also well over 300 kbit/s. And this is by far not an exception, I've had similar experiences in other countries as well including Austria, where mobile broadband Internet access is very cheap and usage is accordingly.

And I am not alone with such numbers. Once a year, German telecommunication magazine “Connect” performs a test drive through all of Germany to measure voice quality in wireless networks, call drop rates and data throughput speeds. And they don't do it from a stationary location but from a driving car. And even under those conditions, they get average (!) 3G speeds in all German HSPA enabled networks of beyond 2 Mbit/s.

So why are the numbers so low in the US? It's not the technology and it can't be the iPhone either because almost everyone these days has one over here as well. So whether it's cell sizes that are to big, not enough carriers used, non-optimized radio engineering, insufficient backhaul capacity, core network congestion, etc., etc. it's difficult to tell from the outside. But there is ample proof that it can be done differently while still making a profit.

6 thoughts on “3G Speeds in the US Measured in kbit/s???”

  1. Hi,
    I have heard that in the US it is “common” that the backhaul is not dimensioned to match the air interface…
    I am not sure this fully true, but it could explain the lower performance.


  2. The report you cite doesn’t seem to jive with this one: http://www.pcworld.com/article/167391/a_day_in_the_life_of_3g.html However, I did notice a big footnote in the latter report which implies that samples showing data rates slower than dial-up were discarded (why?), which would have pushed up the averages.

    I am certainly no apologist for American cellular carriers — their crummy data performance is inexcusable when they are *netting* some $4 billion per quarter — but I think there are a number of reasons to explain the difference, including the ones you’ve pointed out. Additionally, in America, many data users are on flat-rate data plans, because that’s what the market demands. But this provides some perverse economic incentives for the carriers. If you know you’re not going to make a single dollar from a subscriber over and above his monthly tariff, then why bother to expand capacity? The most profitable strategy is to pack as many users as possible onto the existing infrastructure. Only when your customers complain too loudly or when the competitors take you to task in an ad campaign does it make sense to invest more.

    One thought about your performance results: Given your 7.2 Mbps device and your perceived performance (presumably at application/TCP payload level) of 3 Mbps, I think we can conlude that your device is being assigned somewhere between half and all of the HS timeslots. This would imply that the system(s) you’re using are extremely lightly-loaded.

  3. Hi David,

    indeed, speeds over 3 MBit/s in todays HSPA networks indicate that a cell is not overloaded. And thats how it should be for a good user experience from my point of view. In other words, cell sizes, number of carriers, neighboring cell interference, backhaul, core network etc. are dimensioned for a good user experience. It doesnt necessarily mean a network is under-used. When a cell runs into a congestion scenario for a significant time each day, its time to think about an additional carrier, another sector, splitting it up, using specific in-house coverage (e.g. in train stations), etc. It can be done and from what I can see here in Germany, it is done. Unless, of course, a carriers focus lies elsewhere as you have pointed out. I guess its a matter of where a carrier puts the congestion threshold.


  4. Hi Martin-

    Your experience implies that you’re getting between half and all of a cell-carrier to yourself, everywhere you go. To me, this can only mean one of two things: Either the German carriers are making a mind-blowing investment in mobile broadband, or the HS channel isn’t being used much. The former is certainly possible — you can make the mobile broadband experience arbitrarily good if you’re willing to spend enough money — but I think the latter is more likely.

    Unfortunately, even with TEMS, this would be hard to verify since a mobile can only decode HS scheduling events bound for itself. In other words, we can’t determine how many unique mobiles are being scheduled on a cell-carrier with only mobile-side data to go on. 🙁


  5. Hi David,

    I think in practice most mobile network operators in Europe are between the two extremes you describe. Both operator profits and the user experience are o.k. With a 3 sector site, 2 carriers and a modest cell radius a lot of ground / capacity can be covered. And by using tiered contracts with monthly allowances between 5 GB (ex. Germany) and 20+ GB (ex. Austria) you keep most of the streamers at bay that download content 24 hours a day and clog a cell. That probably helps a lot as well. As traffic of most people (except that of the streamers) is quite bursty in nature theres a fair chance you get good throughput when you need it.

    For the fun of it I did some end user performance measurements in Colognes main train station today. Its one of the busiest spots in town with lots of business travelers using their notebooks and thousands of people using their 2G and 3G phones. My mean throughput while downloading a large file was 2.1 MBit/s and it was interesting to see how my bandwidth varied +/- around 6/700 kbit/s over time, presumably when others also heavily download data and the network shared the bandwidth away.


  6. PC World recently did a study (http://www.pcworld.com/article/171458/network_woes_hate_the_iphone_not_atandt.html) in the US and determined that AT&T’s wireless network was the best. Their point was that they did not use the iPhone to measure network performance from the user perspective. They specifically said the iPhone in the US is the reason AT&T has a poor reputation for their network.

    It is not surprising that a newcomer to the cell phone handset industry does not make the best cell phone.

    As for iPhones in Europe, of course the networks there are different from the US, because the industry is different from the US.

Comments are closed.