The Fastest ISPs In the US

adeelarshad82 writes "PCMag recently put Internet browsing speeds to the test to see which ISP was the fastest. The results were based on a quarter million tests run between May 1, 2009, and April 30, 2010, by more than 6,000 users. The tests were carried out using SurfSpeed, which takes into account the complete, real-world download time of a web page to a browser. According to the results, Verizon's FiOS took the top spot as the nation's fastest ISP, with a SurfSpeed score of 1.23 Mbps. Interestingly though, of all the regions where Verizon's FiOS is available, its dominance is only seen in the northeast and the west, whereas cable service from Cox and Comcast won out in the southern region. Moreover, cable through Cox and Optimum Online beat AT&T's fiber optic service in the nationwide results, with SurfSpeeds of 1.14Mbps, 1.12Mbps, and 1.06Mbps respectively. The worst results mostly consisted of DSL providers, bottoming out at 544 Kbps from Frontier and going up to 882Kbps by Earthlink. Other interesting facts noted in the test were that broadband penetration was highest in Rhode Island and lowest in Mississippi, while the average Internet bill was highest in Delaware and lowest in Arkansas."

Re:Only 1.23 Mbps?

[ffejie]

From the article: "Keep in mind, when it comes to the speeds reported in this story, SurfSpeed takes into account the complete, real-world download time of a Web page to a browser. We're not saying your own ISP's claims of double-digit megabit-per-second (Mbps) throughputs or more are false. But those are marketing numbers, based on direct downloads from their own servers, using some abstract math like the number of users divided by the theoretical line speed. The numbers in the SurfSpeed tests compare everything you get in the download of a Web page, not just a single, contiguous file, so the numbers are smaller than the data-rate numbers quoted by your ISP. They provide an example of the real-world throughput you're experiencing when you browse and with speeds comparable to what others customers of the same ISP would get."

But we wouldn't expect you to read the article.

In my opinion, there is a lot of ISP fraud.

[Futurepower(R)]

VERY good question. QWest in Portland, Oregon is currently advertising 40 Mbps. There is, however, very fine print saying "Connection speeds are based on sync rates."

Of course, QWest knows that most people won't understand that. QWest is saying that the advertised speeds are only the speed that the customer's modem synchronizes with QWest's equipment. The actual speed that QWest supplies data over the internet can be anything QWest likes, with those fixed synchronization speeds.

The same ads call the service "Fiber Optic Fast Internet". The fine print says, "Fiber optics exists only from the neighborhood terminal to the internet." That means NOT to your house or business.

The quotes are transcribed from an ad I have on my desk.

Re:Only 1.23 Mbps?

[Anonymous Coward]

This may come as a shock, but iTunes and SurfSpeed are not the same thing! It's almost like comparing apples to oranges doesn't give meaningful results!

Re:Only 1.23 Mbps?

[adolf]

It's not about downloading -- it's about browsing. The question is not about "how many bits can one shove through this pipe," but instead "what is a quantitative measurement of the actual speed one can expect when going clicky-clicky on links on web sites."

So instead of maximum aggregate speed (which is easy to determine with speedtest.net and the like) this "Surfspeed" figure includes latencies for things like DNS. Round-trip times. Route lookups. Geographic caching (Akamai). The time it takes for the geolocation service to figure out where you are. Hops to the host(s) in question. Congestion of those hops. How long it takes for the fucking ad servers to wake up and start spitting out ads.

Should any of that matter? Of course not. But over here in the really real world, things aren't perfect, and it all makes a difference.

Get it? It's not at all intended to be an idealized measurement of maximum throughput.

To use a car analogy: Given a selection of different vehicles of different performance characteristics, how long does it get a bushel full of DVD-R from point New Jersey to San Francisco, including refueling, maintenance, personal needs (more comfortable cars == less stopping), road conditions, weather, traffic, and dodging kids on bikes?

It's easy to come up with an idealized [yahoo.com] route and ETA. But it it's much harder to include some real data [google.com].

And all of that theory is meaningless compared to actually measuring how long it takes a given vehicle to do that job, which is what this Surfspeed measurement tool proclaims to do.

Re:Latency more important than bandwidth

[raddan]

It really depends on the application. I recently went over this as we surveyed the network capabilities of 450 of our field reps in order to determine whether doing virtual meetings was feasible, i.e., something like WebEx. With an application like WebEx, once you meet the minimum bandwidth requirements (roughly 700Kbps down and 300Kbps up for the kinds of meetings we were looking to do), latency is indeed the most important factor. Call quality deteriorates fast when you're looking at 100ms or greater RTT. WebEx also will "fail" into using TCP if it cannot establish a UDP connection, which means that it suffers horribly on wireless connections, where dropped packets are common.

But other protocols, e.g., rsync, which was specifically designed to avoid RTT costs, perform quite well on high-latency network connections, by minimizing round-trip communication. In that case, bandwidth is the most important measure.

BTW, our survey showed Verizon coming out on top by a hefty margin. On average, FiOS users got about 15Mbit down, 7.5Mbit up, and under 10ms latency, with some being quite a bit higher. Of course, offices with Cogent fibre connections trashed everybody, but that's not really surprising-- our test site was running on Cogent, too.

speakeasy are LIARS

[ClioCJS]

They say "unlimited", then they kick you out if you actually dare to download more than 100G. And they lie about it in pre-sales: http://www.flickr.com/photos/clintjcl/76331293/ [flickr.com]