So lets talk broadband speeds, we have in the last couple of months (Feb 2013) launched a new broadband speedtester that now uses multiple tests to help gather more information on broadband speeds in the UK and also help people difference understand how their connection is performing. Possibly the most common question for broadband is are my speeds OK? At which point everyone rushes for a speed test results table such as the one below:

Broadband Provider HTTPx6 Average Upload Average
Virgin Media 38.4 Mbps 3.3 Mbps
BT 23.8 Mbps 6.2 Mbps
Plusnet 21.3 Mbps 6.1 Mbps
Sky 12.2 Mbps 2.2 Mbps
TalkTalk 10.5 Mbps 1.5 Mbps
EE / Orange 9.3 Mbps 1.7 Mbps
O2 / BE 7.6 Mbps 1 Mbps

There are some impressive speeds in that table, and it does reflect what you expect if you are looking at the big picture. The issue is that it is comparing the providers en-masse and not directly comparing products, and if you look at the following composite image, it will become apparent that the reason BT Retail has such a high average is that it has the lion’s share of FTTC products in the UK.

Click image to see results from all 7 providers clearly

Speed test results from thinkbroadband speed test
 (Y-axis identical in all graphs 140000 Kbps down to 0 Kbps) 

 

So what do we learn from these graphs, mainly that fibre (FTTC) services are more popular with Plusnet and BT customers who visit our site than with Sky and TalkTalk, something that is reflected in the financial reports from the operators and the simple reality that around 9 out of 10 FTTC customers in the UK are with BT or Plusnet. An interesting aspect is that you can use the upload speeds to potentially narrow down which broadband product a customer is using, this is because upload speeds appear a lot less affected by factors such as congestion, line lengths, wireless speeds etc.

virgin-media-results

 

The stepped nature of the graph for Virgin Media shows the capped nature of the 120, 100, 60, 30, 20 and 10 Mbps tiers that are in operation with the provider, as their speed upgrade programme progresses we should see the long tail of people with 10 Mbps and 20 Mbps connections vanish.

sky-results

The FTTC products offer three upload speed variants 20 Mbps, 10 Mbps and 2 Mbps, our speedtest results suggest that the 2 Mbps version is pretty rare, but as the graph for Sky above shows while they sell the 80/20 FTTC service as their premium product, those who have ordered it are seeing very good upload speeds. The downstream speed curve between 38 Mbps and 18 Mbps reflects the variation in connection due to line length from the street cabinet.

o2-be-results

O2/Be would seem to be the poor man of the group, with such a low speed, but their plot does very nicely show the performance range for an ADSL2+ service and how consistent the upstream speeds can be. Certainly when ADSL2+ was the premium technology the BE service did offer some of the highest connection speeds both for download and upload.

So what for the future? Well we will return to look at how the various providers are performing, particularly as it appears to give a reasonable indication as to what is actually selling. The graphs for BT and Plusnet show a glimpse of the future with a handful of testers heading above the 100 Mbps mark and even above 140 Mbps as parts of the UK get access to FTTP services, with just 200,000 properties in the UK with access to an FTTP service at this time it is difficult to get meaningful data but we expect that to change as the year progresses.

plusnet-results

For the critics of Openreach and their FTTC heavy superfast broadband roll-out, the curves from people using the up to 40 and up to 80 Mbps products tend to undermine statements that the service is useless, particularly if you compare the performance difference it makes compared to O2/Be.

All the graphs so far have shown the downstream results from the HTTPx6 test we now use, which uses six HTTP downloads to flood a connection. If we compare the single thread test versus our newer HTTP test the correlation is still pretty good, and certainly we plan to continue to run both tests.

tbbx1-versus-httpx6-results

The reason for running both tests is that it can be used to assess whether an ISP is suffering congestion, i.e. if we see a big difference between the two across the board for an ISP we may be able to infer a widespread congestion problem, and when looking at an individuals connection you can actually spot whether the connection was busy.

example-of-local-congestion

In the sample button from our speedtest, you can see that the graph for the single thread is consistently slower than the HTTPx6 test. This is because when this test was carried out, we were also streaming a HD stream at around 3 Mbps. The average figures are pretty straightforward and such that for the HTTPx6 average this speed was the average over an 8 second period (the same test time is used no matter how fast your connection is), the burst figures approximate to the speed your connection was seen to manage for just 1 second, generally we always quote the average figures as this reflects much better your ability to stream and dowload material. The duration of the TBBx1 test will vary from test to test, but is usually in the 7 to 15 second timeframe.

We all love a conclusion after a long read and it is time to make one, perhaps it should be this, currently amongst the big providers the difference in speeds between them is less about the ISP and more about the physical connection you purchase, i.e. an 80/20 FTTC connection from BT will perform very similar to one from Sky and TalkTalk. The differentiating factors are much more price, level of support and the other less tangiable factors.

 

Tags: , , ,

9 Responses


  1. PhilT on 12 Mar 2013

    Bit slow today – what is the x-axis on these graphs ?

    • andrew on 12 Mar 2013

      The graphs are speed profiles, so you if you took a point half way along the x-axis you could cross reference to find the median speed for any provider.

      • PhilT on 12 Mar 2013

        Thanks, speed test result distribution curve. Got it. You should have put the OFCOM / ASA 10% line on it :-) or at least some tick marks at 10% intervals.

        • andrew on 12 Mar 2013

          Next time I will. This blog was about starting to make more information available from the speedtest data we have.

          Plus it also means we have a nice historical data set to verify the claims made by the politicians in 2015.

  2. Michael on 23 Mar 2013

    I think it’s wrong to calculate an average speed using different “modes of transport”. It’s like saying the average speed between London and Birmingham is 120 mph when this is made up of speeds achieved by road, rail, and plane. How can I figure out how my ADSL2+ line compares with ADSL2+ line uswed by others?

    • seb on 23 Mar 2013

      People have a choice between taking the car or train when heading from London to Birmingham, and many people have the choice of cable versus fttc versus adsl2+ services; so comparisons are still useful, but you’re right.. comparisons between providers on the same technology are still also important. The problem we have is many users don’t know what service they’re on – our detection of ISP is also more accurate than the user’s recollection of who their provider is :)

      • Michael on 25 Mar 2013

        If only I had a choice. I’m on an ADSL2+ line. There’s no cable or fibre in my road. According to my account on the ISP web site my line is enabled “up to 16Mbps”, my estimated speed range is 2 to 4 Mbps, and my typical download speed is 3 Mbps – all very honest; my sync rate is 4,763 Kbps, my upload speed is 867 Kbps. My monitoring over many months shows that on average I achieve a download rate of 3.6 Mbps with peaks at 4.2 Mbps. Now is that excellent, average or poor, should I switch?

        • andrew on 25 Mar 2013

          Whether that is good or indifferent depends on whether the 4.7 Meg sync is roughly right for the attenuation of the line.

          The ISP may have managed the line to a slower sync to improve stability.

          We can probably do similar plots like this blog, but excluding the users on faster services. That would let you see where you sit with respect to others who use ADSL2+ based services.

  3. Michael on 26 Mar 2013

    Line stability is very important to me and I believe that the ISP has synced my line accordingly. But, what I think this also means, is that other ISP’s would be in the same boat and unable to increase the sync rate without sacrificing stability. So there’s a trade-off which we are not measuring. There seems to be no evidence that changing ISP would improve both my line speed and quality.
    Thank you for your views.


Leave your comment