Internet providers caught inflating speed test results

ADVERTISEMENT

So why is it important to have fast throughput over a single connection to a server? Most video streaming services such as YouTube operate over a single connection to the content server. This means that the Internet connection must be able to sustain that video bitrate over a single connection, otherwise periodic buffering will occur.

ADVERTISEMENT

For example, a consistent 20Mbps test result on speedtest.net does not mean an 8Mbps video will play without periodic buffering, as the connection may only deliver 5Mbps over a single connection. With 4K video streaming set to take off with the growing number of sales of 4K TV sets, many ISPs will need to deliver considerably higher throughput over a single connection to reliably stream 4K video.

The following is an example of a 4K 60FPS video stream playing on YouTube which peaked at nearly 35Mbps and did not drop below 24Mbps during playback according to DU Meter. According to our senior reviewer Wendy that streamed this video, it consumed roughly 24Mbps on average and did not stutter once during playback with her BT broadband VDSL connection due to it maintaining a significant buffer ahead of playback:

YouTube 4K bandwidth consumption

ADVERTISEMENT

An ISP that cannot sustain this bandwidth over a single connection will result in YouTube dropping to a lower resolution or periodically buffering if forced to play in this 4K 60FPS mode.

For anyone that would like to check if their connection is fast enough to stream 4K and has a 4K screen, the following is a link to this video:

https://www.youtube.com/watch?v=iNJdPyoqt8U

Conclusion

Based on checking how Ookla’s speedtest.net connects to its server, it does so with multiple parallel connections to its test server over port 8080, a port number rarely used by real world web traffic. As we demonstrated, some of the ISPs we checked including two mobile operators delivered substantially faster throughput over port 8080 than the regular HTTP port with just a single connection to the same test server. However, many other ISPs did not show any variation at all between the regular HTTP port and port 8080.

One ISP we checked was only able to deliver its advertised 38Mbps throughput when multiple connections were made over port 8080, closely matching the network activity of Ookla’s Speedtest. However, when we repeated this test over port 80 to match real world conditions such as a loading a multimedia-heavy webpage, we only achieved 6.5Mbps.

Does the VW’s emissions scandal ring a bell?

We’d also like to hear if our members experiences reflect the above, so if you’ve any interesting results, or you’ve discovered another ISP that isn’t playing fair, please post your results here.

ADVERTISEMENT