Thinkbroadband TCP latency test

I find Thinkbroadband speed tests the most reliable, especially for upload speeds where many others seriously underestimate.

However, its TCP latency results are always rather curious. I usually get values around 15-20ms (unfortunately Speedtest history in my profile does not show the latency results), and just got a value of 5ms. However, when I use 'ping bbc.co.uk' I get much more consistent results. There is usually only 1ms difference over a group of results, and the range for tests taken at different times is almost as consistent (currently 9-10, typically 10-11, sometimes 11-12).

Thinkbroadband tests are still more sensible than most others (eg just got 34ms on BT wholesale test, which is not atypical). Speedtest.net is better, giving me 14ms at the moment; for some reason it is choosing a Sheffield server, which typically adds around 4ms.

I'm not worried at all by these results. Just interested in why there is so much variation when ping is so consistent.

Because we are not doing an ICMP ping, it is a small TCP packet, so testing two different things.

Latency of a small TCP packet is pretty important in a world where a webpage may be fetching lots of small updates now.

This is why we call it latency and not ping

Thanks. Still seems odd it managed to get 5ms less than ping time, but that was just a one-off.

Is this small TCP packet via HTTP?

I just installed Privoxy on my router (and force all HTTP traffic through it) and noticed that the new speedtest dropped my download bufferbloat score from 'A' to 'B'.

If the TCP packet is via HTTP then it probably explains the change because there will be a slight increase in web latency as Privoxy attempts to pattern match the URL.

Currently, but subject to change at anytime