Broadband News

Peak and off peak performance at the largest UK broadband providers in September 2017

The proposals under consultation for providers who remain signed up to the Ofcom Broadband Speeds Code of Practice makes our monthly reporting on what we are seeing in terms of peak versus off-peak speeds for the major providers even more important. Of course there are variations from day to day in provider performance and which part of the UK your connection is in but at knowing the national picture and how much speeds vary does at least help people to see if the performance variations they are seeing are par for the course or something exceptionally bad.

The new Ofcom proposal suggests monitoring peak speeds for 8pm to 10pm which for consumers if probably the most critical part of the night since this is when people are most likely to be streaming video, and especially for live streaming our single thread tbbx1 test results are a good proxy for how well or not video should stream (this is especially true for live streams where no significant buffering is possible). We know some providers say to ignore slow or very variable single thread speed test results as web access is multi threaded, but as websites gain ever larger images and javascript libraries which are individually delivered as single downloads then having a very variable tbbx1 result will be felt as web pages loading slowly. Our data currently uses a definition of 6pm to midnight for the peak period and we will explore an additional definition of peak to match the Ofcom proposals.

Off-Peak Tests Results September 2017
Off-Peak defined as midnight to 5.59pm
Median Average
Provider

tbbx1 Test
(1 download)

httpx6 Test
(6 downloads)
% differenceUpload Speed

Quality

Lower is Better

Grade A = Best

Latency
BT 23.8 Mbps 26.8 Mbps -11.2% 4.9 Mbps 0.3 - Grade A 41ms
EE 12.2 Mbps 14 Mbps -12.9% 1.0 Mbps 0.7 - Grade A 51ms
Plusnet 19.5 Mbps 22.1 Mbps -11.8% 1.7 Mbps 0.3 - Grade A 44ms
Sky 13.1 Mbps 15.2 Mbps -13.9% 2 Mbps 0.5 - Grade A 50ms
TalkTalk 12.6 Mbps 15.1 Mbps -16.6% 1.5 Mbps 0.4 - Grade A 51ms
Virgin Media 43.4 Mbps 64.6 Mbps -33.7% 6.0 Mbps 0.9 - Grade B 37ms

Peak Tests Results September 2017
Peak time defined as 6pm to 11:59pm
Median Average
Provider

tbbx1 Test
(1 download)

httpx6 Test
(6 downloads)
% differenceUpload Speed

Quality

Lower is Better

Grade A = Best

Latency
BT 26.1 Mbps 29.2 Mbps -10.7% 5.7 Mbps 0.4 - Grade A 41ms
EE 9.9 Mbps 11.3 Mbps -12.4%

0.9 Mbps

0.7 - Grade A 57ms
Plusnet 19.4 Mbps 22.6 Mbps -14.2% 1.7 Mbps 0.4 - Grade A 48ms
Sky 13.2 Mbps 16 Mbps -17.5% 2.6 Mbps 0.6 - Grade A

49ms

TalkTalk 11.9 Mbps 14.2 Mbps -16.2% 1.2 Mbps 0.5 - Grade A 54ms
Virgin Media 38.7 Mbps 56.8 Mbps -31.9% 5.9 Mbps 1.0 - Grade B 40ms

The thinkbroadband speed test has been taught about the grading system for the quality score making it easier now for people to see what we thought of their individual test, to see your quality grade and a chart of the latency measurements during the test along with how you compared to others using the same connection technology simply click the analysis button once the test has finished.

The differences between single thread and multiple thread tests at both peak and off-peak times were fairly consistent compared to the figures from August, so looks like it was pretty much business as usual for the six largest providers, alas Vodafone Home Broadband still needs to add a whole load more customers to give us enough data points to be confident enough to present their results.

We know that some people worry that the quality score favours slower connections, but the scoring is agnostic to the actual speed of the connection, so for example here is a Gigabit broadband connection with a < 0.1 quality score and another probably Gigabit connection that only scored 2.0 with a significant delay in the test starting and then the speed drops off over time. 

Comments

I would venture to suggest that sustainable single thread streaming is by some distance the most important QoS measure for throughput, as the most annoying issues are surely with streaming video. Constant buffering can make things unwatchable. Also, if the single thread works well, then multi-thread invariable does too.

Next I tend to think of responsiveness and consistently low latency. Lots of web pages are made of many small elements.

  • TheEulerID
  • 14 days ago

One potential issue with single v multi- threaded tests is that some computers are just bad at single- threading. I now own a laptop with this problem. It's only six months old and runs Windows 10 but for some reason it struggles to get to full speed on the single- threaded test.

It seems to have some kind of lengthy 'ramp up' time. My other computers in the same house don't do the same :-/

  • AndrueC
  • 13 days ago

(cont'd) leave it to download a large file and it will soon get up to full speed but the TBB speedtester just doesn't quite give it long enough :-/

https://www.thinkbroadband.com/speedtest/1507658143266629655

I've spent several evenings trying to fix it to no avail.

  • AndrueC
  • 13 days ago

@AndrueC

Have you tried any other speedtest sites? It seems that testmy.net uses a single-thread download. If that's OK then there is presumably something different between the way TBB and testmy.net are running their tests which might give a clue to where the issue lays.

  • TheEulerID
  • 13 days ago

It really worth spending several evenings over something that isn't impacting to standard use and is only seen on certain speed tests? Tad OCD :)

  • CarlThomas
  • 13 days ago

@AndrueC We could give you longer, but given we see Gig people ramp up to full speed within a second taking eight plus seconds to reach top speed seems odd.

Something manipulating RWIN and in a manner that is slower than usual, or some QoS that takes time to recognise it can give full speed.

  • andrew
  • thinkbroadband staff
  • 13 days ago

I’m struggling to understand what this pointless test is meant to prove? I’m in agreeement with TheEulerID, what is the point of measuring the difference between off-peak and peak-time speeds? Surely, if I can’t stream my 4K video at peak time then that is an ISP that is not performing well, yet you score these highly.

I’m wondering why a Virgin user would be worst off if they have a much faster throughput speed and a lower latency level?
It’s shameful that the slow Openreach network has meant we in the UK are performing far worst than some poorer countries.

  • StanAboutTown
  • 5 days ago

Post a comment

Login Register