Broadband News

Peak and off peak performance at largest broadband providers in May

With the main speed test results growing longer it was time to move the supplementary table that covered peak and off-peak speeds to its own news item, and we have made some changes in how the results are shown.

The time definition for the peak period has remained the same, but the off-peak period has now expanded from 7am to 3pm to cover midnight to 5:59pm, this increases the same size, with the intention that in subsequent months we may start to list peak and off-peak speeds for the different technologies at the larger providers, the idea being so people can see if FTTC is performing better or worse at peak times compared to the older ADSL/ADSL2+ services.

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

tbbx1 Test
(1 download)

httpx6 Test
(6 downloads)
% differenceUpload SpeedQualityLatency
BT 24.6 Mbps 27.9 Mbps -11.9% 6.4 Mbps 1.00 43ms
EE 11 Mbps 11.9 Mbps -7.6% 1 Mbps 1.06 51ms
Plusnet 21.8 Mbps 25.8 Mbps -15.6% 1.7 Mbps 1.00 52ms
Sky 12.9 Mbps 15.6 Mbps -17.3% 2.2 Mbps 1.01 57ms
TalkTalk 12.2 Mbps 13.7 Mbps -11% 0.9 Mbps 1.00 61ms
Virgin Media 37.2 Mbps 56.9 Mbps -34.7% 6.6 Mbps 1.42 40ms

The difference column is now different, since it is a direct comparison between the single thread test (akin to live video streaming) and multiple thread downloads. Previous difference figures were made between the two peak and off-peak period, but there was always the worry that the testing population may be different, we never thought they are significantly different, but the way of showing the results now makes it clearer. Additionally by splitting into two tables we have been able to add the upload and latency columns, which help to give a better understanding of any differences between the two groups of testers e.g. if upload speeds were significantly different a potential reason could be that you are seeing a different split of products.

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

tbbx1 Test
(1 download)

httpx6 Test
(6 downloads)
% differenceUpload SpeedQualityLatency
BT 25.9 Mbps 28.9 Mbps -10.4% 6.6 Mbps 1.00 44ms
EE 10.3 Mbps 11.2 Mbps -8.1% 1 Mbps 1.24 54ms
Plusnet 21 Mbps 24.5 Mbps -14.3% 1.7 Mbps 1.00 50ms
Sky 12.6 Mbps 15.3 Mbps -17.6% 2.6 Mbps 1.04 59ms
TalkTalk 10.9 Mbps 13.2 Mbps -17.4% 0.9 Mbps 1.00 63ms
Virgin Media 29.7 Mbps 51 Mbps -41.8% 5.9 Mbps 1.57 41ms

So what are the conclusions we draw from this, essentially the same as Ofcom when it has looked at similar comparisons, that the cable broadband services have a more variability at peak times, signified by the larger difference and poorer quality score. Plusnet after a few months where the system showed things were looking bad appear to have turned a corner with no major change between peak and off-peak when we looked at their results using the old comparison system.

The single thread test is very sensitive to issues such as Wi-Fi speeds, PC and router performance plus issues within a providers own network and looking back over time a drop of around 10% looks relatively normal but if you get everything right we do see people pulling high speeds from the single thread test.

The figures of -34.7% and -41.8% for Virgin Media indicate that a good number of customers are seeing worse performance and while it can be said that a service that providers a single thread speed of 29.7 Mbps should not be one that anyone complains about, with the number of people saying they cannot stream video or its constantly buffering its clear that the test is reflecting what lots of people (but not everyone, since we do see some with good speeds) are experiencing. 

Comments

Can you state what you were downloading for these tests??

EG where from(your site/ users reports/ BQM or speedtest) , what size, and what tariff.. :)

  • comnut
  • 6 months ago

Results are based on the analysis from the speed test at http://www.thinkbroaband.com/speedtest

Don't understand what you mean by what size? Or what tariff, since no tariff is implied in this item.

  • andrew
  • thinkbroadband staff
  • 6 months ago

Those numbers are ugly for VM. They have a lot of work to do.

  • CarlThomas
  • 6 months ago

I ask because my speedtest results are always around 100 Mbps (TW1)

- At work at the mo, big outage in VM due to clueless support leading dumb guys who cannot see the cable, no clue about LOADS of pictures of we gave months ago!!! :( :(

Are there any other **NON BT** 500Mbps ISPS in the area???

  • comnut
  • 6 months ago

Andrew, that is my HOME speed... :/

  • comnut
  • 6 months ago

We are fully aware that some people on Virgin Media have faster speeds, and this item is not about the top speeds, but median i.e. 50% get faster, 50% get slower and the difference between peak and off peak periods.

At peak time, 25% recorded at 26.3 Mbps or slower, and top 25% are 88.5 Mbps or faster.

At off peak these are 34.2 Mbps and 102.8 Mbps respectively.

So congratulation to those who are happy with their Virgin Media connections, but it would be remiss to ignore the figures highlighting issues that some also are reporting are impacting their experience.

  • andrew
  • thinkbroadband staff
  • 6 months ago

It interests me that despite many complaints about single threaded performance on VM, it's still on average higher single threaded performance than all the other's multithreaded average.

I mean, yeah, there is a larger delta between VM multithreaded and VM singlethreaded than for other providers, but in absolute terms VM singlethreaded is still higher than X Multi or Single threaded.

It's also lower average latency than any other provider.

Yes, I can't get 345mbit single threaded on my connection, but it's rarely below ~85, which is better than the ~55 that FTTC can offer.

  • rtho782
  • 6 months ago

Interesting, genuinely expected a bigger delta of VM results, having until last year always having used DSL as there was no VM down my road, switched over speed diffrerence is great 200mb/15mb vs the 40ish mb/5.8mb I was getting on BT, but my god WHY WON'T VIRGIN LEARN and actually build out their network with suffcient capacity before giving people more and more speed and screwing themselves over. Interestingly its not just Virgin as a cable provider suffering, a friend I game with daily over the net lives in Belgium and he also has the exact same issues on cable, another Liberty Global ISP

  • LudaLuke
  • 6 months ago

@rtho782 If the issues around single thread were just that, and not the variations in latency, packet loss and video streaming that people are reporting and people were getting a good clean single thread performance then we would be saying things like 'even though single thread is dropping significantly there is not impact on tasks such as HD and UHD video streaming since it is well above the threshold to reliably stream'.

Reality has been that the low single speeds are also impacting video streaming, and if we run tests for much longer i.e. minutes at a time we might see much wider dips

  • andrew
  • thinkbroadband staff
  • 6 months ago

Have we now finally hit the point where people are beginning to realise that a high headline speed is meaningless without quality?

  • WWWombat
  • 6 months ago

@WWWombat nope, the 40 down and about 6 up I was getting on BT Infinity 2 just doesn't compare to the 200 down and 15 up I get from Virgin Media despite a slighty higher ping and the irritance of the packet loss - I appreciate I'm lucky to have that choice however

  • LudaLuke
  • 6 months ago

VM have a few issues that need time, money and work to resolve in a few places. When they've fixed some issues more have appeared.

Not good but hopefully light at the end of the tunnel and solutions in progress.

  • CarlThomas
  • 6 months ago

@andrew I'm just going by the data you have given here, which shows VM to be a better option than anyone else.

You're not showing packetloss and jitter here.

  • rtho782
  • 6 months ago

@andrew I'm just going by the data you have given here, which shows VM to be a better option than anyone else.

You're not showing packetloss and jitter here.

  • rtho782
  • 6 months ago

rtho782, I agree with that sentiment: I have just run a a couple of tests and an getting 20-25% faster than I am paying for, and most of the time it is pretty good, but sadly VM is like Longfelow's little girl
"There was a little girl,
Who had a little curl,
Right in the middle of her forehead.
When she was good,
She was very good indeed,
But when she was bad she was horrid.
So I would be even happier if I simply got the speed I paid for AND avoided the occasional "Horrid" when VM collapses to less than 1MB.

  • Mikejc
  • 6 months ago

Mike, you're right, if I was getting below 1mbps I would be most displeased.

That said, I pay £42.25 a month for 300meg. The router is rate limited to 345m. Most things I do (Steam, torrents, Origin, Newsgroups) are multithreaded and get more than I pay for.

Those that are not (streaming from Amazon, VoIP, etc) don't need more than the 85 or so I get single threaded.

No other ISP can provide me 85 multithreaded let alone single threaded, I'd get 60 at best. Many are just as expensive, including line rental for a phone line I don't want.

  • rtho782
  • 6 months ago

The latency values given in the tables ( 40 - 60 ms) look far too high. Ofcom/SamKnows report typically 20-30 ms for ADSL lines and < 20ms for VDSL/Cable, which is what I see on my line and what one would expect given the kind of networks and technologies involved. Can't understand why ThinkBroadband average values are so high ?

  • johnmiles101
  • 6 months ago

Why so high?

TCP Latency as opposed to ICMP ping latency will have some effect. With a full Linux box you get access to ICMP, in a web browser you don't.

An average across the whole UK, and latency is not uniform e.g. Kingston-upon-Thames KT postcode median latency 34ms Aberdeen AB 52ms for FTTC when ignoring providers

On ADSL/ADSl2+ even higher KT 72ms AB 90ms

In theory corrections could be applied, but prefer to report what we find rather than attempt modelling that may introduce other issues e.g. are longer lines seeing higher latency due to interleaving?

  • andrew
  • thinkbroadband staff
  • 6 months ago

I really think you need to look at latency discrepancies. Thinkbroadband speed test on my line shows 59ms latency whereas a ping to your site (or other large sites) gives 8 or 9 ms. Your speedtest gives an accurate estimate of the line speed ( > 20Mb, ADSL) but the latency is way off. ( I'm actually surprised the ping latency is quite so small but its a very fast ADSL line with a short route to Telehouse)

  • johnmiles101
  • 6 months ago

I recommend clicking the Analysis button and looking at the pings for your tests, last one

40,40,39,39,38,38,40,38,32,250ms

That final figure brought the average up a fair bit.

In fact latency while also downloading was just an average of 41ms, i.e. there was no outlying result. That would have been obvious in the Analysis view

  • andrew
  • thinkbroadband staff
  • 6 months ago

Post a comment

Login Register