Poor speed test results since the site updated

I used to get reasonably accurate test results before you updated the site But now my results are consistently between one third and half half of my 300Mb.

I can use another site and select the Birmingham server and get the correct results
So it would appear that it is not a Pipe issue!

If in news Groups I consistently get over my 300Mb

and My Du meter program agrees with all the tests!!

Your tests such as http://tbb.st/1494080973874374355
http://tbb.st/1495068447724873855

All have indications that their is some contention on the connection, and in particular last week Virgin Media had a major issue with any sites linked to them via LINX

While the latency issues have been resolved it appears, it is highly probable that the path before it reaches our network is still congesting.

The new test is more sensitive to congestion, which is the whole point of running a speed test, i.e. to give people feedback on how a connection is performing. Some testers that exist are known to peak/burst speeds only, or give speeds beyond the actual connection speed.

The site change meant we went from the single thread been a custom protocol to plain old HTTP, and if a provider is having http issues then it might reflect that.

We can see other people with Virgin Media on the some browser getting 200 Mbps speeds (and look like that package based on upload) but with wide variations in the single thread.

Tester is certainly not broken e.g.
http://tbb.st/1494779969612608855
http://tbb.st/1494989676283855755
http://tbb.st/1495307612486934655

So given the known issues with Virgin Media and complaints that actually do mirror the performance our test is indicating happy to say the tester is working well.

No
This is More Accurate!
SpeedTest Result

As opposed to this Poor excuse

24-05-2017 00:02:30 (GMT)
150.7 Mbps
20.6 Mbps
Virgin Media

Did you actually read MrS's post? The fact that some of Virgin's interconnects are known to be running hot and therefore the interconnect they have that is used for the TBB traffic is congested whereas other interconnects are ok? It is more likely that it is down to the known issues with VM than a problem with the tester. You have to remember that the speed is testing everything between you and the server and that the paths taken are not the same for all destinations.

Yes I read the reply

But Still no answer to why this has ONLY occurred since the Site Revamp!

My quality Monitor has only showed Latency issues on 1 Evening in the Last 2 weeks
Just more Virgin Bashing and that Lasted for 2.5 Hours

and I don't work for them! Just a Loyal Customer of 19 Years

MrS also said the new tester is more sensitive to congestion. The Virgin problem only surfaced fairly recently on some of their links. I am not bashing Virgin - it just happens they have had a problem with some interconnects, it happens to other ISPs too it just happens to be Virgin's turn at the moment. This even points to a comment from Virgin made to ISPReview confirming they had seen an issue and MrS is postulating that the congestion could be continuing signs of that.

I have no axe to grind with Virgin, never been a customer of theirs although it is possible I might be in the future as lightning may be coming to my area.

Ookla system operates on a find lowest latency server system, and if provider does host data sources it will use those, so it is entirely possible you are testing to somewhere inside the Virgin Media network only.

This is why Virgin Media prefer to recommend ookla testers as they know it will almost always give a higher speed.

Are we wrong? No - but reflecting a reality where not everything is perfect. If a provider has reason to believe our tester is unfairly showing slow speeds they are free to talk to us and on network solutions are available, but to avoid skewing national statistics tests on those solutions do not feed into the national picture.

Surely on a 300Mbps connection a throughput speed of 150Mbps is more likely to be correct than one of 331Mbps?

We also need to see the graph link please from your thinkbroadband test. It runs two different downstream tests following each other, and the shapes of all three graphs are an important part if diagnosing line/ISP performance.

Ookla does not provide that level of data, and is well known to discard extreme spikey values. Which makes the 331 even more suspect, as it suggests there were consistent readings above the maximum possible.

NOTE: Virgin Media over provision by 10% so a speedtest of 330 Mbps is possible on a 300 Mbps connection.

The ookla algorithm is drop 30% of slowest samples and 10% of fastest samples and then deliver an average.

We store the graphs, and show the median

Thanks. I didn't know about the over-provision. We still need the poster to link to the graph.

Edited by RobertoS (Wed 24-May-17 09:20:55)