In an earlier post I did fully accepted that contention is the most likely cause but I am still open minded to other possible issues just as I was when as a test and design engineer and a trouble-shooter for complex control systems I never closed my mind to all possible causes of a defect and sometime to what others had considered were impossible causes. - (With intermittent faults I mostly never saw the actual fault arise but after I left the recurrent fault had all gone).

It is now past 4:30AM in the early morning and I have just carried out a speed test that does not support the theory that it is all down to connection and nothing else.

https://www.thinkbroadband.com/speedtest/16023866981...

At this time in the early morning contention should not be an issue since most sensible people, (who do not have a fault), are in bed but the speeds are worse now than they were last evening. The speeds of 438Mbps down and 32 Mbps Up are nowhere near as bad as they have been previously but are still below the BT Minimum Threshold and are lower than they were mid-day yesterday.

Thank you again and sincere thanks to all contributors for your assistance.

Regards,
Fido.

.

You are all confusing contention with congestion. https://support.aa.net.uk/Congestion_and_Contention

Are you sure your computer and router can actually transfer data at 900Meg?

This may not work for you but i was also getting similiar speeds of around 300-400mb on the 900 package, so changed my ethernet from the ONT to router for cat7 instead of cat5e and now getting max of 930Mb/115

Although they both support 1gig the cat5e was probably down to being defect? who knows.

They are mostly consistent i would say about 80% of the time with the lowest being around 680

Edited by deleted (Sun 11-Oct-20 15:12:21)

That the graphs never flatten out suggests maybe the RWIN is not scaling correctly on the computer, sometimes a side effect of tweaking for previous speeds.

If it was contention, you would expect more peaks and troughs rather than a line that is trending up.

Have ran fast.com before and have seen 1.1 Gbps on a 900 Mbps connection which also has a Gigabit Ethernet connection to the router, so as always just because it gives a nice number it is not necessarily correct.

The Fast.com speed test was the one that the Openreach engineers were using too. Will the TBB speed test always give correct figures?

"Correct figure" is a difficult term.

Every actual speed test from any source is only a snapshot at the time of the interaction of many variables over which the test provider has no control. The provider minimises these as much as they can, by (at a guess), two main methods. Installing more throughput and processing capacity than they can reasonably expect, and setting up peering links to the main centres and ISPs in their own country.

If you have ever used speedtest.net for instance, the server it chooses automatically for you by testing a few latencies to you I find almost always gives me a lousy actual speed. It also rarely chooses the same server on two different days.

I use one I choose manually that I find consistently gives me high speeds, as it has to be nearer to "correct" that the lower ones. I also use the thinkbroadband one, and am pleased to report that when the two are used within minutes of each other they are generally in close agreement.

Even when the server company has the best possible setup as described above, the test data still has to go through your ISP and many intermediate routers all with immense amounts of other traffic to handle. For instance, extreme cases are when a Wimbledon final is on or a popular figure is playing, and a Formula 1 race is on, or the common clash of the Boat Race some other sport occurs, a speed test is likely to be poor.

Though the major ISPs are these days installing orders of magnitude more throughput capacity as the cost of it per gigabyte has dropped dramatically. So the effect of such clashes is decreasing.

There must also be a limit to the bandwidth available to the server that the TBB speed test runs on.

I didn't exclude tbb from the above. See the first "main" paragraph of the post.

I used the speedtest.net example to illustrate my point, because it is the main source with multiple independent businesses supplying the service. It isn't a single provider with multiple servers.

Every server has a limit, difference is we run the network the server is on and can be sure the links we control and data server is not a congestion point.