Fibre Optic Speed

When I test for the speed of my TalkTalk Fibre optic connection using Speedchecker.co.uk (and others) I get a download speed of around 26Mb/s - no real complaint there. BUT when I use speedtest.net the figure is around 47Mb/s. Can anyone explain the difference? Afaik the units used are the same.

Cheers

MerC

What does http://www.thinkbroadband.com/speedtest.html show? In particular when its finished there will be a link you can share on the web page.

Post that and we can see the shape of the graphs which might help answer the question.

That seem to have sorted something out in that it gives me TWO download results :
Avg TBB = 21.4
Avg HTTP = 46.1

(these results are compatible with my previous figures)

I'm assuming this means the speedtesters I have been using actually measure different things. I should like to know what the difference is though.

On that, what was the burst speed against the 21.4Mbps one?

Ideally Andrew/we need the link to the graph button, so we can see what shape it is .

Edit - Like this.

Edited by RobertoS (Thu 21-Mar-13 20:37:32)

It may be that your connection is busy doing something else and the HTTPx6 with six threads will tend to overwhelm the other activity, or the ISP has congestion.

Or it may be you have an RWIN limiting your speeds for each individual thread

http://www.thinkbroadband.com/speedtest/button/13638...

That's not the "Direct graph" button. You want the first link below the result. But taht does seem to agree with Andrew's suggestion it might be the RWIN setting.

Assuming you are using Windows, which version of it please?

http://www.thinkbroadband.com/speedtest/results.html...

is the link directly below the result display.

I'm using Windows 7 Home Premium 64bit SP1 build 7091 with IE 9 32 bit

Lines are too flat too be a busy line or ISP congestion

Latency is OK too and OS is not XP so RWIN should be automatic.

So in short don't know why single thread is slow on that PC

Brilliant. (I got it wrong! I didn't meant that one, I meant the one labelled "Direct Graph Button", but it's included in the one you just gave ).

That's really wierd, as Andrew says. Are you testing wired or wireless to the router, but that shouldn't affect it either.

My connections are wired, although I do have wireless connectivity if I wanted to use it - as for my mobile. I take it from your replies that there's nothing really I can tweak. Seems odd that speedtesters (apart from yours) test either one thing or the other and don't say which it is.

Tx for the help anyway.
Cheers

MerC

Keep an eye on this thread for a few days. Someone might come up with an idea.

ISP traffic management ? Do you see this on other TT fibre results ?

Don't recall seeing it, and too late to go and dig through the various data sets I have now

In reply to a post by yarwell:

ISP traffic management ? Do you see this on other TT fibre results ?

Is there any chance the OP can test at different times of day showing us the graphs with the different results in it?

I have two theories:

The first being that your circuit is set so that one thread is equal to your IPv4 TBB speed so it will be impossible to max out the connection with one thread. Speedtest.net uses two threads (four threads sometimes if needed) hence the higher (double) speed with their tests.

The second is that TalkTalk are using traffic management as previously suggested. Completing tests at different times of the day should rule this out.

Edited by chris6273 (Fri 22-Mar-13 01:19:38)

In reply to a post by MrSaffron:

So in short don't know why single thread is slow on that PC

Just a thought but I had a problem like that with my ISP last year. It turned out to be a configuration issue that occurred after they updated one of their boxes. I don't have any more information than that but here's what tech support said:

"Thanks to input from Andrue we have identified a possible issue with one of the LNS servers so we're going to reload it at 04:30 tomorrow morning. Consequently those customers currently connected will see a brief PPP drop as your connection is shifted to one of the other LNS servers. "

That is a possibility, some nifty analysis on a few weeks worth of test data might help.