Speedtests - BT versus Thinkbroadband

After my broadband speed dropped to less than 0.5mbps a couple of months ago, I checked out my ISP (freedom2surf) help section and was told to use BT or Thinkbroadband to do a speedtest before contacting technical help. I decided to try both. Much to my amazement, BT and Thinkbroadband results were vastly different. BT says I was getting 1.5mbps and Thinkbroadband says I was getting about 300kbps.

I then call my ISP technical help and was told to run another few speedtests using BT and I compared them with Thinkbroadband. The speedtests, they told me, should be every 6 hours and I should do this for a few days, which I did (what a mug).

The days dripped passed and my broadband was so slow that I couldn't even watch the news on BBC iPlayer or listen to any music without experiencing the dreaded buffering. I then phoned my ISP technical help again. They suggested I do some more speedtests, but this time they asked me to remove the cover from my BT socket on the wall and plug into the test socket inside the caseing as, (they say) BT require this to be done. OK, the mug that I am, I ran some more speedtests. I ran the comparison speedtests using BT and Thinkbroadband for another month. I then sent my ISP tech help (by this time it has changed to Opal help) a support ticket with more than a dozen results (and links) from BT and Thinkbroadband, and drawing special attention to the date and times of these tests, as many were done back-to-back and with the results being so vastly different. My ISP help replied by telling me to only use the BT test... and could I run some more, as (they claim) BT will only take the complaint seriously if I use the BT test results.

Patient as ever, I ran some more speedtests. BT results were in the 2000k-2500k range while within a few minutes before and after, Thinkbroadband results were less than 500k.

Can anyone explain to me what makes these test results be so vastly different when tests were run back-to-back?

Also, can anyone suggest a decent ISP who has decent speeds and decent helpline with decent human beings in charge?

Here's a short list of my speedtest results from thinkbroadband, for anyone who's interested...

http://www.thinkbroadband.com/speedtest/results/id/1...
http://www.thinkbroadband.com/speedtest/results/id/1...
http://www.thinkbroadband.com/speedtest/results/id/1...
http://www.thinkbroadband.com/speedtest/results/id/1...

All those are on the default port 8095. Have you tried speedtests using the "Port 80" option? It could be the ISP are either prioritising standard port 80 traffic or shaping port 8095 traffic. This can make a big difference and show the difference between a slow connection and a traffic controlled connection.

I used to think that ian, but it's not necessarily fool-proof.

The point being the ISP may be prioritising http traffic, which normally comes in on port 80. But the tbb speed test, even when using port 80, is not http.

Try some known file-size downloads. For example here.

Thanks for your advice. I downloaded a 50MB file using tbbMeter and the result was the following... Download time: 1102 secs. Rate:48KB/sec (380kbps). I don't know which port tbbMeter used, just the default one I guess.

I'll wait until my off-peak time (1-5am) then try again using various ports.

Cheers.

Several ISPs have similar I think, perhaps Zen. TBB has more than one method - do you have tbbMeter?

The thing is ISPs can fairly easily prioritise the BT speed test .

The test files in tbbmeter are all HTTP based, default port 80.

Try the multiple thread option, a common simple traffic management technique is to limit each thread to a certain size.

Also with some providers the BT Speedtest bypasses some of the ISP network, so if nothing else is the cause, it might point to ISP congestion.

While I was able to get the maximum out of my 8Mbps connection, 7150 profile, in normal use, I was getting a similar situation with regards to speed tests on this site, and some others, compared to download tests and the BT speedtester site.
If I used my download manager to download a test file from here, I could get an average of 850 KBytes a second download, the tester would give a maximum of 200 KBytes a second, the BT Speedtester site was in the 600 to 700 KBytes a second range.
My download manager allowed me to make multiple connections for the download, so I restricted it to a single connection, result was that maximum throughput on the download was about 200 KBytes a second.

This meant that the sites which were showing slow results used a single connection for the test, where I got a high result there were multiple connections, my slow down was due to a single connection being throttled to about 200 KBytes a second, but where was the throttle ?

I tried a different router, I usually use a customised IPCop build, same issue on Voyager 205 router.

That left the local machine or the ISP side being the problem. I did some checking into the type of issue and found the TCP/IP analyser at http://www.speedguide.net/analyzer.php

This told me that the maximum throughput the machine would allow is about 200 KBytes a second due to the TCP Receive Window (RWIN) settings within Windows. There was info on solving this, and they also have a tool to download that allows you to tweak from an interface, rather than have to hand edit registry entries. I have set this to optimal, according to the site, settings for my connection, and now the speed test on this site, a single connection download test and the BT Speedtester all give similar results, 800 KBytes a second or so, when done at the same time of day.

May be that the buffering you are getting is due to single connections being in use and the machines current own limit for a single connection being reached. It might be worth trying this test to see if it is possible.

In reply to a post by intro2:

Thanks for your advice. I downloaded a 50MB file using tbbMeter and the result was the following... Download time: 1102 secs. Rate:48KB/sec (380kbps). I don't know which port tbbMeter used, just the default one I guess.

I'll wait until my off-peak time (1-5am) then try again using various ports.

Cheers.

Did you get any joy with different ports?

The problem with speed tests is that they are only 1 moment in time. This packet loss tool will send pings to your IP address and every hop in the ISP leading to it.

Although it is not testing speed, you will get a much better idea of what is going on and you can test over an extended period of time. Look for the hops in the network that have packet loss and latency from the previous hop. Those are likely the routers that are killing your speed. This tool is based in the US so ignore the latency of the hop where it is coming overseas.