Could be a Java7 and ipv6 issue: this copied from a post of mine on BT forum (http://community.bt.com/t5/BT-Infinity/Warning-java-7-and-BT-speedtest/m-p/558115)

Further exploration, and comments on http://community.bt.com/t5/BT-Infinity/Speedtester-b... showed me that my Java 7 theory wasn't completely correct. It worked fine for some people. It seems there are complications if you have VPN connections or other network adaptors (VirtualBox in my case). I got the hint for the V6 issue and possible solution from http://jonathangiles.net/blog/?p=1126 I tried disabling ipv6 in the various network connections as suggested, but no help. The solution for me was to change the default Java7 ip stack.

One way to apply this patch is:
- make sure no Java running
- go to the Java control panel
- Java tab
- View ...
- insert -Djava.net.preferIPv4Stack=true under Runtime Parameters
- close View and java control panel

I'm not sure if this will automatically be undone when you next install another Java. Any suggestions for appropriate semi-permament alternative patches welcome.

This fixed both the BT retail speedtester (http://speedtester.bt.com/) and the wholesale on (http://speedtest.btwholesale.com/) for me. As mentioned elsewhere on this thread, neither of these is very good as a speedtester, but they are useful to find your ip profile.

Edited by StephenTodd (Tue 10-Jul-12 16:07:43)

Aha, that link gives a very useful bit of information

A couple of the posts refer to another tester at http://speedtest.btwholesale.com/ , I tried that and, if you select "Further Diagnostics" after the first test has completed, it runs the "usual" speedtest and gives your IP profile

The one at: http://speedtest.btwholesale.com/ is, I believe, the Beta Tester initially, which notoriously over reports - 41Mb on a 38.72 sync, and the additional is the original standard test. It can also be run more frequently than once an hour - every few seconds for the first part and every 10 minutes for the second which requires the phone or circuit number.

http://speedtest.btcentralplus.com/ and http://www.speedtester.bt.com/ are the original performance tester. The tester is hosted at several location (3 or 4) and either URL could take you to any one of them.

In reply to a post by MHC:

and the additional is the original standard test.

Yes, but if accessed direct it has Java problems on a Mac, if accessed via the first one it works.

I'm not usually interested in the speed, it's my profile that I want to get!

It can also be run more frequently than once an hour - every few seconds for the first part and every 10 minutes for the second which requires the phone or circuit number.

Nope, the second one still won't let me run more than once per hour.

Have just tried it:

First Phase - 38.89 Mbps (I also have a background application running with a continuous 200k feed that is not interrupted) so teh true figure would be higher still.

Second Phase - 38.89 Mbps with a maximum achievable speed (profile) of 38.72 Mbps but interstingly, no actual download showing in either Task Manager or TBB meter.

Then tried again:

First Phase 39.37 Mbps

Second Phase: 10 minutes

In reply to a post by MHC:

Second Phase - 38.89 Mbps with a maximum achievable speed (profile) of 38.72 Mbps but interstingly, no actual download showing in either Task Manager or TBB meter.

I think the second phase just interacts with the BT/OpenReach database, to save the result of the first phase against your phone number (for future diagnostics), and to read the ip profile (for presentation to you). For many of us it is just the ip profile we want, and we would happily skip the actual upload/download part if we were allowed to.

Edited by StephenTodd (Tue 10-Jul-12 17:33:12)

Odd... the first time I tried a repeat run it said there was a minimum of 60 minutes between tests.

Just tried it again and got the same as your link- a minimum interval of 10 minutes. The pages had slightly different layouts- it wasn't the same page with a different figure.

Depends on the luck of the draw as to which server you get routed to I suppose.

Edited by billford (Tue 10-Jul-12 17:35:59)

In reply to a post by StephenTodd:

For many of us it is just the ip profile we want, and we would happily skip the actual upload/download part if we were allowed to.

Very much agreed, especially for FTTC (and FTTP?) where you can't easily get at the modem stats and the profile is simply a percentage of the sync, and changes immediately.

It would make it vastly easier to see if a re-sync (user-initiated or otherwise) had improved things or made matters worse.

Edited by billford (Tue 10-Jul-12 17:44:22)

Good news, of a sort.

The Java IP stack fix made no difference to the retail test site, it still fails on my PC with the same error at the same point, and does not actually send any data upstream.

However, the wholesale test site works like a dream (or is that nightmare?), and returns results and IP profiles for both up- and downloads.

For the record this is following a 21CN upgrade to my exchange a couple of weeks ago, nothing exciting like fibre.:-

Download (result / IP Profile / Max possible Mbps) 11.85 / 11.92 / 21
Upload (result / IP Profile / Max possible Mbps) 0.86 / 0.83 / 0.83

Note that upload result is higher than max possible.

My router (Netgear DG834 v2) shows:-

Up time 138 hours
Connection speed (Down / Up) 13510 kbps / 1059 kbps
Line attenuation (Down / Up) 22 / 4 db
Noise margin (Down / Up) 5 / 5 db

Thanks for the pointers and other comments on this thread.

In reply to a post by Axolotl:

Note that upload result is higher than max possible.

Did you note my comment about the first test probably being the Beta tester which is notorious for over reporting? And you have just seen it first hand.