80/20 Infinity up and down

Hello,
Since upgrading to 80/20 from 40/10 (which gave me 12 months very good service - 30ish/8ish) - my performance varies from 8/1.6 to 64/16.
I've been on to BT continuously since 24 April, with varying results - which seem to be cured for an hour or so, then drop - typically to 20ish/12ish.
So many variations on speedtest results: BT variously say their "BT Beta" is accurate - other BT staff say use the normal BT speedtester. Which is likely to be more accurate - if at all?
I've uploaded the TBB result from a few minutes ago, when the BT tester showed only 28.15. Who do I believe? The general Infinity performance is fine ie everything works pretty much as it should, but the figures leave me confused. Any comments? Cheers, Les.
(URL below)


http://schirmfoto.de/4fb90cbd33c7

I've been through exactly the same as you (getting rock solid 30+/8+ from the 40/20 product, to wildly variable results from the 80/20 product - to the point I'd come to the conclusion I was being throttled).

After some time (and exhaustive tests) it appears Windows 7 was largely to blame for my speed issues (although quite how my speeds returned to normal after peak times - i.e. outside of evenings & weekends - is still an unanswered question).

To shortcut you (close to the end of my "adjustments") you may as well carry out two speed tests.
The first is a single thread Java based speed test (the regular TBB test is also Java based).
The second is a multi thread (four; I think) Java based speed test, from the same provider.
Single thread test
Multi thread test

The regular BT speedtester is also Java based (their beta test is Flash based, along with the TBB Flash test and places like speedtest.net, speed.io & speedtest.bbmax.co.uk ).

See if you get close to full line speed with the multi-thread test, and much lower with the single thread test - we'll take it from there.

In reply to a post by lelboy:

Hello,
Since upgrading to 80/20 from 40/10 (which gave me 12 months very good service - 30ish/8ish) - my performance varies from 8/1.6 to 64/16.
I've been on to BT continuously since 24 April, with varying results - which seem to be cured for an hour or so, then drop - typically to 20ish/12ish.
So many variations on speedtest results: BT variously say their "BT Beta" is accurate - other BT staff say use the normal BT speedtester. Which is likely to be more accurate - if at all?
I've uploaded the TBB result from a few minutes ago, when the BT tester showed only 28.15. Who do I believe? The general Infinity performance is fine ie everything works pretty much as it should, but the figures leave me confused. Any comments? Cheers, Les.
(URL below)


http://schirmfoto.de/4fb90cbd33c7

Don't bother with speedtesters.

Start a multithreaded download (use a download manager like Free Download Manager) from a decent wesbite with bags of bandwidth (I.e. download a service pack from Microsoft) and see how fast it goes.

I get all sorts of variable results on all the speedtesters I use but actual real world downloads show that my connection delivers the full IP-Profile throughput all the time

In reply to a post by adebov:

After some time (and exhaustive tests) it appears Windows 7 was largely to blame for my speed issues (although quite how my speeds returned to normal after peak times - i.e. outside of evenings & weekends - is still an unanswered question).

That pretty much shows it can't be a defect in WIn7. Something I think would have been found by now in corporates given how many are moving fast from XP to 7.

That said we have no knowledge on congestion in the OR FTTC part, the wholesale WBC part, or the ISP part of the complex networks we use for broadband internet.

In reply to a post by jchamier:

That pretty much shows it can't be a defect in WIn7. Something I think would have been found by now in corporates given how many are moving fast from XP to 7.

My thoughts exactly.
Having said that, I was getting very low results right up to the point I tweaked.
Almost 30Mbps difference between a single thread speed test (and the multi thread test from the same provider) to a few Mbps difference immediately after tweaking (and with repeatability - essential for accurate diagnosis).

I was sticking to my guns (about BT throttling) but kept getting shot down in flames ("not possible", "BT don't throttle", "no congestion", "BT can't do local throttling", "BT can't throttle individual threads", "mine's OK so it must be you", blah, blah, the usual "helpful" responses).
Oddly enough, all the way through I was getting maximum speed from Usenet and other multi-thread downloads (which proved congestion wasn't an issue).

The poor peak time speed, and it only affecting single thread downloads, was reinforcing my view it was BT throttling individual threads.
Then, suddenly, all is OK following my tweaking the other day (now single thread downloads are almost as fast as multi thread - which has never happened before, during peak time, with my 80/20 product).

In reply to a post by adebov:

Then, suddenly, all is OK following my tweaking the other day (now single thread downloads are almost as fast as multi thread - which has never happened before, during peak time, with my 80/20 product).

I wonder if its anything similar to the problem BT had years ago with packet sizes being too large. Where people found setting MTU to 1458 was beneficial. Its around 2002/2004 time frame - but there was a misconfiguration on some core routers in the old 20CN network.

One of my tweaks was to fix MTU at 1492 (the recommended PPPoE setting) - Windows 7 kept overriding my setting of it (and kept going back to 1500), until I turned off the Win 7 auto-adjustment.

Quite how an MTU of 1500 (when it should be a little smaller), on its own, made single thread downloads slow and didn't affect multi threads I'm not qualified to judge.

Of course all this is irrelevant if BT have their packet size set incorrectly for PPPoE connections

My guess is it's something to do with peak time congestion adversely affecting the Windows 7 auto-tweaks (particularly the one when W7 checks the speed of the first few packets and tweaks to suit).
Maybe just turning off auto-tweaks stops this?

All I know is my XP laptop has always managed faster single thread downloads than my Win 7 desktop (and both PCs are about the same vintage and spec - other than the Win 7 desktop is full 64-bit and has 6GB of RAM, compared to a 64-bit laptop with 3GB RAM running a 32-bit OS).