I regurly download files from some gigabit servers I run, and I can max out my line single threaded.

Why didnt u do single threaded filesonic tests?

Yes they are but as I understand it there should be no difference and if anything the download server has more control as some people sit there downloading the GB file constantly.

RWIN tuning can also affect single versus multiple too.

The speedtest game is more complex with software firewalls messing with traffic too sometimes. The taking highest sample is the way to produce happier users of a test, is it correct though.

In reply to a post by Chrysalis:

Why didnt u do single threaded filesonic tests?

I didn't think of it at the time and it's becoming obvious that the TBB speedtest won't be changed whatever tests I do. Far simpler if I just don't use it.

Changed how?

We test it on 100Meg and VM connections relatively regularly and produces good results. Seb showed a plot that seemed to show a reasonable distribution of speeds the other day, i.e. lots of VM users on slower packages, but a good number getting high results still

In reply to a post by MrSaffron:

The taking highest sample is the way to produce happier users of a test, is it correct though.

I imagine it wouldn't be beyond the wit of the TBB coders to do both. My own tests show that the difference between TBB and other testers is a lot more than just whether or not the maximum or average are reported though as the Networkx graph clearly shows TBB peaks come nowhere near those from the other speddtests.

Speedtest comparisons. My conclusions from these are that the BroadbandMax test is pretty useless on high speed connections as it's far too short. It clearly reports absolute peak speed too as it is reporting higher than my Networkx chart is showing. For some reason or other my upstream is lower than it should be. It could be someone on the IP camera which wouldn't show on Networx but I think it's more likely congestion as I have seen this before in the daytime. It may even be the logmein access which isn't using a lot of bandwidth according to the chart but may have some other impact. TBB is managing less speed as usual but is actually showing a bit better than the other day at 6:30.

Results from http://www.speed.io
(Copied on 2011-08-11 11:10:23)
Download: 45997 Kbit/s
Upload : 4377 kbit/s
Connects : 2109 conn/min
Ping: 16 ms
http://speed.io/pics/4536/7489/speed.io.png

Speedtest.net
49.71 / 3.77
http://www.speedtest.net/result/1427705920.png


BrodbandMax
Download Speed: 46350 kbps (5793.8 KB/sec ) Upload Speed: 3784 kbps (473 KB/sec )
http://www.speedtest.bbmax.co.uk/results.php?t=13130...

TBB
29.8 / 4.4


Networx bandwith chart while tests were run...
http://www.kwikbreaks.webspace.virginmedia.com/speed...

In reply to a post by MrSaffron:

Changed how?

We test it on 100Meg and VM connections relatively regularly and produces good results. Seb showed a plot that seemed to show a reasonable distribution of speeds the other day, i.e. lots of VM users on slower packages, but a good number getting high results still

I've no idea what would need changing because I don't know what's wrong as, after all, I don't operate a speedtest website. My own comparisons prove to my satisfaction that something obviously isn't right. Fortunately there are alternate testers which I can use if I suspect I have a problem with my connection.

What should happen is a good ISP will run a speedtester just inside their own network so you can test your connection, rather than testing connection, internet network, peering, destination third party server etc

Oh and not use traffic management to ensure good results from some places

There is no capacity based throttling on 50Mbps downstream and VM claim shaping only runs from 5pm to midnight on weekdays and midday to mignight weekends. It is supposed to impact only P2P and NNTP although there are claims that it can also hit "unidentified" traffic including some gaming traffic. That said I seriously doubt that traffic management is the reason the TBB test is consistently slower than others on my connection.

VM suggest using the game test files they host for speedtesting. It's far more convenient to run a quick speedtest though. So long as it can keep up.

Does TBB run a speedtest website or is it an add-on to the rest of what they do?

In reply to a post by MrSaffron:

Yes they are but as I understand it there should be no difference and if anything the download server has more control as some people sit there downloading the GB file constantly.

RWIN tuning can also affect single versus multiple too.

The speedtest game is more complex with software firewalls messing with traffic too sometimes. The taking highest sample is the way to produce happier users of a test, is it correct though.

all good points but generally if things like RWIN are tuned badly then it would affect more then just one speedtest site.

Windows 7 tuning is fine by default, except in some cases of incompatible routers, nics been used alongside the autotuning.

There is VM peering issues, but I think before going further on that tbb need to get it fixed in that at 6am it provides full speed.