It looks like we might have spotted a common link between problematic connections on the VM forums. So far everyone with the problem who has checked is on a CMTS with a 'RiverDelta' MAC address which I assume means Motorola.

We don't have a ton of data yet but it does seem to be the case so far. Ideally we could use some more people with/without the problem to post what CMTS they're on to see if the trend continues.

If you're interested, you can find the CMTS MAC address in a Superhub 3 by looking in network logs (be careful not to confuse it with the modem's MAC address though). If you're on a SH2 I think you need to be in modem mode and look at ARP tables in the connected router - that worked for me anyway.

And use a MAC lookup site to enter the first 6 digits and see what company it's linked to: http://www.macvendorlookup.com/

I'm assuming that TBBs monthly summary of download speeds shows the results for the multi-thread downloads. It would be interesting to know what, if any, slowdown is seen on multi vs single thread speeds for different ISPs as that would give a clue about congestion impacts.I realise that the tables are already a bit complex, but it would be intriguing to know (and it also shows the problem of characterising BB performance with a single metric).

Single thread performance can be an issue as many protocols are inherently single threaded. It also reminds me that the TCP congestion control protocols were designed on a "play nice" good neighbour basis. It was designed to treat all TCP connections on an equal basis. Once multi-streaming protocols arrived that "good neighbour" assumption went out the window (subtle TCP flow-control pun unintended).

Have stopped fiddling with single threaded now, so once we have a clear month or two of data to look at will talk about the overall results more.

Good. It might prove interesting. Services like 4K UHD streaming seem to need a sustainable 20mbps or so. Not that I personally see much point in UHD for consumption in a typical UK living room (rather than authoring), but some will consider it important.

I definitely experience the single threaded issue and I'm on a Cisco CMTS.

It seems like Virgin are continuing to just brush it under the carpet as long as they can point to speedtest.net results and act like they don't know of any problems. If speedtest started showing single threaded results I bet they'd do something about it!

How bad is it for you? My single threaded speed is around the 10Mb mark or lower at peak times now.

I find the Netflix speed test gives very similar results the TBB single thread test...

Netflix Speed Test

Mine drops to about 20-30Mb/s at peak.

I also note that the single thread TBB test starts higher and then drops off.

Hmm.

I think they might have fixed something in my area, because I just got 180/210 on the x1 at 19:40;

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

35Mbps here using the Netflix speedtest.

I'm on a Riverdelta CMTS and see the same single thread issues. It's pretty pathetic.