As has been pointed out already, speedtest.net use multiple threads - https://support.ookla.com/hc/en-us/articles/23457582...

In reply to a post by arronlowley:

im not moving my pc downstairs again just to do another ethernet test when i personally know it wont make the slightest bit of difference.

Which says you're not in the same room, so I'm quite impressed with your wireless throughput.

I myself have the SH3 in modem mode, using an Asus RT-AC87u, so 802.11ac 4x4 MIMO as the router.

I have an old Netgear 802.11ac 3x3 MIMO router I can put in as a bridge tonight to test. Will be interesting to see if I can get such high speeds (even in the same room!) over wireless.

On the main page, you can see the 'hosts' of the speed test for each area of the country. For Norwich, it's:

Intouch Systems
Freeclix Ltd
Voicehost Ltd

Edit: Actually looks like Freeclix is in Norwich

Edited by deleted (Tue 31-Jan-17 14:14:31)

In reply to a post by MrSaffron:

And doing an in the head Ookla algorithm suggests using same method we'd report around 260 Mbps

Also interesting its picked a Norwich server (given your location sort of logical), but we don't know whose network it is on either. But then how much of your data download is from Norwich?

Some of the Virgin Media problems in the past have been not just local contention being visible, but some of the routes out of their network being congested too.

That test was "Voicehost Ltd".

I do get similar speeds to my own server hosted by Redstation, I'm not quite sure where that is, London area I guess.

what makes the tbb speedtest so different? and why is it miles apart in terms of speed from the majority of other speedtests, basically what youre trying to tell me is, my connection has never been 300/20 because tbb says so.

Edited by arronlowley (Tue 31-Jan-17 14:58:40)

And

Download
"7.The fastest 10% and slowest 30% of the slices are then discarded (see * below for more detail)"
Upload
"5.Chunks are sorted by speed, and the fastest half is averaged to eliminate anomalies and determine the result"

So when we take a median on download for example, we are almost guaranteed to be slower than Ookla

It is not 'miles apart in terms of speed from the majority of other speedtests' but it is doing things in a different way that is designed to be more reflective of the user experience (outside apps such as Steam and P2P)

The test itself is perfectly capable of showing your speeds e.g. http://tbb.st/148342930123956785509 and can go a lot faster http://tbb.st/1483303592226028855

What you'll notice on those two tests is how stable the samples are for the httpx6 test, and on a good connection that is what you are looking for, which indicates Ethernet (or good wifi) and no bad contention issues. The first one does have a slow ramp on the single thread, so might be some light contention in that area.

The speed does ramp to 300+ quite quick on Speedtest.net - https://www.youtube.com/watch?v=W_qjY11sTkE

As you can see it's over 300 for basically the entire time. I'm doing this over RDC as I'm at work so there is some other traffic happening.

In reply to a post by arronlowley:

what makes the tbb speedtest so different? and why is it miles apart in terms of speed from the majority of other speedtests, basically what youre trying to tell me is, my connection has never been 300/20 because tbb says so.

Your connection - and mine - would probably very rarely if ever be able to hit 300mbit with a single connection, e.g a VPN or a single source download. There is a level of contention. So from that perspective, it's never been 300/20.

But, in my case I can hit 80-90 mbit on such a download, which is more than is possible from any other provider as BT estimate ~40mbit on FTTC for me, and many of those other ISPs will slow down at peak times themselves.

AAISP are about the one ISP that will truly give you your line speed on a single connection day or night... but for that I'd pay as much as I pay for 300mbit from VM and only get 80mbit (actually likely to sync at more like 60) with a 1TB cap.

And that leads us back to, you're not happy going down to 160mbit at peak times, but what other ISP can you switch to that will give you more than that, at any time?

VM support suck. Nearly as badly as Plusnet do. Again, you get what you pay for, I think the number of ISPs with "good support" these days could probably be counted on one hand, even if I'd lost a finger or two in an accident.

And all of them are too expensive for me

As it usually does on our httpx6 test too

What is interesting is that the needle stayed at 316 Mbps at the end of the download test, but the little graph showed things dropping away a lot.