New DSLReports speedtester

[qasdfdsaq]

Originally Posted by Chrysalis

something up with your system qas, is consistent here.

Also its a test probably aimed at consumer's not university connections? why are you using your work connection?

Yes, clearly it's my system because other speedtests work fine on my system...

There's specifically multiple options for gigabit, fibre, and work/corporate connections though as you can clearly see from the ISP box I am not using a university or work connection. But if it wasn't aimed at all connections (not just consumer) there shouldn't be a 32-stream option or big buttons for "Corporate" "Edu" "Fibre" "Gigabit" and "10Gigabit"

tl;dr: It's a beta. It can't be expected to be consistent and reliable all the time.

[Kushan]

Your connection is probably faster than theirs

[qasdfdsaq]

Originally Posted by Kushan

Your connection is probably faster than theirs

They use 4-6 servers including between 4-8 connections to each server, and that's for the non-10G test. 32 connections across 6 1G+ servers should certainly be enough to max out any connection I have. I dunno what they use for the 10G test, that's members only.

Given 4 streams is enough (on my line) to hit 900M+ I find 32-streams is rather excessive, but should have no problem with "my connection is probably faster than theirs".

[Martyn]

Sky Fibre Pro


http://www.dslreports.com/speedtest/1350371

[Kushan]

Originally Posted by qasdfdsaq

They use 4-6 servers including between 4-8 connections to each server, and that's for the non-10G test. 32 connections across 6 1G+ servers should certainly be enough to max out any connection I have. I dunno what they use for the 10G test, that's members only.

Given 4 streams is enough (on my line) to hit 900M+ I find 32-streams is rather excessive, but should have no problem with "my connection is probably faster than theirs".

Depends how their load balancing those connections and there's a bit more to it than total bandwidth available per stream. I'd like some more technical details of how they are generating their test data.

[qasdfdsaq]

Originally Posted by Kushan

Depends how their load balancing those connections and there's a bit more to it than total bandwidth available per stream. I'd like some more technical details of how they are generating their test data.

The whole thing is HTML5 - just look at the source yourself?

[Kushan]

Me and HTML5/CSS/Javascript don't get along

[qasdfdsaq]

Originally Posted by Kushan

Me and HTML5/CSS/Javascript don't get along

Well then you may have difficulty "getting along" with the technical details of how the test works - i.e. HTML5 websockets.

[Kushan]

Originally Posted by qasdfdsaq

Well then you may have difficulty "getting along" with the technical details of how the test works - i.e. HTML5 websockets.

More interested in the technical aspects of packet size, UDP vs TCP, etc.

[qasdfdsaq]

Originally Posted by Kushan

More interested in the technical aspects of packet size, UDP vs TCP, etc.

That's defined in the HTML5 websocket.

[Kushan]

And the nature and configuration of the NIC's accepting those sockets? There's a lot of variables to consider.

[qasdfdsaq]

Given there's dozens of servers across just as many countries, the NIC configuration is going to be different between all of them.

[Kushan]

That doesn't sound like it's going to produce reliable, consistent results then. NIC configuration can make a hell of a difference.

[qasdfdsaq]

Hence the brute force method, pile in 32 streams from 6 different servers.

For reference, Speedtest.net have no control over NIC parameters either, or even what kind of web server or connection is used, let alone the settings within. All servers there are just set up whatever way the sponsor/provider feels like.

[Ignitionnet]

Comcast's 2Gb service.