Testing FTTC speeds under Linux

I was upgraded to FTTC earlier this week and decided to test my speeds using the TBB speed tester. This showed that a 1x connection was running at about 20mbps with the 6x test showing about 37mbps on a 40mbps connection which I found rather strange. I was trying all sorts of things and then decided to test using Windows 8.1 rather than Linux which is my usual OS, so I connected my wife's laptop to the lan via cable (my linux is a desktop wired in anyway) and then discovered that the x1 and x6 speeds were very similar as one would expect. It seems that under Linux using any Flash based speed tester I can find the speeds for a single thread is reported as much slower than it is using Windows. I have had someone I know check this out on their Linux system (and a different ISP) with similar under reporting of x1 speeds. The problem does not seem to happen when using testers which are NOT flash based.

I am wondering how many others have a similar issue with speed tests under reporting actual speeds when running Linux, for me the problem exists using both Firefox and Chrome with both under reporting the x1 speed using the same tester. I wonder how many others have been initially misled when running such tests with Linux.

Stuart

No significant difference here- x6 was 63.1Mbps, x1 was 62.6Mbps:

Scrub that, I misread it.

http://thinkbroadband.com/speedtest/results.html?id=...

(Firefox, Linux Mint 16, 64-bit)

Edited by billford (Sat 15-Nov-14 13:53:30)

Thanks for try that, what I dont understand is the link takes me to a page which shows x1 at 50mbps and x6 at 62mbps which is I would suggest quite a difference.

Stuart

See my edit- I had a moment's brain fade

If another PC on the same connection gives similar speeds then issue is generally a RWIN scaling issue.

OSX flash has an oddity that single thread connection max's out at 12 Mbps.

The HTTPx6 and tbbx1 are using different socket methods in flash, and it is feasible the Linux flash implementation has an issue just like OSX. The OSX one is easy to spot we warn users about.

Most speed testers do not show the issue as they only do multiple thread tests.

The box to the right of the graphs says:

The HTTP x6 test carries out six downloads at the same time to help maximise the usage on the connection.

so the difference is probably not unexpected. I see a bigger difference using Chrome under OS X (which is Unix-based underneath):

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

but the machine is pretty heavily loaded atm so may not be a fair comparison.

I don't have a Windows machine here, so can't check that.

Edited by billford (Sat 15-Nov-14 14:10:25)

That theory about the problem in Flash could be correct. The issue for Flash and Linux is that it is no longer being supported by Adobe and Linux is stuck at V11.2. It certainly seems that at least under Linux one should avoid an Flash based speed testers especially if communicating with an ISP about download speeds. My ISP TalkTalk has a Flash based speed tester!

Stuart

We have an alternate at http://labs.thinkbroadband.com/speedtest

I used to have a big difference between x1 and x6 speeds; something like 20Mbps/70Mbps (I forget the details now). I was connected PC->gigabit switch->HH3 gigabit port. The problem turned out to be an oddity in compatibility between the switch and the HH3 gigabit port. For some reason, that did not impact most real life browsing/downloading or most tests, but did hit the TTB x1 test.

When I connected the gigabit switch to a 100Mbps port on the HH3, all tests ran at full speed and there were no downsides I can see.

You could be seeing a similar issue, for example if the ethernet ports or drivers on the different machines are causing some mismatch with your router.

Yes I have used that and will in future as I realised it does not use flash.

As to the comments about connectivity I did try my Linux Laptop connected by wire directly to the router (a ZyXel VMG8324-B10A ) which gave the same problem. My desktop is into a gigabit switch and then into the router but since the laptop did not work properly direct to the router I did not try changing the desktop.

Stuart