Internet providers caught inflating speed test results

We’ve just posted the following review: Internet providers caught inflating speed test results[newsimage]http://www.myce.com/wp-content/images_posts/2016/01/Feature-image-95x75.png[/newsimage]

With the help of TestMy’s speed test, we take a look at the real world throughput of several Internet providers against happens when we attempt to simulate the network activity of Ookla’s Speedtest.net, the preferred speed service most Internet providers recommend using. A few ISPs delivered 3 to 6 times greater throughput despite using the same test server each time, while many other ISPs behave properly, delivering a consistent throughput.

            Read the full article here: [http://www.myce.com/review/internet-providers-caught-inflating-speed-test-results-78458/](http://www.myce.com/review/internet-providers-caught-inflating-speed-test-results-78458/)

            Please note that the reactions from the complete site will be synched below.

Go to testmy.net:8080/multithread. If it does not show the server list, then click the “Enable Multithreading” button again.
Tick the nearest test server to you. Ensure the rest of the check boxes are clear.
Click the ‘Test My Internet’ button below the list to run the multithread test over 8080.

Ok I did this. My isp is TWC here is my result…Download :: 5.2 Mbps 655 kB/s

65% slower than my average Global Multithread US 15.1 Mbps
82% slower than my host average 28.8 Mbps
68% slower than my city average 16.2 Mbps
80% slower than the US average 26.9 Mbps
68% slower than Index Speed 16.5 Mbps
You might have a problem…
Your connection scored only 18% of the Time Warner Cable average (82% slower). This does not always mean that you’re in bad shape. If the Internet package you subscribe to is slower than most people using your ISP you’ll see this message.

My TWC subscription is Turbo 20/2Mbps.

I get this or a similiar result multiple times. My laptop is the only device on the net. Result is via wifi.

What could be happening here?

Interesting discovery.

In my experience, downloading from TestMy.net is generally quite a bit slower than downloading real content from some other sites, so for me TestMy.net is not an accurate test for true download speed.

So while Speedtest.net may artificially inflate download speed for some, TestMy.net may deflate download speed because their servers can’t keep up or are too far away.

[QUOTE=DrageMester;2767288]TestMy.net may deflate download speed because their servers can’t keep up or are too far away.[/QUOTE]

From a quick check over recent results on their German and UK servers, there a are fair number of pretty impressive test result figures such as the following two picked out from their most recent test results list. So it’s quite possible along the route that there is a slow router or link.

On the other hand, I did run into a user complaining on their forum about only getting about 900Mbps on his 1Gbps fibre optic link. I’d be more than happy if I could get 2% of that… :wink:

The following is what TestMy’s webmaster asked me to point out:

TMN represents some of the most common connections, in the most popular hosting areas in the world. And all of the testing locations have a minimum of 1000 Mbps uplink. These datacenters are very popular, well connected and are regularly tested for quality. Speedtest.net has people believing that you need to test off a server that’s close to you. In my opinion testing close does provide good information but only when compared to a test further away… 1000 or more miles away can EASILY provide great results IF the provider is delivering. I do it EVERY day.

To answer valkyrie1968’s question above - I would be interested in seeing what he gets with the standard http test. To me it looks like a poor signal picked up by the cable modem box or it didn’t sync properly, so would be worth checking that its cable connection is tight and rebooting it.

I’m getting from 6 Mbps to ~ 60 Mbps when using TestMy.net depending on server.
When downloading real stuff from a fast site (just tested an Ubuntu download), I’m getting around 90 Mbps on my 80/50 connection.

So for me the TestMy.net results are not indicative of my real internet speed.

:clap:
This was interesting! Criminal corporate activity anyone?

I have to agree with DrageMester in his points though, as the server suggested I should use the UK servers from Norway - Here’s the result from the Download test, both lines tested:
Minimum :: 0.22 Mbps | Middle :: 0.63 Mbps | Maximum :: 0.98 Mbps
Minimum :: 0.11 Mbps | Middle :: 0.51 Mbps | Maximum :: 0.77 Mbps

Weeeell, that’s puny and so I switched to the DE server instead:
Minimum :: 45.84 Mbps | Middle :: 64.87 Mbps | Maximum :: 89.44 Mbps
Minimum :: 36.12 Mbps | Middle :: 54.66 Mbps | Maximum :: 65.03 Mbps

I’d say the results are a tad above average different :bigsmile:
The lines are in reality 100Mbps and 70Mbps.
Then again, there are other network traffic going on at the time of testing and so I think the results tells what earlier tests have shown, I get what I pay for.

Nevertheless, this was very interesting and a piece of information I will keep in mind when testing myself.

Thank you for a great in depth study to all involved :flower:

Your article certainly comes at the right time. With many ISP’s starting to offer gigalan speed, 300mb and so on. I have found it amazing that I barely even see 1mb speed transferrring data from one TW connection to another. Connections that are supposed to be 50/5 . I have recently been seeing 100/10 with ooklah and even dslreports during the day, but at night my d/l speeds plummit to 27/7 . Of course when you call TWC they tell you everything looks fine, yet they really do not have any tools to check with and they won’t send someone out at night to check it. Now with your testing,that you have done, it sounds like I would have real benefit if you were running a plex server and set the default port to 8080. Is this correct. Also with you testing, did you try DLSReports speed test or have an opinion about it? I have been using it for a while. It also seems to be able to test for multi streaming and bufferbloat. But if it is doing the same thing as ooklah speed test then it is really useless. Any additional info would be great and thanks for putting this report together.

None of this is really “news”, although its interesting to know some of the nuts and bolts of how the cheating is done. But most folks have been aware for years that the popular flash-based speed tests were unreliable.

That said, its pretty rare to find a web server in practical use that can deliver anything close to 50Mb speeds on sustained transfers. So in practical terms it really doesn’t matter if your pipe delivers 50Mb or 100Mb or 1Gb.

Those international tests are fairly inaccurate all depending on the destination server (responder)

Melbourne:

Singapore:

Pretty close results even the distance is multitude - watch the routing:

It depends on whether the Internet provider has prioritised incoming traffic only or in both directions on port 8080. For example, the Ziggo ISP only delivered slightly better uplink performance over port 8080, which means a home server operating on port 8080 would probably only deliver a small improvement.

On the other hand, one way to exploit the extra bandwidth would be to make a VPN connection over port 8080 to a nearby server (e.g. privacy VPN that supports connections over 8080). Last year when the Three mobile network here had something like a 10x speed variation between HTTP and HTTPS, I use to connect to a VPN to get much faster Internet access on my mobile. Now their network has got so congested that the high latency degrades the VPN performance.

[QUOTE=dun4cheap;2767313]Also with you testing, did you try DLSReports speed test or have an opinion about it? I have been using it for a while. It also seems to be able to test for multi streaming and bufferbloat. But if it is doing the same thing as ooklah speed test then it is really useless. Any additional info would be great and thanks for putting this report together.[/QUOTE]
I’ll do some further checking, but from a first glance their test is very customisable which is nice, i.e. you can select the number of threads to use, whether to run over http or https, limit data usage (for metered ISPs), select a server location and various other settings.

By default, DSLReports runs its test multithreaded (multiple connections to test server) and according to TCPView, it does this over the http on the downlink going by TCPView, which is also nice. So without changing any settings, it aims to measure your ISP’s maximum throughput and using port 80 for the download (not 8080). This is similar to downloading a large Linux ISO using a download accelerator that downloads multiple parts of the file simultaneously. On the other hand, it runs its uplink over port 8888 for some reason, but at least it’s different to what speedtest.net uses.

In the settings page, there is an option to specify how many streams (their word for threads) to run over, so you can select ‘1’ for the download and upload to run a test similar to how TestMy runs by default, i.e. like the bandwidth you get from streaming video or downloading a single large file in your web browser.

TestMy can run a Multithread test also as mentioned on page #6, however, it doesn’t provide the ability to specify how many threads to use. By default, it runs its test over a single connection to the test server.

Basically the main point of this article is to show that some ISPs are giving priority to traffic on port 8080 to inflate the speedtest.net figures. However, as TestMy has support for running its tests over both port 80 (http) and port 8080, I used it for the demonstrations in this article.

I also prefer TestMy over Ookla’s Speedtest due to its ability to test with a single connection, particularly with wireless ISPs that are quite common in Ireland. Some of these ISPs deliver very little bandwidth over a single connection (e.g. 4Mbps) even to a nearby server, probably to prevent YouTube and streaming services going into 1080p mode. However, if one tests the same ISP with Speedtest.net, it will give the impression that there’s plenty of bandwidth for 1080p video since Speedtest tests over multiple simultaneous connections to its test server.

[QUOTE=CDan;2767314]None of this is really “news”, although its interesting to know some of the nuts and bolts of how the cheating is done. But most folks have been aware for years that the popular flash-based speed tests were unreliable.

That said, its pretty rare to find a web server in practical use that can deliver anything close to 50Mb speeds on sustained transfers. So in practical terms it really doesn’t matter if your pipe delivers 50Mb or 100Mb or 1Gb.[/QUOTE]

Yeah, this has always been assumed but I think it’s still genuine ‘news’ mainly because we’ve now got proper evidence, thanks to Seán’s expert detective work. :clap:

None of it surprises me either but I still think it’s pretty cool that we were able to catch these guys out. :cool:

[B]Wombler[/B]

Thanks for the info and article. In the past six months I have tested my internet speeds across many ISP’s. Cox, ATT Uverse, TWC, and VZ 4g on a more regular basis. I have access to all of those connections all the time. Your report in my opinion has come at the perfect timing. Especially with all of these main ISP players talking 300 and even giga speeds coming. What amazes me is I simply cannot get the 50/5 I am paying for now. I have even spent a couple of hours on the phone with TWC asking them how they test their lines themselves and of course they push you to speedtest.net and you run a test and it looks good however I cannot get more the 1mb from one TWC to another TWC in the same area. Will DSLreports for me it actually keeps a record for me so I can give this info to them. (There appears to be a load balancing issue in my neighborhood.) But at this point there is no real accountability for the ISP. But this article points this out and explains how they go about it. Hopefully there will be a 2nd part article to come in the future.

My Tests are flawed, I pay for 60 down and 30 up, but if its at night I get less, in the morning I get about the right amount.

[QUOTE=alan1476;2767456]http://www.speedtest.net/result/5036842303.png

My Tests are flawed, I pay for 60 down and 30 up, but if its at night I get less, in the morning I get about the right amount.[/QUOTE]Can you run a test from this link, pick the server closest to you. Click the “Test My Internet+” button towards the top right and click ‘Combined’. Then post the result.
http://testmy.net/mirror

Then run the same test over port 8080, with the following link, and post the result.

http://testmy.net:8080/mirror

[QUOTE=Dee;2767464]Can you run a test from this link, pick the server closest to you. Click the “Test My Internet+” button towards the top right and click ‘Combined’. Then post the result.
http://testmy.net/mirror

Then run the same test over port 8080, with the following link, and post the result.

http://testmy.net:8080/mirror[/QUOTE]Hi Wendy:
Boy this is really screwy
http://testmy.net/pR5cvOn80.OBCwuXpNI.png - combined

URL: http://testmy.net/db/w58JC4Zyu.gqLVjIDEb 8080

I noticed you can change the size of the data on this test, what size would you like? I just went with the default amount.

The size of the data block would be best set to around 50MB, but it’s more important that you run the ‘combined’ test on both normal, and port 8080.

[QUOTE=Dee;2767477]The size of the data block would be best set to around 50MB, but it’s more important that you run the ‘combined’ test on both normal, and port 8080.[/QUOTE]
Okay here is the combined test on 8080

Why are my results so terrible?

Alan’s results.

Normal on the left, port 8080 on the right.