Internet speed lower than Speedtest.net results during actual download

1

1

I am subscribed to VDSL2 with 10 Mbps Download and 1 Mbps Upload. My sync speeds on Modem's page is 11264 kbps down, 1119 kbps. On speedtest.net, I always get same result all the time at any given day; 10.54 Mbps download, 0.92Mbps Upload which is about right. However, when I download from websites like Mega.nz, Google Drive, Steam, Mediafire etc. my actual speed is only about 1.1 MBps (In Bytes) which translates to about 9 Mbps (8.8 Mbps) which does not seem quite right. On the other hand, on Steam application, my maximum speed is only 9.6 Mbps as shown here:

Download speed in Mbps

SNR Margin and Line attenuation seems good though:

Connection Status       
Help
DSL synchronization status  Up  
DSL up time 09:39:18  

Line Status     
Help
Line standard   VDSL2
Downstream line rate (kbit/s)   11264
Upstream line rate (kbit/s) 1119
Downstream SNR (dB)     13.4
Upstream SNR (dB)   30.1
Downstream line attenuation (dB)    22.7
Upstream line attenuation (dB)  10.1
Downstream output power (dBmV)  11.2
Upstream output power (dBmV)    4.9
Downstream CRC  4
Upstream CRC    0
Downstream FEC  12428
Upstream FEC    36

I use Ethernet cable connected directly to my PC, and also I am the only person who uses the internet. Is this considered normal in ideal conditions? Thanks.

John Hark

Posted 2018-03-05T17:36:37.200

Reputation: 35

1A download speed of 1.1MB/s translates to a line speed of 9.2Mbps. Download speeds use binary units, line speeds use decimal units. Accounting for a typical coding efficiency of 90%, it's more like 10.1Mbps. – David Schwartz – 2018-03-05T19:19:14.867

Answers

4

There are many factors that impact the speed of a specific download.

Among these are distance to the server, number of hops between you and the server, the capabilities of the server, the load on the server, the number of devices on your connection, other bandwidth needs of your own computer, and a myriad of other things.

Speedtests show you a baseline optimized speed and are generally built to indicate the maximum possible connection speed.

Speedtests are NOT a guarantee or even an indicator of the speeds you should expect to receive in real downloads.

music2myear

Posted 2018-03-05T17:36:37.200

Reputation: 34 957

Very informative. In that case, is it technically possible to deliver max speeds on an actual download? I would love to see something like 1.3 MB at least close to my speedtest results which can be useful in most cases. – John Hark – 2018-03-05T18:16:00.603

Yes, it is possible. The server will have to be both capable and have the available resources to send this rate, and there will have to be nothing that would impede this rate between your computer and the server, and your own computer would have to have nothing else using the connection. – music2myear – 2018-03-05T18:20:38.443

1

Further to the answer above there is also a bandwidth overhead associated with A/VDSL somewhere in the ballpark of 10% as well as another figure for the generic IP packet overhead.

A study Encapsulation Overhead(s) in ADSL Access Networks puts the overhead at anywhere from 10 to 20% of your line speed. In the conclusion of that study they state:

Generally speaking the “line efficiency” of a protocol stack/encapsulation is roughly 80% given the Internet packet distribution.

Which means that your "effective" line speed is actually somewhere below (11264kbps - 1126 = 10138kbps) which is approximately 10Mbps.

The fact that you are getting pretty close to 10Mbps suggests that you are much better off at closer to 10% overhead rather than the 20% encapsulation overhead.

Mokubai

Posted 2018-03-05T17:36:37.200

Reputation: 64 434

This is good info. However, OP appears to be asking about the disparity between the tested speed and the observed speed for a real-world download, both of which are "post-overhead" measurements. – music2myear – 2018-03-05T22:54:20.477

Aye. I was aware of that and your answer was perfect for describing that side of things, I just thought it worth mentioning why line speed and speedtest measured speed could be different. – Mokubai – 2018-03-05T23:56:52.337