Reliable Internet Speed Test

[smitbret]

I usually use DSLReports.com's tools to do my Internet Speed Tests. It has consistently shown a DL speed of 12Mb/s and an upload of 1.5Mb/s for my connection. My advertised speed is 50/2.

Out of curiosity, I went to my ISP's website to look at my account info and they have a link to Speedtest.net. I ran the speed test and, not surprisingly, I got the advertised speeds or slightly higher for both up and down. Now I was suspicious, so I bypassed the website link and went straight to www.speedtest.net via the address bar and got the same results there.

So, my question is who do I believe? Are the tools at DLSreports outdated and inaccurate or is Speedtest not real? Is there a good, reputable way to test my speeds that I am not aware of?

 

[Gryz]

If you really want to know how fast your link is, check the "trained speed" on your DSL-router. Your ISP configures a maximum speed on the DSLAM (the equipment on the other side of your copperwires). When your router at home boots, it connects to the DSLAM. They will do some tests to see how good the copper is (distance, quality of the copper, interference, etc). This is called "training". The link will then run at the maximum speed that the DSLAM thinks it can run at. (Of course it won't go faster than the max configured speed). This is not necessarily the speed that is configured on the DSLAM.

Find the IP address of the router.
One way is to look at the routing table of your PC, and find the default gateway.
Open a cmd.exe (aka dos-box). Type "netstat -r".
Find the IPv4 Route Table. Find the entry 0.0.0.0. See what the Gateway's IP address is.
That is your router's IP address.
Now you can type that IP address in your browser.
You should now have access to the routers web-interface.
Maybe you need a password. Check the manual, or ask your ISP what the default password for your router is.

Find info about your DSL link.
It should have upstream and downstream info.
Speeds (in kbit/sec or mbit/sec).
Min and Max DSLAM throughput (as configured by your ISP).
Attainable throughput (max speed the DSLAM thinks is possible over your copper).
And Current Throughput. This is the speed at which your link is actually running.
This is an example of DSL info on a FritzBox. (I happen to have one of those).
http://screenshots.portforward.com/AVM/FRITZ_Box_Fon_WLAN_7570/Internet_DSL_Information_DSL.htm

I always look at the Current Throughput on my router. That is the actually speed your line is running at.

Your measured data transfer rates are lower. That is too be expected. DSL has some overhead. As a rule of thumb, the data transfer (in Bytes/sec) should be about your link's speed (in Bits/sec) divided by 10. Not by 8 (as you expect with bytes->bits). That's how much overhead DSL gives you. Example: my own DSL link runs at 7.5 Mbps. And the download rates I see (and expect) are ~ 750 KiloBytes/sec.

If the speed at which your DSL link is running, is really 50 Mbps, then you should see data rates of 5 MBytes/sec. You see 12 Mbit/sec. I expect your DSL to run at 15 Mbps or so.

Note, normal ADSL run at 20 Mbps maximum. (ADSL, ADSL2, ADSL2+, etc).
You need VDSL to go higher.
Are you sure you have a VDSL subscription ?
Are you sure you have a VDSL capable router ?

 

[smitbret]

It's actually a cable internet connection. DocSis 3.0 setup.

My issue isn't so much the speed, it's perfectly adequate for what I need. The disparity between tests just piqued my curiosity. Then, I started thinking, why am I paying for 50Mb if I could just pay for a 12Mb or 20Mb and get the same results.

I'll check my connection tonight. Thanks for the info.

 

[Mushkins]

Because if you were paying for 12Mb or 20Mb, you'd be getting 2 or 6.

This is the nature of how cable works, you're on a shared network segment that shares x amount of available bandwidth. You are in no way, shape, or form guaranteed that 50Mb at all times. That is an advertised estimate. As other people on your node use bandwidth, yours can be impacted. Run a speed test in the morning and it could be totally different than during dinner time, both of which are accurate. Speedtest.net is pretty much the go-to web based speed tester.

Not only that, most of the higher tiers (Xfinity Blast Plus, Optimum Boost, etc) aren't just allocating more bandwidth to you, but there are enterprise level QoS and traffic shaping going on to make sure you get that extra bandwidth over the guy next to you who isnt paying for it.

Right now the only truly high-speed broadband internet that gives you exactly the advertised speeds (minus overhead) is FiOS, because of how fiber optic works as a technology the bandwidth allocated is strictly dedicated to you.

 

[Red Squirrel]

Speaking of speedtest, anyone know why it would give higher results for my phone, than my desktop? Well download is the same, but upload on the phone is somehow faster, which is odd because not only is it faster, but it's faster than my actual obtainable. I can go up to 1mbps and I apparently get 1.4. Both going through the same firewall/connection, which is the bottleneck.

 

[Gryz]

So on speedtest.net you got your 50 Mbps ?
And on DSLReports you got 12 Mbps ?

The Internet does "best effort delivery".
There are no connections on the Internet. No end-to-end bandwidth reservations. Nothing of the sort.
There is something called Quality of Service. But that just prioritizes some types of traffic over other types of traffic. And AFAIK there is hardly any QoS between ISPs.

Speedtest might be close to your ISP. And there might be more than enough bandwidth along the path to their servers. While the path from your ISP to DSLReports might be longer, lower bandwidth, more congested, etc.

If you'd use a website to measure your speed, and that website happens to be in Australia, the results will probably look a lot worse than if you were testing on a website that is inside your own city. (Or to be more precises: a website in your own city that is also on your ISP). On the Internet, things are a bit more complex. But you can understand that not every destination is equally far away.

You can use traceroute (tracert on Windows) to see more info about the path between your own PC and some remote host on the Internet. Interpreting the result is non-trivial, so I won't even try.

 

[kornphlake]

Several years ago certain speed tests were only capable of testing certain speeds because of the way the speed test was measured. I can't recall which technology needed what, but I know some tests were java based and some were flash based, higher speed connections got more accurate results with one test and slower connections got more accurate results with the other test. DSLreports.com used to have a pretty good explanation of which test you should used based on your advertised speed on the test page, I'm not sure if it's still there or not.

 

[ScottMac]

So on speedtest.net you got your 50 Mbps ?
And on DSLReports you got 12 Mbps ?

The Internet does "best effort delivery".
There are no connections on the Internet. No end-to-end bandwidth reservations. Nothing of the sort.
There is something called Quality of Service. But that just prioritizes some types of traffic over other types of traffic. And AFAIK there is hardly any QoS between ISPs.

Speedtest might be close to your ISP. And there might be more than enough bandwidth along the path to their servers. While the path from your ISP to DSLReports might be longer, lower bandwidth, more congested, etc.

If you'd use a website to measure your speed, and that website happens to be in Australia, the results will probably look a lot worse than if you were testing on a website that is inside your own city. (Or to be more precises: a website in your own city that is also on your ISP). On the Internet, things are a bit more complex. But you can understand that not every destination is equally far away.

You can use traceroute (tracert on Windows) to see more info about the path between your own PC and some remote host on the Internet. Interpreting the result is non-trivial, so I won't even try.

"And AFAIK there is hardly any QoS between ISPs."

There is NO QOS on the internet. None. Zip. Nada. Bupkis (etc.).

Within a specific provider's distribution system (like A&T U-verse), some QOS and non-QOS services may share the distant segment (i.e., VoIP & video have QOS, Internet does not, all on the same wire from the house to the DSLAM). This is done with tunnels (QOS defined and non-QOS defined) and split out upstream, like at the CO. The QOS services continue through their own private internal network, the Internet is routed to the Internet.

 

[smitbret]

So on speedtest.net you got your 50 Mbps ?
And on DSLReports you got 12 Mbps ?

The Internet does "best effort delivery".
There are no connections on the Internet. No end-to-end bandwidth reservations. Nothing of the sort.
There is something called Quality of Service. But that just prioritizes some types of traffic over other types of traffic. And AFAIK there is hardly any QoS between ISPs.

Speedtest might be close to your ISP. And there might be more than enough bandwidth along the path to their servers. While the path from your ISP to DSLReports might be longer, lower bandwidth, more congested, etc.

If you'd use a website to measure your speed, and that website happens to be in Australia, the results will probably look a lot worse than if you were testing on a website that is inside your own city. (Or to be more precises: a website in your own city that is also on your ISP). On the Internet, things are a bit more complex. But you can understand that not every destination is equally far away.

You can use traceroute (tracert on Windows) to see more info about the path between your own PC and some remote host on the Internet. Interpreting the result is non-trivial, so I won't even try.

Thank you, someone that understood my question. I wasn't questioning real world availability, just the difference between speed tests.

I used the speed test at DSLReports with the Palo Alto location

When I did SpeedTest.net, it first tested from Salt Lake City, UT which is the nearest geographic location and got the 50/2 results. Then I switched to a Palo Alto server to mirror what was happening (at least distance-wise) and got the same 50/2. I actually got better ping response from Palo Alto than I did from SLC.

 

[murphyc]

Use an NDT server. Google 'ndt server'.

 

[Arush]

Best way to speedtest is using Torrent client (If your ISP does not block port forwarding) , because when you download a torrent, for example, say ubuntu ISO image file using BitTorrent. It will download from diverse seeders/peers which are on different networks, country, and location. This will give you REAL case scenario of what maximum speed you can achieve or capable of.

Many times what many ISP do is they claim to give you 100mbps but you only get those speeds only when accessing content within your country/region and when you download or stream anything outside of your country your speed drastically falls to 10-20mbps , this is because your ISP gives you 100mbps within your country and 20mbps international bandwidth.

So downloading torrent with lots of seeders/peers will give you actual speed you can expect.