[SOLVED] How reliable are on-line speed tests?

[daviddunford]

I've just replaced my ancient Cat 5e cable from a Virgin 3 Hub with a Cat 6 so I decided to test my speeds. Previously Ookla speedtest gave around 140 mb/s (it's a nominal 350mb/s connection). Now Ookla on line gives 250, the Ookla app 150 but various others much slower rates, including one (on an ISP switching site) that recorded only 12 mb/s. All the tests were done within a minute or two. I know results will vary, and much depends on the server they use, but it's hard to avoid the thought that the ridiculously slow results were faked in order to persuade me to change my ISP. Any thoughts?

 

[rgd1101]

it depend on where the server located.

 

[daviddunford]

Thanks. I appreciate the importance of where the server is located; I tried Ookla again and with a server only 4 miles away it gave me 290 mb/s. But what sort of location or distance could give me only 12?! I'm still suspicious.

 

[bill001g]

They are actually very reliable but you have to be very sure you understand what is being measured.

You will see large difference even within Ookla if you change server locations. Even choosing different servers in the same city can give you different results. Partially this is the serer itself. Some servers are very large machines on very fast internet connections that are say 10gbit. Others are much smaller machines and connections.

Next you can also see huge difference between machines that are in the same city. This is mostly related to how your ISP is connected to the ISP the server is using. Lets say you want to connect to a pc across the street at your nieghbors house. You use different ISP but they are very cheap ISP and do not pay to have interconnections in a lot of locations. So lets say they are really cheap and only connect in say Australia. That means to go across the street your traffic must actually go all the way to Australia and back.

Speedtest is mostly affected by the location of the server. Even if you choose your own ISP server if they do not have a server located very close it will be impacted.

In general your ethernet cable should not have much impact on this unless it is actually defective.

Your traffic actually will run at 1gbit between your device and the router all the time. So the data would run at say 200mbps between the location and the buffer in your router. The router would then send it to your PC at 1gbit from the buffer. It would not be real useful if the PC reported this as 1gbit since it is always 1gbit. It gives you a average rate which includes the time it is waiting running 0mbps for the data to be placed into the routers buffer. This is somewhat of a over simplification but speedtest are seldom affected by ethernet cables.

 

[rgd1101]

distance could give me only 12?! I'm still suspicious.

don't use them

 

[daviddunford]

Thanks particularly to bill001g for a very helpful and detailed answer. In real life terms, I already had very good speeds (but the connector on my cable was broken so tended to fall out). Now they are slightly faster, even allowing for variation between servers. So no complaints!