kfc wroteHow valid are the results given by speedtest.net?
I am on a 1 Mbps plan and more than 75% of times, I am getting speeds below 0.5 Mbps on speedtest.net while my actual download speed exceeds 100 kb/s 75% of times!
I tried to do a speedtest while downloading a 30 MB File at a speed ranging between 108 and 115 kb/s. I got 0.24 Mbps!
If you are using download accelerator, you will get the max speed. Speedtest AFAIK is only one connection to the server ( no threading ). If the internet connection is shared and there is congestion, a single tcp connection will not be able to take all the allowed bandwidth. If we want to really simplify things, you think of it this way :
If you are sharing a 10mbps and the router is having 20 requests from 20 different users to transfer at full capacity, you will get 0.5mbps download rate. When you increase the number of connections downloading from the same source to 10, you will get your 1mbps. IE limits the connections to the same server to two.
What all this mean is that if you try to load a youtube video, you will not be able to stream at 1mbps, you will get the maximum available bandwidth per tcp connection that speedtest is reporting.
( IF anyone doesn't agree with me, feel free to discuss since I'm not that much into bandwidth sharing algorithms
)
One interesting observation is that some modems perfom better on speedtest.net;
When I use the Speedtouch and run speedtest.net I get most of the time 3.4Mbps download rate but sometimes 1.8 - 2.3mbps.
Since changing the setup to Thomson used in bridge mode and router handling the pppoe, I've
never had a speedtest download rate below 3.5Mbps ( regardless of peak hours ).
I run some diagnostic tests and noticed the CPU load of the modem increasing to 100% when running a test depending on the number of connections it's handling ( NAT ).
In bridge mode, the CPU load is always around 14-20%.
When I changed the setup to Speedtouch in Bridge mode and PPPoE on the router side
