Comcast Speed Test Vs Speedtest Net Speed Test

CT is virtually always slower than ST by a factor of 2 to 5, sometimes even as high as 10 in some locations; therefore CT should not be used for estimation of internet speeds but only for troubleshooting purposes when there is a problem.

We will look at the relative speeds of CT and ST in multiple locations around the US. Not just my speed, but many other people's speeds who have reported to various internet forums, particularly DSLReports.com . That should give us some idea about how much variation (spread) there is from the norm.

Why Speeds Are Slower On CT Than ST?

(1) CT uses TCP congestion for control. This is a timer which starts at around 5 seconds and increases in 10 second increments for each packet loss after that. If there is no packet loss, the timer typically increases by 1 every 10 seconds. The default is 30 seconds when there are packet losses. When the congestion window reaches a preset limit (3 times that of an uncongested connection), then it stays at that limit for additional losses or until congestion goes away .

(2) CT uses Minimum MTU of 1232 octets. The ST MTU is 1500 by default, but can be changed to 1480 if you want to get the best transfer rate possible (in some cases).

(3) CT will not use more than half of your connection speed for downloads and uploads combined. This is a serious limitation, since it means that on a 100 Mbps connection, you will only be able to download at 50 Mbps max and upload at 25 Mbps max . So even though your internet speeds say 70+ down/10 up or whatever, CT caps download and/or upload speed at around 30 Mpbs in most locations.

(4) In addition to those issues above with TCP congestion control algorithms which are applied in all other major consumer ISP's , CT also has an additional choke point in the network where all uploads from residential customers are throttled to around 5 Mbps, no matter how much bandwidth is available.

Although CT has been applied only to Comcast customers for the past couple of years, other major ISPs have used similar controls on their networks for many years prior to that, so this is not a new practice. I remember when I had cable internet back in 2000-2003 and my connection speeds were limited to 7 Mbps/1.5 Mbps down/up (no more than 1.5 MBps upload) despite paying $60 per month!

The data set includes every test run by myself and others over the past several months from multiple locations in various states around the US. So although I'm using my own CT and ST speed tests for illustration, the data set is much larger. All of the runs were made late at night to minimize any real-time issues with congestion or other factors that would affect speeds (or connections) adversely. 

Downstream Speed Tests

For download speeds, I have collected over 100 runs from multiple people on DSLReports.com , with most of them showing a median value of 20 Mbps down on CT vs 30+ Mbps down on ST. The spread in reported results shows a very large variation in download speed between CT and ST, but does indicate that the average median difference is around 2x slower than ST speed when downloading. The graph below shows an example of one location's data set which shows this gap in download speed:

Variation from median is shown by the red circles on the graph. Two small (10%) outliers are marked by yellow squares. The average difference from the ST median value should be about 2x, but generally is larger than that for most people's data sets which I have seen (including my own) 

The variation range of values reported is astonishingly large for some locations! It's important to note that it does not seem possible to predict exactly what your downstream speeds will be when you connect online via CT versus ST. There seems to be too many variables involved upstream from your ISP connection (e.g., at interconnection points with other ISPs) and within your local network/equipment (e.g., router/modem, cable modem, switch ports on equipment).