I have this set to yes and I am running a CEP task on each core and I am not getting any messages regarding the bandwidth.

Each and every time BOINC downloads a file, the client keeps track of the size of the file, and how much time it takes to download the file. After finished downloading, the speed is calculated as file_size / download_time. At the end, the new speed is weighted together with the old average, and this is the new average download-speed.

One of the reasons the measured download-speed is low, is that WCG on most tasks uses many small download-files, and downloading a 1 KB-file often gives less than 10 KB/s in download-speed, and these small files therefore drags-down the average.

Other reasons is that either the WCG-server, or somewhere between WCG-server and you there's something that slows-down the transfer.


You can see it as a car-analogy, there "Speedtest" is basically to put your car on a closed track, there no other cars interferes, and measure the peak speed the car can make, example 200 km/h.
A peak of 200 km/h on the other hand has little meaning if the car in it's day-to-day usage most of the time sits in so heavy traffic that it's crawling-along at 10 km/h...

Also, no cars will go instantaneously from zero to full speed, it will always take some seconds to accelerate. So, if a car starts from zero and drives 1 km with no futher stops, it will maybe use 30 seconds on this km (120 km/h).
If on the other hand a car starts from zero and drives 10 m, for so stopping and driving another 10 m and so on, the car will take much longer to drive 1 km, even if you do stop the clock for each time the car stops. So, chances are you'll use multiple minutes driving 1 km if you'll need to re-start the car every 10 m.