I think you mean the same thing: A Ferrari F1 car is excellent for the race track (but maybe not just this season) but will make a lousy shopping car. A Smart is much better suited for picking up some groceries, while keeping down the fuel costs and saving a lot of time when finding a parking spot.

An 8051 chip would probably be quite hard-pressed to handle the load of a real 3G module. Not only should it be able to send or receive 100-300kB or more every second. It must also have the RAM space for the storage, and the processing capacity to generate the outgoing data, or process the incomming data.

The next thing is that networking requires a stack. A number of nice, cheap GPRS modems have a built-in stack where you use simple AT commands to open a TCP connection and send/receive data. Some of them even has an AT layer for simulating ftp or similar. This will save a lot of cost for a TCP/IP stack, the learning time how to use the stack, and the RAM requirements needed for the stack. It will not be suitable for a high-end solution, but then again: No sane human would try to develop a high-end networking solution with an 8051 chip when ARM chips can be had for the same cost. Besides - the module will probably already contain an ARM chip way more powerful than the 8051 chip on the outside.

But Bing has a history of quickly deciding on a solution, even when there are other alternatives much better suited for the problem. It is a question of: "How do I use this module for something it isn't suited for" instead of "Anyone who can recommend a good solution for solving xx" or "Anyone have any opinions of problems or advantages with using xx for solving yy". Bing has already made his design decisions before starting the threads...