Most of the time, the solution is easy, but finding the problem or underlying issues is one of the hardest challenges. I would say that they are polling the sampling device for one value and then sending that one value over the UART. This will inherently limit the sampling rate because this is the absolute worst case cycle. If they are doing the one sample and one process method, I could see how even with a fast baudrate that they are incapable of utilizing those baud rates to their fullest. The processor must cycle between the sampling device, temporary store the sample, load the value into the UART, wait until the UART is complete. A better approach would be to create an array of memory and fill that array with samples, then transfer the bulk. I am also confused by the explanation of the system's workings. A possible block diagram would greatly help in this regard. The way the system sounds is that a 16 bit value appears from some place. This 16 bit value is taken as two 8 bit values presumably by the 8-bit MCU. These values are pushed out through a UART to a PC that displays the two 8-bit values as a 16-bit value. Some extra sampling takes place on this display waveform after a form of processing has taken place.