Solved! Go to Solution.
I looked at your code, and made a few changes. First, I couldn't tell why you were dividing the samples/ch by two for the read sound function. I disabled that. Also, the "Initial Data" you were feeding to the first "sound write" VI and at the default data input of the DataSocket Read function was a waveform cluster. The sound data sent over the DataSocket connection is an array of waveform clusters (one cluster per channel). This probably was causing DataSocket problems for ya.
I set the sound configuration values to what I found in the "Continuous Sound Input.vi" example VI. I also added a few charts for visual indications.
Note that you must have DataSocket Server running (on windows: Start->Programs->National Instruments->DataSocket->DataSocket Server).
If these don't work, I'd look at network/firewall issues. Also, if you want less latency, consider using straight-up UDP. It should have less overhead than DataSocket (and with a crossover connection, I'd guess you won't have much data loss to worry about).