When Pipelined Streaming I/O is selected and no cyclic prefix is used, the core can overlap the loading of a frame with the processing and unloading of earlier frames. If the upstream master supplies the first symbol for a new frame immediately after the last symbol for the previous frame, the core starts loading it immediately.
The following figure shows the general timing for back-to-back frames in the Pipelined Streaming architecture.
There is a latency between a frame being loaded and the processed data for that frame being available. This latency depends on the options chosen in the Vivado IDE to parameterize the core. However, when that latency has passed, processed frames appear back-to-back.