Typically, the AI Engine array runs at a higher clock frequency (between 1 GHz and 1.25 GHz) than the internal Programmable Logic. Within the AI Engine core the streaming data-width is 32-bit; whereas between the AI Engine interface tile and PL interface, it is 64-bit by default. To balance the throughput between AI Engine and internal programmable logic, it is desirable to pipeline enough data by choosing wider stream data paths for PL blocks. For example, to use an AI Engine array running at 32 bits / 1 GHz rate to its full potential, PL blocks can use 64 bits / 500 MHz rate or 128 bits / 250 MHz rate and so on. Such wider (> 32 bits) stream data is sequentialized automatically into 32-bit streams within the AI Engine interface tile.