Two approaches can be taken when implementing a GPGPU design: chip-down and mezzanine. Each path has its own characteristics that affect the suitability of that implementation for a particular application. This white paper examines these trade-offs to aid a system designer in making the right choice for their program.