For the airborne ultrawideband synthetic aperture radar (SAR), the motion errors caused by trajectory deviation are rangevariant. This manuscript put up with an improved rangvariant stripmap phase gradient autofocus (SPGA) algorithm to solve the problem. With the increase of resolution for airborne SAR, the size of echo data and the complexity of imaging algorithm grow rapidly and the traditional computing platform cannot meet the requirement of fast imaging. This manuscript focuses on the improved motion compensation and autofocus issues using a collaborative architecture, combining central processing units (CPU) and graphical processing units (GPU). By taking advantage of compute unified device architecture (CUDA), the improved SPGA is much more efficient and robust, thereby making it operable to work with high efficiency. Experimental results show a speedup of about 48 times compared with a nonoptimized CPUbased approach.