Technologies for touch point detection include a computing device configured to receive input frames from a touch screen, identify touch point centroids and cluster boundaries, and track touch points. The computing device may group cells of the input frame into blocks. Using a processor graphics, the computing device may dispatch one thread per block to identify local maxima of the input frame and merge centroids within a touch distance threshold. The computing device may dispatch one thread per centroid to detect cluster boundaries. The computing device may dispatch one thread per previously identified touch point to assign an identifier of a previously tracked touch point to a touch point within a tracking distance threshold, remove duplicate identifiers, and assign unassigned identifiers to closest touch points. The computing device may dispatch one thread per block to assign unique identifiers to each unassigned touch point. Other embodiments are described and claimed.