I see several possibilities of processing the data, if it cannot be performed on the single computer:
- just split project according to the reference data into two parts (probably, with overlap) and process the projects separately,
- align photos on the single computer, then by varying the bounding box size create two chunks based on the same alignment and bounding boxes next to each other, then you can generate dense cloud on two computers and merge the results back after both parts are generated,
- use network processing feature and both computers as processing nodes.