Technical problems with shrinking process node for semiconductor manufacturing has generated considerable interest in
the use of multiple beams as an advance manufacturing technique in the context of direct write to mask / wafer. This
paper examines the data preparation bottlenecks associated with this process. The various steps in the data preparation
flow are described. Particular attention is paid to the large increase in data volume and the associated issues in
processing power, transfer speed, storage requirements, and overall turn-around-time. Further, the use of commercial
graphic processing units (GPUs) is examined as a possible solution to some of these issues and results of tests conducted
as part of the MAGIC (MAskless lithoGraphy for IC manufacturing) initiative are summarized.