Multi-generation diffusion models—such as those of Norton and Bass (1987), Mahajan and Muller (1996), and Jun and Park (1999)—were developed to describe the inter-generational substitution and diffusion process of high-technology multi-generation products. Among these, Jun and Park (1999) dichotomized the sales data obtainable from multi-generation products into two types: data that can distinguish between first-purchase demand and upgrade demand (Type I) and data that cannot make this distinction (Type II), and developed a Type I model and a Type II model applicable to Type I data and Type II data, respectively. They also argued that the Norton and Bass model and the Mahajan and Muller model are applicable only to Type I data, and that only their Type II model is the sole model applicable to Type II data. In the case of durable consumer goods such as PCs, identifying whether individual consumers are making a first purchase or an upgrade requires considerable costs in the data collection process; therefore, most sales data for multi-generation durable consumer goods provided by market research agencies are Type II data. Accordingly, this study proposed a new multi-generation diffusion model that can be applied to the more generally obtainable Type II data. Furthermore, using worldwide PC sales data, an empirical comparison was conducted between Jun and Park's Type II model (i.e., JP2) and the new model proposed in this study (i.e., NB2). In terms of face validity of parameter estimates, model fit capability, and predictive validity provided by the NB2 model and JP2 model, the empirical comparison results demonstrated that the NB2 model is superior to the JP2 model.