In large cohort studies, generalized case-cohort design has been proposed when measuring exposures is expensive and events are not rare in the cohort. In such design, expensive information is collected from a random sample from the full cohort and a portion of subjects who have events. In this paper, we propose a proportional subdistribution hazards model for the generalized case-cohort design and examine optimal choice of weights to improve efficiency gain. The proposed estimators are shown to be consistent and asymptotically normally distributed. Simulation studies show that the proposed methods work well and gain estimation efficiency when using extra information from other causes.