The present disclosure provides a data processing apparatus and method. The data processing apparatus comprises a scheduler, configured to schedule at least one processing task to a first processing unit and/or a second processing unit; the first processing unit, configured to process at least a first portion of the at least one processing task by using a first convolution parameter configured before processing the at least a first portion; the second processing unit, configured to process at least a second portion of the at least one processing task by using a second convolution parameter, the second convolution parameter being read from a storage unit in the process of processing the at least a second portion; and the storage unit, configured to store data related to the at least one processing task. In this way, the first convolution parameter used by the first processing unit is configured before the processing task, such that in the task processing process, the first convolution parameter does not need to be read or updated again from the storage unit so that data transmission can be reduced, the power consumption overhead can be reduced, and the processing energy efficiency is improved.