The packet processing performance of the general multi-core architecture is plagued by many factors, including high packet IO cost, multi-core share memory and process scheduling competition, TLB entries failure rate, etc. Therefore, a TABM (thread affinity for buffer management mechanism) based on multi-core network packet processing system, which is oriented to high-speed packet forwarding application was proposed and completed on network dedicated co-processing engine. The TABM adopts thread affinity scheduling strategy with no interrupt, sends each packet data and descriptor which contains control and buffer address information to several successive shared buffers according to the corresponding thread ID, and organizes the packets and descriptors which processed in the same thread in the form of a chain. The packet forwarding performance was tested on the basis of general multi-core and field-programmable gate array platform. The experimental data show that the average packet forwarding performance is promoted by about 12.4% and the IO cost and the TLB entries failure rate are reduced by adopting the TABM. 