Due to the characteristics of large bandwidth, low latency, massive traffic, and diversity in 5G communication scenarios, it has become a trend for 5G services to shift from traditional base stations to data centers. To solve the deployment problem of 5G workloads, we use the open-source OpenLTE for performance analysis as its representative benchmark. Since open-source OpenLTE has poor performance, and it cannot reflect real scenarios and behavioral characteristics in performance analysis, we first optimize its codes according to the characteristics of general-purpose processors (GPPs) and achieve the speedup of 2.5x. On this basis, the workload execution behavior is analyzed considering the characteristics of GPPs, and the analysis shows that in the physical layer, the processing flow of the 5G downlink process is computationally intensive with a maximum port utilization rate of up to 90%, and memory access is not intensive with extremely short program response time. Finally, we analyze the workload behavior in concurrent scenarios in combination with the unique parameters of GPPs (multi-core, SMT, and turbo boost, etc.) and put forward deployment suggestions for improving data center resources. It is found that 5G workloads can only run in an exclusive machine mode on account of their strong real-time nature, and there is little competition among internal concurrent bodies for shared cache and memory access bandwidth but fierce competition for execution components. Thus, the concurrency value should not exceed the number of processor cores for deployment, and the impact of TurboBoost cannot be ignored.