To compare the data generated at the LHC with the standard model, it is necessary to perform HEP simulations. These simulations can be done in various steps: event-generation, digitization, etc. To understand which architecture is most suitable for certain kinds of workloads, we need to benchmark these workloads for the various experiments at CERN. My assignment involved benchmarking these workloads, especially MadGraph, on both CPU and GPU. The objective was to determine the relationship between energy consumption and performance on different architectures.