With the data output from the LHC increasing, many of the LHC experiments have made significant improvements to their code to take advantage of modern CPU architecture and the accompanying advanced features. With the grid environment changing to heavily include virtualisation and cloud services, we look at whether these two systems can be compatible, or whether improvements in code are lost through virtualisation. We compare the runtime speed improvements achieved in more recent versions of ATLAS code and see if these improvements hold up on various grid paradigms.