The OpenACC standard joint effort represents a step forward in the way of alleviating the programming effort when exploiting heterogeneous HPC using hardware accelerators. The lack of OpenACC annotated codes is a drawback for the adoption of the standard. Before real life OpenACC codes become available, benchmarking with synthetic codes is a useful tool to evaluate compilers and the OpenACC standard itself. In this work we use the recently published EPCC OpenACC benchmark suite to evaluate available OpenACC compilers from both industry and academia. We present computational result for the benchmark using both Fermi and Kepler NVIDIA architectures.