For GPUs to achieve their peak performance, effective and efficient usage of memory bandwidth is necessary. To this end, programmers invest extensive development effort to optimize a GPU program, specially its memory bandwidth usage. The OpenACC programming model has been introduced to tackle the accelerators programming complexity. However, this model's coarse-grained control on a program can make the memory bandwidth utilization even worse than the utilization achieved under CUDA. We propose an extension to OpenACC in order to reduce the traffic on the memory interconnection network, using a compression method on floating point numbers. We examine our method on three case studies and achieve up to 1.36X speedup.