This approach is quite effective in terms of both performance and memory use in the GPU architecture. In addition, GMiner solves the workload skewness problem from which the existing parallel methods suffer; as a result, its performance increases almost linearly as the number of GPUs increases. |