the majority of those SDCArds can perform better than that (mainly read). it seems to me the boards are the bottleneck.