This paper considers a diffusion-based molecular communication link assuming the receiver uses chemical reactions. When the signalling molecules reach the receiver, they react with the chemicals at the receiver to produce output molecules. We consider the number of output molecules over time as the output signal of the receiver. This output signal is stochastic because both diffusion and reactions are stochastic processes. If this output signal has high variance, then it can lead to degradation in communication performance. Inspired by the spatial partitioning of receptors in cell membrane, this paper investigates the effect of spatial partitioning on the variance of the output signal. By modelling the diffusion and reaction using linear noise approximation, we show that a spatially partitioned system gives a lower total output signal variance than one that does not use partitioning, especially in steady state.