No performance increase in load/save after installing SSD?

I regularly save and load large (>2Gb) matrices. Using savefast gives me about a 40 fold decrease in in write times and 15 fold decrease in read times. However, there is negligible difference between saving to local SATA SSD vs HDD. System benchmarks indicate the SSD is about 2-3x the speed of the HDD. Furthermore, numpy arrays are saved at 2-3x speed with SSD vs HDD, though the baseline speed is slower.

Is there something I'm missing? Do I have to flip some preference to unlock performance gains from SSDs?

I am using R2015B running on CentOS 6.

Thank you.